@SLICK MA=1 74 1 , TABS=1 3 , WWS=1, IWT=0, ST=0, IN=0, WC=A-Za-z0-9_$, Release 19.2.11, July 22nd 2006 - Built by Steve Rance ====================================================== Changes Since 19.2.10 --------------------- [rls] 19:35:00 13 July 2006 - SRC mode in setup file AES2 cards now default to SRC off. SRC is not required for plug-ins and for AES2 cards which are used instead of AES1 cards. If you need to use Sample Rate Conversion with AES2 cards you will have to enable SRC. Edit /nvr/setup and add "SRC = YES" to the the [Options] section like this: [Options] SRC = YES When the system boots you will see the following lines printed I/O Resources: ... 2 AES2 cards (SRC is enabled) Bugs Fixed Since 19.2.10 ------------------------ None Release 19.2.10, June 30th 2006 - Built by Steve Rance ====================================================== Changes Since 19.2.09a ---------------------- None Bugs Fixed Since 19.2.09a ------------------------- [mcc] 18:20:00 30 June 2006 - mdrplay bus trap in Overwrite/Tape modes After reducing the number of tracks (by opening or importing from a project with less tracks than the current project), a bus trap could occur if the Record Mode was Overwrite or Tape. [mcc] 13:00:00 30 June 2006 - MediaLink drive capacity A 32-bit overflow bug caused MediaLink server to incorrectly report the drive capacity for volumes with 4294967296 or more sectors (2TB with 512-byte sectors). A more fundamental limit of 4TB still applies. Volumes larger than that may still work correctly but the displayed capacity will not be right. [rls] 16:00:00 26 June 2006 - license file not found A fresh install of 19.2 software will print the following error: "license file NOT Found - MFX will not operate" This is because it was looking for the file /dd/license.191 instead of /dd/license.192, to work around this problem in 19.2.09a or earlier, create a dummy file named /dd/license.191 Release 19.2.09a, June 13th 2006 - Built by Steve Rance ======================================================== Changes Since 19.2.08a ---------------------- [rls] 14:15:00 8 May 2006 - 60Hz video refresh rate The video output refresh rate is normally 75Hz which is incompatible with some LCD monitors. The refresh rate can be changed to 60Hz by setting PXY DIP Switch #B8 on. This switch is located at the back of the PXY and is at the bottom of the 16 switches behind a small panel. The 60Hz refresh rate can be temporarily tested by using the command "cg5refresh". The command "cg5refresh -r" reverts back to 75Hz. [mcc] 18:45:00 16 Apr 2006 - more info in free space debug The debug messages printed when drive free space is obtained now include the file system, total capacity, sector and cluster sizes. This makes it easy to check the cluster size of FLFS drives. Bugs Fixed Since 19.2.08a ------------------------- [rls] 18:54:00 20 Apr 2006 - MDR record metering The disk recorder meters on armed inputs were only detecting negative peaks, positive peaks were being ignored. This can cause the meters to read low with unsymetrical waveforms. MDR playback meters and all FMC meters were OK. [mcc] 12:15:00 19 Apr 2006 - netman read skip overflow A bug in netman code to recover from lost FLFS_READ packets could corrupt memory, with potentially fatal results. The bug occurred whenever netman printed tvt message "WARNING: read skip overflow". Now the message says ERROR not WARNING and corruption does not occur. This fault only manifested with a severely malfunctioning network. [mcc] 18:45:00 16 Apr 2006 - cluster size units The cluster size of a drive as returned from FSLIB and dfnserver was in bytes for some file systems and in sectors for others. Now it is always in bytes. This caused backup/restore/copy to sometimes overestimate destination disk space consumption. Release 19.2.08a, April 11th 2006 - Built by Steve Rance ======================================================== Changes Since 19.2.07a ---------------------- None Bugs Fixed Since 19.2.07a ------------------------- [rls] 14:30:00 11 Apr 2006 - Distortion in 96/48/48/48 Two of the 48 busses in 96/48/48/48 were distorting, fixed by redistributing the meter resources. [rls] 9:31:00 11 Apr 2006 - EQ resource loss Fixed a bug created in 18.2.22a, sometimes the EQ/DYN resource could be lost on some tracks or lives causing a break in the signal path. This happened after creating link groups in a certain order. [mcc] 11:35:00 6 Apr 2006 - Satellite 96-tk input patching When there are no more than 64 inputs of any one type installed, the Bank A/B selector no longer applies to inputs for the Arm/inputs 'patch' and 'group' operations. This fixes a problem suffered by users with 48 or less inputs when trying to patch tracks 49-96. The behaviour when there are more than 64 inputs is unchanged. Release 19.2.07a, April 4th 2006 - Built by Steve Rance ======================================================= Changes Since 19.2.06a ---------------------- None Bugs Fixed Since 19.2.06a ------------------------- [mcc] 12:00:00 31 Mar 2006 - serial driver Fixed anthem console-type support. Only one level-6 interrupt per mouse event. Increased 422 receiver efficiency. [mcc] 11:00:00 19 Mar 2006 - PY/TCS interrupt mask mgt Fixed bugs in mgt of CPU interrupt mask by PY/TCS software. One of these caused the obscure "add/subDMfrac failed" tvt messages which indicated a malfunction in setting up the quarter-frame interrupt. Another caused an extremely rare bus trap in TBASE. Release 19.2.06a, March 16th 2006 - Built by Steve Rance ======================================================== Changes Since 19.2.05a ---------------------- [mcc] 09:20:00 4 Mar 2006 - EMU 68K sets write lower bound After vector space is initialised the 68K issues SETWLB cmd (10) to set write lower bound for bus trap to end of vector space (0x400). [mcc] 09:20:00 4 Mar 2006 - Anthem The Anthem product and console types are now distinguished by the QDC. The "anthem" startup script should be used to launch the QDC software. ATOW it has only two configs: 48/48/48/48 and 96/48/48/48. The console ID should/must be 4 for EMU 68K. [mcc] 12:25:00 3 Mar 2006 - SCSI timeout error msgs SCSI Driver Timeout error messages now include the port address, from which the offending bus & drive can be determined. Network DMA Acquistion timeouts are no longer reported as SCSI timeouts. Bugs Fixed Since 19.2.05a ------------------------- [rls] 13:00:00 16 Mar 2006 - TSB DMA resync Added TSB DMA resync code to large mixer configs only (6 and 8 QDCs) to work around the problem of the TSB DMA getting out of sync after a locate. The TSB phase is read from the TSB CPLD and used to force the DMA to the correct phase at every sample interrupt. [rls] 14:00:00 3 Mar 2006 - Direct Out for Return feeds Enabled Direct Out patching for Return Feeds [mcc] 13:50:00 3 Mar 2006 - faulty FATAL cmd In Dream consoles, the 68K was not sending the correct command code to FMC to signal a fatal error. Release 19.2.05a, March 1st 2006 - Built by Steve Rance ======================================================= Changes Since 19.2.04a ---------------------- None Bugs Fixed Since 19.2.04a ------------------------- [rls] 17:30:00 Feb 29 2006 - allocate fxgroups Reversed the order of processing feeds when allocate_fxgroup is retrying. This means Tracks will be done first and high numbered Lives last. If there is a shortage of EQ resources a high numbered Live will miss out. [rls] 18:30:00 Feb 28 2006 - 96/48/48/48 resource shortage There was a shortage of EQ resources in 96/48/48/48 causing some low numbered tracks to be disabled until a BLUE-STOP was done. After BLUE-STOP the upper lives would be disabled instead. [mcc] 18:10:00 Feb 28 2006 - Blue M1/M2 under super or hyper mode When Blue M1 or M2 menu is "covered" by a super or hyper mode, changes to machine status caused loss of control in editor console. When the covering mode was a mixer mode this was always fatal. All consoles were affected. Release 19.2.04a, February 14th 2006 - Built by Steve Rance =========================================================== Changes Since 19.2.03a ---------------------- [mcc] 17:10:00 13 Feb 2006 - WAV export for VSR The XPC WAV export function will now generate filenames derived from the clip name(s) if the seed name is NULL or blank, and will only export top-layer clips (used to be all layers). Bugs Fixed Since 19.2.03a ------------------------- [mcc] 13:50:00 11 Feb 2006 - timecode editing in 29.97, 23.98 The fix in 19.1.11a was incorrect, there were still residual inaccuracies. Now add and subtract are done in the normal, non-rundown rate and the result is converted back to the actual rate. [mcc] 13:50:00 11 Feb 2006 - WAV export filenames too long WAV export could generate filenames that were too long (more than 28 chars) causing Error 215 Bad Pathlist. [mcc] 13:50:00 11 Feb 2006 - Entering filenames with dots A bug was introduced in 19.1.40 that caused filenames containing dots to be sometimes mishandled, characters following the last dot were discarded. [rls] 16:30:00 8 Feb 2006 - Console over TCP Fixed bug when "External PC" and "Console over TCP" both enabled in setup. Conclt logs "Trying to connect to " message only once. Improved message from mfxload if sat or dream/stn console has wrong version. Release 19.2.03a, January 19th 2006 - Built by Steve Rance ========================================================== Changes Since 19.2.02a ---------------------- [mcc] 13:20:00 17 Jan 2006 - sony_id HDCAM additions Several HDCAM and HDCAM SR mahines have been defined in the sony_id file: set da=4 0x20 0xE0 "Sony" "HDCAM" "HDW-F500 1080i 29.97" 0x21 0xE0 "Sony" "HDCAM" "HDW-F500 1080i 25" 0x22 0xE0 "Sony" "HDCAM" "HDW-F500 1080PsF 24/23.98" set da=12 0x20 0xA0 "Sony" "HDCAM SR" "SRW-5000 1080i 29.97 & 720p 59.94" 0x21 0xA0 "Sony" "HDCAM SR" "SRW-5000 1080i 25" 0x22 0xA0 "Sony" "HDCAM SR" "SRW-5000 1080PsF 24/23.98" 0x20 0xA1 "Sony" "HDCAM SR" "SRW-5500 1080i 29.97 & 720p 59.94" 0x21 0xA1 "Sony" "HDCAM SR" "SRW-5500 1080i 25" 0x22 0xA1 "Sony" "HDCAM SR" "SRW-5500 1080PsF 24/23.98" [mcc] 10:40:00 12 Jan 2006 - REC LED REC transport key LED is completely dark when off, not dim. [mcc] 19:00:00 16 Jan 2006 - FLFS open by MFT FLFS file mgr can now open a file by MFT number by using a pathlist of the form "/sdXX/.$MFT:nnnn", where nnnn is the decimal MFT no. This can be used to copy whatever remains of a deleted file to another drive. For example: copy /sd10/.$MFT:123 /sd20/mydeletedfile.MT The flfscheck -c option is not supported on drives with clusters larger than 4KB, so this is the only way to copy deleted files from such drives. Bugs Fixed Since 19.2.02a ------------------------- [mcc] 14:30:00 18 Jan 2006 - Commit hang Commit could sometimes fatally hang the system, requiring reboot. The project file was not damaged by the bug. The fault was introduced in 19.1.13a. [mcc] 11:00:00 18 Jan 2006 - Super Trim coverage bug Trim menu operations with Super Trim enabled sometimes incorrectly left the selected clips with the artificially high "popped" coverage layer. Trim and Slip operations are supposed to always leave clip layers unchanged. The "pop" when Super Trim is enabled is only temporary. [mcc] 13:20:00 17 Jan 2006 - ML import bug Importing from ML file with more than N tracks yielded no clips at all if there were no clips on tracks greater than N (where N is current Tracks setting: - 24, 48 or 96). [mcc] 19:00:00 16 Jan 2006 - makdir stack overflow The "makdir" command sometimes crashed with stack overflow on FLFS drives. The FLFS file mgr was to blame, did not return E_CEF as expected for device root. Release 19.2.02a, December 22 2005 - Built by Michael Carlos ============================================================ Changes Since 19.2.01a ---------------------- [mcc] 12:25:00 2 Nov 2005 - Cleave stereo and mute clip for MFX/MFK On MFX/MFK consoles, Blue EQ selects a new menu containing the cleave stereo and mute clip softkeys, migrated from Dream-family. When there is a From/To range, top-layer clips that are entirely included within the range are selected (red). Double-clicking mute clip toggles the muting of the selected clips. [rls] 12:20:00 27 Oct 2005 - Extra Track names Tracknames for tracks 65-96 are now sent to the Mixer. [rls] 12:20:00 27 Oct 2005 - Extra Meters Increased number Of Meters from 64 to 88 in 96_48_48_48 [rls] 17:20:00 20 Sept 2005 - New FxModes for Anthem Added two new modes for the FxMode of a FEED: inputmonfx: the track FX in is patched to the input source not the MDR out mdrtoreturnfx: RETN input is patched to the corresponding track's MDR out Bugs Fixed Since 19.2.01a ------------------------- [mcc] 17:00:00 21 Dec 2005 - Boot ROM stack overflow The PX dynamic boot ROM did not have a large enough stack to reliably boot from FLFS drives. This was addressed in the kernel in 19.1.43/19.2.01, but the ROM also needed more stack. The symptom of the fault was that a previously well behaved system would suddenly fail to boot. The boot sequence is now a little more verbose and reports errors better. [mcc] 17:00:00 14 Dec 2005 - DSP restart problems Fixed bugs in the DSP restart procedure when QDC h/w faults are present: Audioman sometimes crashed after QDC sections failed to start correctly. Auto-restart following a DSP error could cause the console to malfunction. In some cases the DSP restarting would automatically repeat forever. A small memory leak occurred on every restart, 1-8K depending on configuration. This was worst in editor-only systems. Since the DSP restart occurs after every ODIF operation the leak could accumulate to become serious. [mcc] 17:00:00 14 Dec 2005 - two second delay after Wave menu op An unnecessary two-second delay after most Wave menu operations has been fixed. [mcc] 11:30:00 6 Dec 2005 - console restart problems Restarting the editor console while engine is sending serial data to it (eg while transport in motion) could provoke a stream of 422 protocol errors that took a long time to clear. This was particularly bad in Dream family consoles. Now all consoles drain the 422 input port until first packet-start code is received, and Dream family consoles use a two-stage handshake. These changes increase the chance that editor console will reboot correctly if you are forced to restart it. Dream family editor indicates startup progress with some messages in the LCD. If it starts normally then most of these will happen too quickly to observe. But if fails to start then the message displayed can help diagnose the problem. Starting... - means program is alive and the LCD works Connecting... - establishing 422 link to engine Name & s/w version - 422 rx purged and ready stage 1 - waiting for 422 handshake stage 2 - handshake received, waiting for init data and ack stage 3 - ack received, initialising menus stage 4 - starting MDR, selecting Project menu [mcc] 17:45:00 2 Dec 2005 - ML import problems 1) In mixer systems, FMC was being notified as soon as ML import began rather than at the end of the import process. This created a deadlock situation that could timeout if the import took too long, with unpredictable results. 2) When importing ML files containing marks, a bizarre bug could crash Satellite. [rls] 14:52:00 25 Nov 2005 - MIDI TC out bug With the latest batch of SIO2 cards the MIDI timecode generator was failing causing TCS TIMEOUT after pressing GEN. [rls] 15:09:00 18 Nov 2005 - sio2 diagnostic bug A bug in the sio2 diagnostic program caused midi test failure with the latest batch of SIO2 cards. [mcc] 12:35:00 17 Nov 2005 - kernel tvt debug msg termination Some kernel modules were incorrectly terminating tvt debug messages. This could potentially cause system instability or a crash. [mcc] 18:00:00 11 Nov 2005 - AB2 search for apostrophe Putting an apostrophe (single-quote) into AudioBase2 search form could cause the QDC to fatally hang, requiring reboot. This has been fixed. However you still cannot search for a single apostrophe in AB2, you will always get No Match Found. A work-around for AB2/SQL is to use two apostrophes. For example, to search for "joe's" type "joe''s". Alternatively, upgrade to AB3 server which has no problem with apostrophes. [mcc] 12:50:00 9 Nov 2005 - restore from tape, no files A bug was introduced in 19.1.18a that caused restore from tape to fail as if there were no files on the tape. Pressing restore again usually revealed the files. [mcc] 12:05:00 8 Nov 2005 - ML import converting more than 4GB ML import did not correctly detect in advance the case where more than 4GB of audio must be converted and copied. It displayed the wrong total in the popup dialog, and proceeded to copy until it ran out of project space after 4GB and then failed. If ML import reports "Too Much Audio to Convert" it means the ML file references too much audio to convert and copy into the 4GB project. If ML import reports "Not Enough Disk Space" it means exactly that, delete some files and try again. [rls] 11:55:00 22 Sept 2005 - ftp server fixed Changes to the startup file in 19.1.43 caused the ftp server to not work. If you are running 19.1.43 or 19.2.01a then a temporary work around is to run the OS9 command 'load shell' before attempting to ftp to the QDC. ========== 19.2.01a BRANCHED FROM 19.1.43 ============= Release 19.1.43, September 11th 2005 - Built by Steve Rance =========================================================== Changes Since 19.1.42 --------------------- None Bugs Fixed Since 19.1.42 ------------------------ [mcc] 12:45:00 31 Aug 2005 - increased system stack The system stack size for a process has been increased from 8K to 12K. The FLFS file mgr may have been suffering system stack overflow with certain disks (depending on depth of dynamic file system data). To compensate for memory consumption, many rarely-used modules are no longer loaded at startup, for a net gain of about 200K. [mcc] 07:55:00 30 Aug 2005 - more stack issues espdeldir: stack overflow when deleting a tree deeper than about 5 levels. bdd: system-state program did not have enough stack. Release 19.1.42, August 25th 2005 - Built by Steve Rance ======================================================== Changes Since 19.1.41 --------------------- None Bugs Fixed Since 19.1.41 ------------------------ [rls/mcc] 15:39:00 25 Aug 2005 - SCSI cache crash A fatal exception can occur when starting "SCSI Caching Subsystem" Fixed by increasing the stack size for scct. [mcc] 14:25:00 16 Aug 2005 - v3.04 diskinit -c option for FLFS The range-checking of the -c=sectorsPerCluster option was incorrect and did not allow a value less than 8 sectors. The correct limit is based on the sectorsize: a cluster cannot be less than 4KB. [mcc] 17:05:00 5 Aug 2005 - quit with project open In Dream-family mix systems, Quit with a project open usually caused MIXER DEAD FORCING CLOSE error and left the embedded mix data unflushed. This was caused by shutting down QDC software components in the wrong order and has been fixed. Release 19.1.41, July 26th 2005 - Built by Steve Rance ====================================================== Changes Since 19.1.40 --------------------- [rls] 15:00:00 22 Jul 2005 - setup documentation improved Revised the Help section of the setup file and added some end of line comments. Removed unused [Surface] SERVER settings. Default HOSTNAME is now QDC and DOMAINNAME now FAIRLIGHT. [mcc] 17:25:00 21 Jul 2005 - sony_id Panasonic AJ-D580 and Fostex D-30 DAT have been added to sony_id file, thanks to GTN, Detroit. [mcc] 17:25:00 21 Jul 2005 - mixint, CLIENT_PORT The mixer interface mode (Dream or fame) is no longer controlled by the presence or absence of [Surface] CLIENT_PORT in the setup file. Now it is never necessary to modify (comment or un-comment) this line in the setup file, for any product. To be specific: fame and prodigy ignore the CLIENT_PORT setting altogether. In all other products CLIENT_PORT is optional. Port 2801 will be used unless another port has been defined with CLIENT_PORT. This fixes a serious performance degradation that occurred when Satellite or mfx3.48 systems had CLIENT_PORT undefined. [mcc] 15:00:00 16 Jul 2005 - AB3 When AudioBase Version is 3 the URL format is /AudioBase/MFX/*.php. Bugs Fixed Since 19.1.40 ------------------------ None Release 19.1.40, July 15th 2005 - Built by Steve Rance ====================================================== Changes Since 19.1.39 --------------------- [mcc] 14:10:00 15 Jul 2005 - Log files A log of fatal exceptions is maintained in /dd/USR/LOGS/crash.log. If a "Red X" Process Terminated error occurs, information about the event is written to crash.log. If crash.log grows too large it is renamed oldcrash.log and a new log is begun (at system startup). A new command called "log" writes a tvtlog with time-stamped name to /dd/USR/LOGS. These logs are never deleted automatically. To delete them you must use "del" command, or just type "log purge". [rls] 15:30:00 8 Jul 2005 - Multiple MADI cards Removed aman_cfg variable MADI96K_24PER. Replaced with automatic setting, now when there is more than one MADI card the number of channels per MADI card is changed to 48 (at 48K) or 24 (at 96K), to make patching easier. [mcc] 16:24:00 7 Jul 2005 - V3.03 upgrade Upgrade no longer installs the user-defined patches file by default. Upgrade no longer uploads MFX & Merlin console software, this is done automatically when the application first starts up. Bugs Fixed Since 19.1.39 ------------------------ [mcc] 14:10:00 15 Jul 2005 - M1/M2 Sony Arm bank Recent changes caused Arm Trks to revert to A1-4 every time the menu was selected. [rls] 17:40:00 5 Jul 2005 - MADI outputs dropout bug Groups of four MADI outputs were dropping out after a wave menu operation. This is fixed by moving the switching of the SSI speed to earlier in the DSP startup sequence. Release 19.1.39, June 17th 2005 - Built by Steve Rance ====================================================== Changes Since 19.1.38 --------------------- [rls] 13:20:00 15 Jun 2005 - Enabled recording on Pyxis With a Pyxis 9-pin slave machine online, entering Record mode on the QDC will also put the Pyxis into record. Previously, the Pyxis would go in to play. Using the Pyxis controls to go into Record puts the QDC into Play but not Record mode. [mcc] 10:40:00 8 Jun 2005 -- Folder sorting in browser The sorting order of folders in the file browser has been restored to what it was before 19.1.09a, specifically: Folders are always sorted by name. The user-selected sort rule and direction applies only to files. The last-edit date is no longer displayed for folders because it is not a reliable or useful value. The created date is still shown. [rls/mcc] 13:48:00 03 Jun 2005 - Pyxis Video Track Arming The video track on a Pyxis can now be armed from the Blue M1/M2 menu using the new "Arm Video" softkey (Dream family only). Bugs Fixed Since 19.1.38 ------------------------ [mcc] 09:45:00 14 Jun 2005 - ROM cfginq not found Fixed error regarding cfginq when booting from dynamic ROM. [mcc] 10:40:00 8 Jun 2005 - Cancel Arm/inputs ops in Satellite Input patch UI was left in a confused state if the patch, group or gain operation was cancelled by an unexpected key, even a lower softkey in the same menu. [mcc] 11:30:00 30 May 2005 - Blue M1/M2 track arming Various minor but confusing bugs in Blue M1/M2 track key/LED mgt: Tracks could be "armed" on the LTC and 9P Rem device types. MFX/MFK console rarely updated LEDs with real arming status. In Dream consoles the LED is now red for an armed track, not green. [rls/mcc] 17:54:00 26 May 2005 - MTC and mixer TC bugs Fixed several bugs in the generation of Midi Time Code (MTC) and the Mixer Time Code. Time code is now generated precisely on the correct frame boundary. [rls] 15:32:00 25 May 2005 - 9P Rem / TC Gen offsets Using 9P Rem was overwriting the TC generator offset. 9P Remote does not have any offset control. [mcc] 09:40:00 20 May 2005 - fades in ML4 export A bug is fixed that may have affected 3rd-party ML4 readers, but did not affect the QDC itself or AVTransfer. A waveseg object with no fades still had fadein and fadeout specified. Release 19.1.38, May 19th 2005 - Built by Steve Rance ===================================================== Changes Since 19.1.37 --------------------- None Bugs Fixed Since 19.1.37 ------------------------ [spr] Thursday May 19 16:23:20 EST 2005 - SCSI Select Timeouts A timeout error noticed with some newer Seagate drives (eg. ST373207) caused mediad to take these newer drives on and off line about every 10 seconds. This has now been fixed by making the SCSI Select Timeout parameter saved in NVRAM common to both the ROM and the runtime SCSI drivers. The value hard-coded into the kernel driver was 7. A value of 8 or the default 9 seems to work. Release 19.1.37, May 13th 2005 - Built by Steve Rance ===================================================== Changes Since 19.1.36 --------------------- None Bugs Fixed Since 19.1.36 ------------------------ [mcc] 12:45:00 13 May 2005 - TAIL GT END Sometimes clips were recorded with a sync point that caused an error message such as this to be displayed: TRK 1 TAIL GT END tail:00:00:03:11/1198 end:00:00:03:11/1196 The message appeared at drop-out, and thereafter everytime the EDL integrity is checked by the software (eg when project is opened). The fault only occurred when Arm/Handle not zero, at sample rate <= 44K1. The fix applies only to new recordings, not to existing clips. If you have valuable projects that produce large numbers of these errors then contact your distributor. [mcc] 14:45:00 12 May 2005 - Blue 5 & 6 The track display sometimes crashed with divide-by-zero exception when more than one Blue number-pad digit was pressed at the same time. [mcc] 14:45:00 12 May 2005 - mdrplay Bus Trap A bus trap in mdrplay is fixed (probable cause of May 5 Red X crash at GTN). Release 19.1.36, May 11th 2005 - Built by Steve Rance ===================================================== Changes Since 19.1.35 --------------------- [mcc] 17:25:00 9 May 2005 - 96 inputs in 96tk Satellite In 96tk Satellite, the selection of inputs for patch & group operations depends on the Bank Select key. Select Bank B to patch inputs 49-96. Bugs Fixed Since 19.1.35 ------------------------ [mcc] 11:28:00 9 May 2005 - Spurious MDR File Server Queue Full A bug could occasionally cause a two-minute stall in MDR file IO, followed by the error: MDR File Server Queue Full. Before 19.1.34 the error would have been Flight - Semaphore Timeout. [mcc] 11:20:00 7 May 2005 - Track,Import,Bounce to 49-96 Tracks 49-96 could not be selected as destination for bounce and some Track and Import menu operations. The bank selection was ignored. [mcc] 09:20:00 6 May 2005 - MediaLink mlfam Not Responding Mlfam.exe did not correctly handle Windows LOGOFF event. [mcc] 09:45:00 30 Apr 2005 - No marks in new project A new project created after changing the Tracks count had no chronological mark list. Marks could not be created until the project was closed and re-opened again. [mcc] 16:40:00 29 Apr 2005 - More MediaLink netman recovery bugs Failure to recover from certain network errors could cause netman to return a shorter-than-requested read count. This would cause dfn_backup to terminate a file transfer (backup/restore/copy) and report "Read Error". Dfn_backup will no longer terminate on a short read. Also, the filesize is reviewed before transfer commences in case modified by 3rd party software. Release 19.1.35, April 22nd 2005 - Built by Steve Rance ======================================================= Changes Since 19.1.34 --------------------- [mcc] 11:30:00 20 Apr 2005 - DFNServer debug option For consistency, the DFNServer z option now has same syntax as other apps. That is, more z's give more vebosity, eg -zzz. [mcc] 15:50:00 15 Apr 2005 - New error message The new error message Transport Control Timeout is returned instead of the generic Flight IPI Timeout when TCS fails to respond. Bugs Fixed Since 19.1.34 ------------------------ [mcc] 15:50:00 15 Apr 2005 - MediaLink netman recovery bugs When reading from remote file, bugs in error recovery methods could sometimes cause corruption of data and/or corruption of memory due to buffer overflow. These critical bugs manifested with unreliable network, and caused a range of different symptoms in the QDC. Network IO functions now timeout after 20 seconds (40 seconds for Write). Previously the app could stall forever under certain conditions. Release 19.1.34, April 8th 2005 - Built by Steve Rance ====================================================== Changes Since 19.1.33 --------------------- [mcc] 17:50:00 7 Apr 2005 - New File Server Errors To aid in diagnosis of IO faults, the MDR file server returns specific errors instead of the obscure Flight Semaphore or Mutex Timeout errors. MDR File Server Timeout MDR File Server Queue Full Either of these errors indicates a fundamental file IO fault. For a file on local disk, a SCSI bus fault or termination error is indicated. For a remote file, unreliable network connection is indicated Bugs Fixed Since 19.1.33 ------------------------ [rls] 18:20:00 7 Apr 2005 - 9-pin bug with slow responding slave Fixed bug when receiving an unexpected report. This could just be a late arriving report. Previously this would cause M1/M2 to go offline and not be able to go online again, requiring a restart. Now a sysmesg is printed "SONY-A DEVICE: UNEXPECTED OR LATE REPORT" and it stays online. This message could indicate that the slave device is not responding in a timely manner. [mcc] 17:50:00 7 Apr 2005 - WAV export lost errors WAV export failed to report errors from certain operations, acted like everything worked okay. [mcc] 14:30:00 4 Apr 2005 - 9-pin remote jammed MDR into play When M1/M2 9-pin remote online, Shuttle and Stop transport modes were internally jammed to Play for MDR device (Disk). This "hid" the end of a recording pass (the first mode other than Play after drop-in), affecting the sync of cached Gate points, the input alignment for Tape & Overwrite modes, and in some cases the bracketing for Undo. Now the jamming-to-play does not occur at the end of a recording pass. Should this cause an unforseen problem, the jamming can be reinstated by setting MDR-PLAYJAM-9PREM to 1 in tcs_cfg: @SETENV MDR-PLAYJAM-9PREM=1 [mcc] 09:00:00 3 Apr 2005 - MediaLink, Error 13 Windows ANetD server returned incorrect error code (13) when attempting to delete an open file. Now returns error 253, Non-Sharable File/Device Is Busy. Release 19.1.33, March 31st 2005 - Built by Steve Rance ======================================================= Changes Since 19.1.32 --------------------- [mcc] 12:25:00 21 Mar 2005 - Blue PlayMenu In Dream and mfx consoles, Blue PlayMenu resets the timebase and TSB, and stops the transport. This will clear certain error conditions. If you are thinking of trying Blue Stop or Blue/Blue/Blue to recover from a fault, please try Blue PlayMenu first, and see if that fixes the problem. Bugs Fixed Since 19.1.32 ------------------------ [mcc] 17:35:00 31 Mar 2005 - lax 422 protocol verification Certain errors were not being detected in the 422 packet protocol received by the (editor) console. Loss of data bytes from the serial stream would result in unpredictable behaviour. Now a FATAL PROGRAM ERROR will be reported in the LCD and (editor) console will reset itself. [rls] 12:15:00 24 Mar 2005 - Sony machine control bugs Added 23.98 FPS case to Sony remote and Sony master. Only force "Unlace off" for Pyxis instead of all "follow_slave" machines. Don't print "stray rx byte" messages unless debug enabled. This condition can occur continuously if a Sony-B and Sony-Remote are both connected due to shared transmit hardware. When "9-pin Remote" is online, there should not be any machine plugged in to the "9-Pin B" port. [mcc] 11:35:00 24 Mar 2005 - 9P Rem / TC Gen coupling There was an incorrect coupling between the M1/M2 9P Rem controller and the timecode generator. Putting 9P Rem online always enabled TC generator. [mcc] 15:12:00 23 Mar 2005 - Gateway to 0.0.0.0 If [Web]GATEWAY was undefined or set to 0.0.0.0, the setup command would install a gateway to 0.0.0.0. This had some undesirable side-effects. In this build, no gateway is installed if the value is undefined, or 0.0.0.0, or 255.255.255.255. [mcc] 12:35:00 20 Mar 2005 - MediaLink transfer between servers An old bug sometimes caused MediaLink file transfers between two servers to progress very slowly or fail altogether, especially on gigabit network. If the transfer failed due to the bug, the engine would report error #039:076 (0x274c), which is Windows error 10060. Release 19.1.32, March 17th 2005 - Built by Steve Rance ======================================================= Changes Since 19.1.31a ---------------------- [rls] 13:57:00 16 Mar 2005 - New SYNC display The Sync display has been moved from the Track display to the Top window A lot more information is now displayed about the sync system such as: Displays the measured frame rate or word clock rate when out of lock. Displays "no input" if no input is detected. Displays the actual system clock rate when locked e.g. 48048 or 47952 Hz. Displays the actual frame rate of the video/ltc source when in lock. [rls] 16:07:00 17 Mar 2005 - HD Video Sync source A new Sync source "HD VIDEO" can now be selected on the Digi menu. This selects the sync source from an external HD Video Sync Separator module (Tri-Level Sync), soon to be available. HD video rates supported are: Any interlaced format (e.g. 1080i or 1080PsF) at 23.98, 24, 25, 29.97 or 30 full frames per second. Any progressive format (e.g. 720p or 576p) at 50, 59.94 or 60 frames/s, in this case the project frame rate will be half of the progressive frame rate. [rls] 13:57:00 16 Mar 2005 - 23.98 and 24 FPS LTC Changes to the SIO Xilinx firmware means that the Sync system can now lock to 24 and 23.98 FPS LTC input in 'Master' mode. [mcc] 12:55:00 14 Mar 2005 - MediaLink file search Enhancements to MediaLink file search will give a faster search time on larger networks. [rls] 17:55:00 14 Mar 2005 - Alternate tapcfg files Some mixer configs will not start if the number of inputs (analog + digital) is too high. For example "dream96 48/8/20" will only work with a maximum of 64 inputs. Rather than physically removing AI or AES cards you can do this: cd /dd/usr/sys copy tapcfg tapcfg.dream_hs48_8_20 (whatever config you want to override) ed tapcfg.dream_hs48_8_20 Now comment out (with a *) some of the AI or AES lines like this: *@qdc 1: mw4-ai1 1 0; save the file This reduced number of inputs will be automatically used for the config specified by the file name. Bugs Fixed Since 19.1.31a ------------------------- [mcc] 17:55:00 17 Mar 2005 - mixint crash after ODIF and blue-stop A narrow race window has been closed that could cause a fatal exception in mixint process following ODIF (Wave menu) operation or Blue-stop. [mcc] 11:50:00 15 Mar 2005 - duplicate cookie confusion Two problems have been fixed that could cause the wrong file to be located when there are files with duplicate cookies on the network. [rls] 16:46:00 9 Mar 2005 - Another LTC Output bug fix Fixed LTC output alignment in PLAY mode. The LTC frames were late by 10% of a frame time. Release 19.1.31a, March 9th 2005 - Built by Steve Rance ======================================================= Changes Since 19.1.30a ---------------------- [rls] 12:30:00 7 Mar 2005 - mfx3 mixer comms Enabled mixer comms for all mfx3 configs. This enables RAPID to be used with mfx3/48 consoles. [mcc] 10:30:00 5 Mar 2005 - MediaLink server max files The theoretical maximum number of files that can be open simultaneously on the server has been increased from 128 to 4096. [mcc] 07:30:00 4 Mar 2005 - mlfam no watch for write MediaLink Folder Watch does not watch for file write, only for create, delete and rename. Reduces the overhead required for the watching. [rls] 15:50:00 2 Mar 2005 - added 96/48/32/72 Added new config 96/48/32/72 for 8 QDC cards Bugs Fixed Since 19.1.30a ------------------------- [rls] 16:16:00 4 Mar 2005 - LTC Output bug fixes Fixed bad LTC output at 96kHz sample rate. Fixed generation of LTC at play speed > 2x and play speed < 0.1x - the LTC output now reverts to normal frequency as originally intended. [mcc] 19:25:00 3 Mar 2005 - DSP track dropout fix revised The 19.1.28a DSP track dropout fix has been revised to avoid some objectionable side effects in playback buffer mgt. [rls] 12:16:00 2 Mar 2005 - 96kHz mixer jogging bug Fixed bug when jogging with 96kHz mixer configs (e.g. dream96). All the tracks would mute requiring a blue-stop to fix. [rls] 15:44:00 2 Mar 2005 - distortion in dream 48/48/64/62 fixed distortion on some busses in dream 48/48/64/62 Release 19.1.30a, March 1st 2005 - Built by Steve Rance ======================================================= Changes Since 19.1.29a ---------------------- [mcc] 14:00:00 1 Mar 2005 - AudioBase3 Support connection to AudioBase3 server. The number of clips returned in each search batch from AB3 server is 250 by default, and can be set in the setup file: [AudioBase] BATCH = nclips. [mcc] 07:25:00 1 Mar 2005 - MediaLink Folder Watch MediaLink2 server can optionally "watch" certain folders for changes. This means there is no need to "refresh" these folders before using files created by other applications, for example AVTransfer. This is a beta release of this feature. It can be enabled by editing the dfnserver.INI file (Properties). In the [appargs] section, add or edit the line for mflam.exe. List the pathlists of the folders to watch. For example, to watch the folder "X:\Incoming\ML4": [appargs] dfnserver.exe = anetd.exe = zeromon.exe = mlfam.exe = X:\Incoming\ML4 More than one folder can be watched mlfam.exe = X:\Incoming\ML4 Y:\OMFDialog To also watch sub-folders within the selected folders, use "-s" option: mlfam.exe = -s X:\Incoming\ML4 Y:\OMFDialog In this release you cannot use -s to watch an entire volume: mlfam.exe = -s X:\ << This will not work If there are spaces in a pathlist it must be quoted: mlfam.exe = "X:\Incoming Files\ML4" "Y:\OMF Dialog" After starting the server apps, inspect the MLFam window to confirm the folders being watched and check for any error messages. Folders can be watched on a mapped network drive, provided the remote host is a Windows machine. On an engine running a software release earlier than 19.1.30a, new files in a watched folder will not appear automatically in the file browser if the folder is already selected. It is necessary to browse away from the folder and then return to it. So it may seem as though "refresh" is still necessary but it is not, all that is required is to leave and return to the folder. Bugs Fixed Since 19.1.29a ------------------------- [mcc] 07:25:00 1 Mar 2005 - meter bank switching The switching of A/B meter banks was broken in 19.1.23a. [rls] 15:16:00 25 Feb 2005 - jog -> play bug with Pyxis Fixed bug when jogging at 1x speed with a Pyxis slave. Both the Pyxis and QDC could unexpectedly switch from jog to play mode. Release 19.1.29a, February 21st 2005 - Built by Steve Rance =========================================================== Changes Since 19.1.28a ---------------------- None Bugs Fixed Since 19.1.28a ------------------------- [spr] - Febraury 21st 2005 - Broken Builds Due to a typo in two makefiles, builds 19.1.26a thru 19.1.28a should be discarded and never used. This current build, 19.1.29a and subsequent has been fixed. Release 19.1.28a, February 17th 2005 - Built by Steve Rance =========================================================== Changes Since 19.1.27a ---------------------- None Bugs Fixed Since 19.1.27a ------------------------- [mcc] 16:25:00 11 Feb 2005 - DSP track dropout A DSP bug could cause loss of playback audio, recoverable only by blue-stop. The auto-recovery from an overload condition only worked once. The overload condition was most likely to occur in configurations that use the mdr6 DSP program. Release 19.1.27a, February 3rd 2005 - Built by Steve Rance ========================================================== Changes Since 19.1.26a ---------------------- [mcc] 10:30:00 2 Feb 2005 - Cannot Import, more information When attempting to Import from an incompatible project file, the error message now includes the filename, track count and file revision. Note that you cannot import from a file that has more tracks than the current project. [mcc] 18:25:00 25 Jan 2005 - muted clips in Dream only In mfx3, merlin, fame and prodigy, muted clips in a project file are played and displayed as normal, as if they were NOT muted. This can be changed by setting the MUTECLIP environment variable to 1: setenv MUTECLIP 1 Then clips will be muted as they are in Dream family products. Bugs Fixed Since 19.1.26a ------------------------- [rls] 18:15:00 28 Jan 2005 - re: meter data enhancements Fixed a bug in the previous build which caused a bus trap in audioman. Release 19.1.26a, January 24th 2005 - Built by Steve Rance ========================================================== Changes Since 19.1.25a ---------------------- [rls] 14:08:00 24 Jan 2005 - meter data enhancements Always use the extended meter data packet format for DREAM products, this gives one more bit of resolution to the dream meters. Also, rounded the meter values up slightly to ensure that a 0dB sine wave generated by recdata sends the exact value. Bugs Fixed Since 19.1.25a ------------------------- [mcc] 12:30:00 24 Jan 2005 - file gopher bus trap on empty drive A bug introduced in 19.1.23a caused file gopher to suffer bus trap when a drive has no files (eg, just after diskinit). A work-around for earlier builds is to create a single file on any empty drives. The "touch" cmd can be used for this purpose. For example: touch /sd20/junk.txt Then reboot. Release 19.1.25a, January 24th 2005 - Built by Steve Rance ========================================================== Changes Since 19.1.24a ---------------------- [mcc] 11:45:00 Jan 14 2005 - mute clip The mute clip softkey presents a sub-menu with ON, OFF and TOGGLE softkeys. Pressing one of these changes the muting of the selected clips and closes the sub-menu. To just close the sub-menu, press mute clip softkey again. Double-clicking the mute clip softkey is a shortcut for TOGGLE. While the sub-menu is active, the track and range keys may be used to change the selection. The All Layers key applies for the muting of clips. [mcc] 13:10:00 Jan 12 2005 - file created date The file created date is shown in the file browswer. Because of the additional column, the desired Sort setting may need to be re-selected. [mcc] 15:45:00 Jan 11 2005 - setup for console with external PC To configure QDC for Dream console with external PC, add or set EXTERNAL = yes in the [Surface] section of /nvr/setup: [Surface] EXTERNAL = yes If "setenv MFXCONSOLE /modem422" has been added to /nvr/profile, then it is no longer necessary, it can and should be removed once the setup file has been updated. [mcc] 15:45:00 Jan 11 2005 - muted clips The muted clip shading is displayed for any number of tracks. Bugs Fixed Since 19.1.24a ------------------------- [mcc] 16:25:00 13 Jan 2005 - missing Merlin shifted features Fixed missing features caused by lack of Shift key in Merlin. To toggle Nudge Sync on or off, use double-click Nudge (press-and-hold Blue, double-click Fade/Level key). To set Level/Mode key default, use Blue-softkey. [mcc] 16:25:00 13 Jan 2005 - muted-clip display Muted-clip shading sometimes overwrote hunks of the clip name. This is fixed. Release 19.1.24a, December 21st 2004 - Built by Steve Rance =========================================================== Changes Since 19.1.23a ---------------------- [mcc] 12:40:00 21 Dec 2004 - Razor fill When Razor is on, the fill and b/f fill operations insert the range to be filled. The softkey labels say "insert" to indicate this. [mcc] 12:40:00 21 Dec 2004 - Mute clip Clips can be muted, causing them to play as silence. Muting is an undoable edit. Following Wave menu operations the source clip is automatically muted. Muted clips are indicated in the track display by a half-tone "darkening" of the clip. Caveats: Muted clips cannot be seen when more than 24 tracks are displayed. The prototype UI toggles the mute status for the selected clips, a method will be devised to explicitly mute or unmute. Bugs Fixed Since 19.1.23a ------------------------- [mcc] 11:05:00 19 Dec 2004 - Import/load-mix range The Import edit range is no longer cleared when library project is closed. Thus fixing a bug in load-mix. Release 19.1.23a, December 15th 2004 - Built by Steve Rance =========================================================== Changes Since 19.1.22a ---------------------- None Bugs Fixed Since 19.1.22a ------------------------- [rls] 16:07:00 10 Dec 2004 - VU meter bug when buss insert enabled Repatch phase meter and VU meter when buss insert changes [mcc] 14:20:00 8 Dec 2004 - AB publish File Rig projects Clips in external sub-projects were not published when the master project was published. Now they are. [mcc] 10:15:00 14 Dec 2004 - FLFS write error vulnerability A write error when flushing FLFS system files could cause a secondary fault. If the first available MFT was corrupted, FLFS would not create a new file, always returned error 1. Now such damage is automatically repaired. Release 19.1.22a, December 7th 2004 - Built by Steve Rance =========================================================== Changes Since 19.1.21a ---------------------- [mcc] 20:00:00 7 Dec 2004 - Gate parameters interactive The Gate menu parameters take effective immediately when changed. The Enter key is not necessary, hence the softkey LEDS do not blink. [mcc] 20:00:00 7 Dec 2004 - Meter window fixes & changes In 48-meter view the meter bars are 40% wider. Armed tracks are indicated by red track numbers, not red input names. Thru mode tracks are indicated by amber track numbers to match console LEDS. Fixed graticule-straddle bug when track N*8 armed in stereo: the graticule labels are displaced if necessary to keep stereo pairs in a group of meters. [mcc] 13:20:00 1 Dec 2004 - File Rig The frig command has been upgraded so as to be useful for merlin/mfx users. It is no longer possible to start recording before all files have been created. After changing the file rig setting, the UI blocks until any necessary files have been created. After opening an already-rigged project, the UI will block when arming a track until any necessary files have been created. When waiting for files to be created, the system status line says "RIGGING FILES - PLEASE WAIT". The File Rig status window can be explicitly displayed with Esc G, or Blue-G. Note that when MDR is shutting down a short delay may occur while unused sub-projects are being deleted. Bugs Fixed Since 19.1.21a ------------------------- [mcc] 12:39:00 3 Dec 2004 - Export ML4 Bad Pathlist When the project file name is the maximum length (28 chars) then exporting to ML4 caused Bad Pathlist because the 3-char suffix makes the name too long. The file name is now truncated by one character in this case. Release 19.1.21a, November 29th 2004 - Built by Steve Rance =========================================================== Changes Since 19.1.20a ---------------------- [mcc] 14:20:00 26 Nov 2004 - Space menu File Rig softkey The File Rig can be set on the Space menu. The value is the max number of tracks recorded to one file, or OFF for normal operation. When the key is active a window shows the meaning of the setting. Sub-projects are now created on demand. New projects are always created with File Rig OFF. [mcc] 07:50:00 23 Nov 2004 - AB2 server port AB2 HTTP server port number can be set in setup file or shell environment. /nvr/setup: [AudioBase] PORT = portnum Shell environment: setenv ABPORT portnum Bugs Fixed Since 19.1.20a ------------------------- None Release 19.1.20a, November 22nd 2004 - Built by Steve Rance =========================================================== Changes Since 19.1.19a ---------------------- [mcc] 11:40:00 22 Nov 2004 - AB2 HTTP fixes and improvements URLs were not correctly terminated, caused IIS/V6 servers to barf. Improved detection and reporting of HTTP errors from server. The ctrl-Q key will terminate an AB2 transaction if the QDC seems to be hung. There is now a timeout when waiting for reply from server, default 180 secs. The timeout can be changed in the setup file or shell environment. /nvr/setup: [AudioBase] TIMEOUT = secs Shell environment: setenv ABTIMEOUT secs Bugs Fixed Since 19.1.19a ------------------------- [mcc] 14:35:00 22 Nov 2004 - file rig stereo When file rig was 2, stereo clips were recorded in the master project. Release 19.1.19a, November 19th 2004 - Built by Steve Rance =========================================================== Changes Since 19.1.18a ---------------------- [rls] 12:13:00 18 Nov 2004 - bus group names Bus groups can now be defined via an xpc call to aman_defgroup(). MDR now just reads the bus group names from aman global structures. Increased number of Aux busses to 36 and MT busses to 48 [mcc] 19:20:00 17 Nov 2004 - File rig Sub-projects are created and deleted by a new process, mdrfcs. It is possible to start recording before all the necessary sub-projects have been created. So don't do that. It is not yet possible to un-rig a project. Keep-borrowed does not import from sub-projects. [mcc] 19:20:00 17 Nov 2004 - Library window The Library window (Esc L) now shows the complete pathlists for open attached files. Previously there was no convenient way for the user to determine exactly what files had been found by MediaLink search. If a pathlist is too long, intermediate folders are replaced with "..". An 'S' is shown in the status column for external sub-projects (file rig). The default view filter is now ALL (see Setup menu, LIBRARY FILE LIST FILTER). To use the Library window, select a Binnacle menu (or any non-browsing window), type "Esc L" on qwerty kbd, use jogger or plus/minus console keys to scroll up and down the list. Bugs Fixed Since 19.1.18a ------------------------- [rls] 18:30:00 15 Nov 2004 - license file Fixed a bug in the license.xxx file which could cause a hang if the machine is unlicensed. Release 19.1.18a, November 8th 2004 - Built by Steve Rance ========================================================== Changes Since 19.1.17a ---------------------- [mcc] 14:00:00 3 Nov 2004 - gfxram process filter In gfxram (Esc R) right-clicking a block sets a filter that causes blocks not belonging to the same process to be grayed-out. It also "locks" the clicked block, same as left-click. To clear the filter and the lock, click again. [mcc] 14:00:00 3 Nov 2004 - pmem The pmem diagnostic tool reports the memory allocation for a process. [mcc] 17:50:00 1 Nov 2004 - browser enhancments Postponed folder refresh When browsing up & down in the folder pane, the refreshing of the file pane is postponed until you stop. This prevents browser "stall" caused by fetching the contents of each folder. The file pane displays "Loading..." when a refresh is pending. Shift/Ctrl Arrow in Dream consles The shift & ctrl keys can be used to browse by drive or node. shift-arrow: browse by drive (next/previous disk) ctrl-arrow: browse by node (next/previous machine) blue-arrow: up/down page These work with focus in either folder or file pane. Blue-arrow is not new, but little known. [mcc] 13:00:00 29 Oct 2004 - FFS RAM size FFS RAM colour size reduced from 4MB to 2MB. Developers that make-install this change must blowall as advised by sysinfo. Bugs Fixed Since 19.1.17a ------------------------- [rls] 12:20:00 8 Nov 2004 - audioman Return feeds above 56 weren't working and were interfering with tracks because of an 8-bit wrap around bug, now that there are 96+96+96 possible feeds. Also reduced MAXMIXOUT from 128 to 72 to save memory. [mcc] 15:20:000 5 Nov 2004 - FLFS FM max file size bug When the space allocated to a file equalled 4GB the clusters became orphaned - deleting the file did not return free space to the drive. The cause was the failure of FLFS to limit file size to the max possible size, which is 4GB minus one cluster. Fairlight project files were not affected by this bug because they are never larger than 4GB-128KB. [mcc] 04:30:00 27 Oct 2004 - gfxram The gfxram diagnostic tool (Esc R) had a bug that could crash the system. Even if no crash, the information displayed was not always correct. Release 19.1.17a, October 25th 2004 - Built by Steve Rance ========================================================= Changes Since 19.1.16a ---------------------- [rls] 15:15:00 22 Oct 2004 - 96 lives and 96 returns 96 Lives and 96 Return feeds now supported. Need to test for Fame compatability. [mcc] 12:30:00 22 Oct 2004 - frig The "frig" command allows the project file rig to be set for testing/demonstration of writing to attached projects. It has one parameter, the number of tracks per file: 2,4,8 or 12. Example: Create new 48-track project. Go to Project menu. Type the command "frig 2". When it finishes you will see that 24 project files have been created. These are attached to the current project and enabled for writing. Now arm 48 tracks and start recording. You can now record for over 6 hours at 16-bits or 4 hours at 24-bits, assuming you have the disk space (24 files X 4GB = 96GB). [mcc] 10:30:00 21 Oct 2004 - writing to attached projects MDR project mgt layer now supports writing to attached projects. Previously attached files were always read-only. This allows waveforms to be created and written (recorded) to the attached files. There is not yet a UI method to enable the feature. But it has been tested, successfully recording to 32 files in parallel. Bugs Fixed Since 19.1.16a ------------------------- [rls] 15:15:00 22 Oct 2004 - 8 QDC engine Decoupled DSP TSB I/O for 6 and 8 QDC mixers to fix distortion problem. [mcc] 09:00:00 11 Oct 2004 - MediaLink filename length Various fixes related to maximum MediaLink filename length. MediaLink folder and file names (including suffix) must not exceed 28 chars. Bugs in client (engine) software would sometimes cause names exceeding the limit to be created on the server. These have been fixed. The server also prevents this from happening to allow for old clients and undiscovered bugs. File gopher incorrectly allowed file names exceeding the limit (but less than 64 chars) to be included in the DFN file cache. Such files are now excluded. Alias filenames for Backup/Copy/Move were too long when original name longer than 25 chars. An alias name is generated if two files are marked for transfer to the same file (eg, mark /sd00/fred.MT and /sd01/fred.MT for copy to server). Also, the popup dialog to confirm the alias was not presented for Copy or Move, resulting in a destination name like fred001.MT. [mcc] 09:00:00 11 Oct 2004 - MediaLink readdir Windows ANetD readdir service did not return entries for "." and ".." in device root directories. This caused programs that always skip over these entries (like os9 dir) to miss the first two files. Release 19.1.16a, October 1st 2004 - Built by Steve Rance ========================================================= Changes Since 19.1.15a ---------------------- [mcc] 09:00:00 23 Sep 2004 - group dimensions MDR bus group definitions expanded to 72 busses, 66 groups (used to be 48 busses, 42 groups). [rls] 18:30:00 22 Sep 2004 - 7 and 8 QDC config Added dream 48/96/96/64 configs for 7 and 8 qdc cards. 8 QDC cards use MDR4 instead of MDR6. 7 QDC config is subject to test results. Increased MAXLIVE and MAXRETN to 96. This needs more testing. Bugs Fixed Since 19.1.15a ------------------------- [mcc] 13:45:00 21 Sep 2004 - Import/load mix The "load mix" softkey did not work correctly, always reported obscure error "DreamMix Not Found". Now the protocol is correctly implemented, but must be activated by FMC (ObjectEnable, A=1). Release 19.1.15a, September 20th 2004 - Built by Steve Rance ============================================================ Changes Since 19.1.14a ---------------------- [mcc] 13:05:00 20 Sep 2004 - range width The range width is shown with subframe resolution when diplay of subframes is enabled. Bugs Fixed Since 19.1.14a ------------------------- [mcc] 13:05:00 20 Sep 2004 - path table extensions for os9 Path table extensions were not deallocated and system path no.s were not closed when MDR exited, resulting in OS resource leak when quitting/restarting without rebooting. [mcc] 13:05:00 20 Sep 2004 - FLFS FM large clusters Oversights in FLFS file mgr caused excessive memory consumption for disks using new large-cluster formatting. MDR could not open 255 file attachments for project due to "No Ram Available" error. Now only about 12.5KB is typically required for each open file, regardless of cluster size. Release 19.1.14a, September 14th 2004 - Built by Steve Rance ============================================================ Changes Since 19.1.13a ---------------------- None Bugs Fixed Since 19.1.13a ------------------------- [mcc] 06:20:00 14 Sep 2004 - too many open objects A bug introduced in 19.1.13a broke mixer-object interface resulting in error "Too Many Open Objects". Release 19.1.13a, September 7th 2004 - Built by Steve Rance =========================================================== Changes Since 19.1.12a ---------------------- [steve] 15:54:00 7 Sep 2004 - QDC Symbol Files Symbol files for the QDC binaraies are no longer included in the release file [steve] 15:54:00 7 Sep 2004 - sysinfo sysinfo now will prevent MDR from starting if there is not enough free space remaining on /dd. [mcc] 14:40:00 7 Sep 2004 - free space on boot drive More temporary disk space is required on the boot drive than in earlier releases. There must be at least 345MB free on /dd. [mcc] 14:40:00 7 Sep 2004 - reclaim PX memory Various changes to reclaim some PX memory. Among these the size of SCSI cache has been halved to 64 entries per drive. [mcc] 14:40:00 7 Sep 2004 - diskinit v3.03 The new FLFS formatting uses a larger default cluster size. For typical 512-byte sectors: disk size sectors per cluster 64GB+ 128 32GB+ 64 16GB+ 32 8GB+ 16 less than 8GB 8 Note: the new FLFS formatting, introduced in 19.1.11, produces a more efficient file system, more suitable for large capacity drives. Less memory and CPU time are consumed for each mounted drive. However the new format is not compatible with any mfx3 or QDC software older than 19.1.11, and so diskinit prints an alarming warning message. This warning can be ignored if you do not intend to move the drive to a machine running older software. [mcc] 09:30:00 27 Aug 2004 - 96-track projects, more clips and waveforms 96-track projects can have 16384 waveforms and 32671 clips. Projects created with earlier 19.1 will expand on demand, if sufficient space remains in the file. A substantial amount of PX memory has been consumed by this change, regardless of actual numbers of tracks/clips/waves. It is not yet known if sufficient memory remains for general use. [mcc] 09:30:00 27 Aug 2004 - more clip info The clip tail time and waveform number(s) are displayed in the Takes window and (Goto) Clipname window. The head and tail times are displayed with subframe remainder (if enabled by Blue-Subf). The Clipname window has more lines. Bugs Fixed Since 19.1.12a ------------------------- [mcc] 12:50:00 26 Aug 2004 - Wave window delays Extensive delays of a minute or more were caused by Wave window (Space menu) under these conditions: Large number of clips and waveforms in project. Few or no unused waveforms. Waveform list filter option set to NOT USED (this is the default). The problem can be avoided by setting waveform list filter to ALL. Release 19.1.12a, August 19th 2004 - Built by Steve Rance ========================================================= Changes Since 19.1.11a ---------------------- [mcc] 08:25:00 9 Aug 2004 - Edit Mark/Macro name The behaviour of 'name' key in Edit Mark & Edit Macro menus has changed. Rather than clearing the input field, the current name is presented for editing, like similar keys in other menus. For the Mark menu, the new behaviour is optional (to avoid impact to user macros, keystroke habits, or mfx console remote serial interface). Pressing Blue-name or Shift-name toggles the new behaviour on or off. [mcc] 17:15:00 3 Aug 2004 - 24bit WAV export format QDC normally exports 24bit WAV files in Microsoft extensible format. Now the alternative 24bit PCM format can be used instead, by setting the WAV24PCM environment variable to non-zero. For clients that want the PCM format, edit their /nvr/profile file as shown: ed /nvr/profile setenv WAV24PCM 1 Save and close the file by typing Esc Z (for mfx consoles: Blue-space Z). Reboot the QDC. [rls] 16:30:00 16 July 2004 - VU meter adjustment Added temporary +9.5dB gain to the VU meters because FMC does not yet have a VU reference setting menu. The 0VU reference level is now approx -11.5dBFS. It is also currently pre main fader. Bugs Fixed Since 19.1.11a ------------------------- [rls] 15:40:00 20 Aug 2004 - Sony_B default behaviour Changing the "default" settings in the sony_id file had no effect on an "Unknown" machine on the Sony_B port. Now it works for both Sony_A and Sony_B. [mcc] 12:00:00 18 Aug 2004 - Audition & Super-Trim while Undo/Freeze on When Undo/Freeze ON, AudioBase Audition and Super-Trim clip/head/tail caused the project EDL to revert to the frozen state, as if Undo had been pressed. Edits done while frozen were lost, no Redo was possible. Following Super-Trim the EDL could also become corrupted producing error messages like "HEAD LT BEGIN..." or "TAIL = HEAD...". Immediately pressing Undo will repair that corruption. Undo/Freeze is now turned off before rather than after these operations. [mcc] 10:43:00 12 Aug 2004 - AB2 edit-all System would sometimes crash or become unstable following AudioBase2 edit-all (edit, Blue-Enter). [rls] 14:50:00 28 July 2004 - Wrong names in MDR for MT busses When MT busses above MT5 were patched to tracks, the names were displayed incorrectly on the MDR meters and the MDR patch page. [mcc] 15:42:00 21 July 2004 - Serato Dialog distortion Wave menu pitch shift operations using Serato Dialog algorithm produced two kinds of distortion: Gross distortion when reducing pitch using Scale 2 or 3. Crackles in left channel of stereo clip. The gross distortion has been corrected (in 19.1.11a). To avoid the stereo crackles, pitch-shifting of stereo clips is no longer permitted when using Serato Dialog algorithm. Release 19.1.11a, July 16th 2004 - Built by Steve Rance ======================================================= Changes Since 19.1.10a ---------------------- [mcc] 15:20:00 13 Jul 2004 - shadow key in Track menu The shadow softkey has been re-instated in the Track menu of Dream family consoles. [mcc] 18:18:00 27 Jun 2004 - Setup menu displays Blue S The option selection window (known as Blue S) is now displayed by the Setup menu. [spr] 14:26:19 22nd June 2004 - ROM Built Device Descriptors The Device Descriptors built by the ROM (/dd,/SCSI0,/SCSI1,/SCSI2,/SCSI3) are now built with the options as specified by the NVRAM settings. Before this change the SCSI options were only set up by the device descriptors built by mediad. [spr] 14:14:12 19th June 2004 - Limit cluster size to 8 for non R&D systems The experimental change to the cluster size has been limited to R&D systems only. This mod will be reverted once more testing has been done to ensure it works as expected. [rls] 17:00:00 17 June 2004 - VU metering for Constellation The Constellation main VU meters can now be driven digitally instead of the current analog method. This frees up two analog outputs from the engine. The integrated VU metering is not as accurate as analog because the meter values are updated only 30 times per second. [spr] 15:19:00 25th May 2004 - FLFS cluster size It is now possisble to specify the cluster size of an FLFS disk. This is experimental so should only be used on test systems not production systems. diskinit can now be used with the -c and -w options. The default is 8 (4 KB clusters) and can be as high as 128 (64 KB clusters). [spr] 16:39:00 17 May 2004 - Changes in the Dynamic ROM re: Interrupts The dynamic ROM now prints out and detects the Interrupt status from INTGLUE. If a situation is detected that will prevent the kernel from booting the system will do a hard reset. This is the same as if the reset button was pressed. The mods to do this require IntGlue revision 48. [rls] 18:00:00 14 May 2004 - Front panel reset feature in IntGlue Rev 48 The engine can now be reset by double-clicking the front panel power button. If the button is pressed twice within a 0.5 second period the engine will be reset instead of powering down. If the PXY has stopped then two presses of the button will reset it regardless of the time interval. Resetting a crashed engine allows the TVT log to be captured. [mcc] 14:35:00 14 May 2004 - Blue Undo menu In Dream family consoles, Blue Undo and Blue Redo present a menu with Freeze and Follow softkeys. Freeze locks the undo point so that all subsequent edits (until Freeze turned off or press Undo) will be undone/redone as a single operation. Follows causes Undo/Redo to restore transport position and track selection. As in MFX/MFK console, the restored track selection only applies until Undo or Redo key is released. [rls] 11:00:00 7 May 2004 - More async XPC functions Added 2 extra async functions: aman_edit_group_a and aman_set_feed_eq_trim_a Bugs Fixed Since 19.1.10a ------------------------- [mcc] 15:20:00 13 Jul 2004 - timecode editing 29.97 The plus key did not correctly increment frames when editing 29.97 timecode with subframes turned off. [mcc] 14:20:00 12 Jul 2004 - quit and restart with less tracks Quitting and restarting the application with less tracks could potentially cause the system to crash or become unstable. Using sat as an exmample: sat 48 quit sat 24 [mcc] 11:19:00 9 Jul 2004 - duplicate cookie confusion It is reasonably well-known that duplicate cookies may confuse MediaLink file searches. In fact the problem was far worse than ever imagined. An old bug caused the filename to be ignored for most searches, only the cookie mattered. If duplicate cookies were present, the application software would be likely to access the wrong file, regardless of filename. The bug was in dfnserver for both engine and server software. Most affected were: Re-finding attached files when opening a project. Finding files returned by AudioBase search. Finding files referenced by cookie and filename (not pathlist) in ML file. [mcc] 18:10:00 7 Jul 2004 - Wave menu VARI op used Serato Wave menu VARI operations used Serato transform if selected for STRETCH and PITCH. This was inadvertently left in place for test purposes and should have been removed. Now VARI uses the standard SRC transform (much faster than Serato). [mcc] 14:20:00 6 Jul 2004 - track display SYNC field refresh The refreshing of SYNC field in track display window was broken in 19.1.04a. [mcc] 09:30:00 4 Jul 2004 - FLFS bitmap When flushing the cluster allocation bitmap to disk, FLFS did not correctly round the transfer to sector boundaries. When scanning the bitmap for freespace, only a multiple of four bytes was scanned. [rls] 18:15:00 2 Jul 2004 - Left/Right swap playing stereo WAV backwards Fixed bug when playing stereo WAV files backwards. Left and Right channels were swapped and right channel clicked periodically. [rls] 14:45:00 30 Jun 2004 - DSP crash playing 24bit stereo WAV files Fixed DSP crash when playing (jog/shuttle) a 24 bit Stereo WAV backwards in stereo mixed out mode. [rls] 11:45:00 28 Jun 2004 - Distortion playing 24bit stereo WAV files A bug in the DSP code caused low level distortion/buzz when playing a 24 bit Stereo WAV file. [mcc] 18:18:00 27 Jun 2004 - MFX & Merlin filesort Fixed filesort default and menu item on Blue S for MFX & Merlin. The menu no longer applies for Dream products. [mcc] 18:18:00 27 Jun 2004 - mediad malfunction Mishandling of SCSI options caused mediad to malfunction at startup. [mcc] 15:30:00 25 Jun 2004 - Windows node daemon timeouts Too short timeout caused Windows version of node daemon to falsely report nodes as offline. Introduced random delay between nodes like os9 version. Increased default timeout for known (online) nodes to 5 secs (was 250ms). New options for dfnserver allow timeout periods to be adjusted: -k=n RPC timeout for known node (in ms) -r=n RPC timeout for unknown node (in ms) [mcc] 14:20:00 11 Jun 2004 - FAT32 Memory mgt bugs in irbf and iformat could destabilise and crash the system during the mounting, use and/or initialisation of FAT32 drives. The bugs could even cause corruption of other drives. Also fixed inability to format certain drives, error 214. [rls] 16:40:00 9 Jun 2004 - Input levels The input gains were reverting to 0dB after a blue stop. MDR now doesn't reassert the input levels after blue-stop in a mixer system. Only aman knows the correct levels if they were changed via XPC. This was introduced when FMC changed to sending input levels via XPC. [mcc] 11:05:00 4 Jun 2004 - mediad SCSI hang When mediad setting SCSI options on boot drive collided with process using /dd the SCSI bus could hangup. A workaround has been implemented that should make this much less likely to occur. A more rigorous solution is still required. [mcc] 11:47:00 31 May 2004 - flfscheck errors Memory mgt bugs in flfscheck caused various errors depending on disk content. File system errors could be reported when there weren't any, or the system could crash or become unstable, with everything returning Error 171 (System Data Structures Damaged). [rls] 16:00:00 27 May 2004 - ROM Diagnostics The Static ROM diagnostics were failing due to changes in IntGlue Rev 48 [mcc] 13:05:00 25 May 2004 - locked-up graphics Possible fix for locked-up graphics symptom (GFXLOCK TIMEOUT error). Lists shared by the Track, Takes and Eq windows became "stale" when the current track is not displayed, resulting in infinite loop. [mcc] 17:58:00 20 May 2004 - Spurious Edit Mismatch errors Edit Mismatch errors were mistakenly reported when project closed with Undo/Freeze still ON after performing some edits. [mcc] 17:57:00 20 May 2004 - cosmetic graphics bugs In Device info window the FREE TIME field overflowed when >= 1000 hours (~322GB @48K/16bit). Now displays only hours if >= 100 hours. In File browser window the pixmaps were sometimes incompletely erased. [rls] 12:00:00 18 May 2004 - Boot halting at start of Dynamic ROM When some engines were upgraded to IntGlue Rev 47 they could fail to boot, stopping at the beginning of the 'Dynamic' ROM. Now that IntGlue has been upgraded to revision 48 it is possible that this problem may still exist. If the engine freezes during bootup with a totally black screen and with four green LEDs (& no red LEDs) showing on the PXY rear panel, then try the following work around: Remove the small access panel covering the LEDs (by removing 4 screws) Turn on DIP switch A8 (bottom switch of the top group of 8) If the machine now boots correctly, you will see a red "Yes" after "Bypass Reset Instruction #A8". Note that there may be side affects of using this mode. Please report this condition for further debugging. Release 19.1.10a, May 6th 2004 - Built by Steve Rance ===================================================== Changes Since 19.1.09a ---------------------- [mcc] 18:13:00 5 May 2004 - cleave stereo New "cleave stereo" softkey splits stereo clips into two mono clips. In this version the waveform is always copied (like blue paste). A later version will avoid this if copy not necessary. [rls] 14:10:00 1 May 2004 - Async XPC functions Added async versions of 15 existing aman functions by adding "_a" suffix. More metering points for busses now available. Bugs Fixed Since 19.1.09a ------------------------- [mcc] 07:50:00 4 May 2004 - Import/keep sync error Under certain specific conditions, Import/keep from another project file would yield incorrect sync point for the imported clip. This occurred only if the clip was originally borrowed from the current project. This can happen if the file is an extension of the current project, or if it contains clips from AudioBase that were published from the current project. [mcc] 13:40:00 30 Apr 2004 - ML import bugs Sometimes audio was needlessly being copied/converted. In particular, 24-bit mono with odd start offset was copied instead of borrowed from the source file. Certain completely incompatible audio formats were not being detected. The copy/conversion would proceed to conclusion but yielded either gross distortion or total garbage. The popup dialog now warns of the presence of incompatible formats. [rls] 18:10:00 29 Apr 2004 - MDR2 jogging backwards Fixed bug in reverse jog/shuttle for sat/mfx with 16 or less tracks per QDC card. Even numbered tracks were muted when jogging backwards. This bug introduced in 17.4.10a and 17.3.12a Release 19.1.09a, April 28th 2004 - Built by Steve Rance ======================================================== Changes Since 19.1.08a ---------------------- [mcc] 16:48:00 21 Apr 2004 - mix-follows-edit V1 compatibility The QDC XPC client interface for mix-follows-edit is now compatible with V1 XPC server (FMC 3.1, build 374 or earlier). This removes the incompatibility described in 19.1.05a release notes. The QDC tries to connect using V2 and if that fails reverts to V1. The objectionable delays are eliminated only if FMC has new V2 interface (FMC 4.0 or later), otherwise the delays will remain. [rls] 16:30:00 21 Apr 2004 - new dream configs Added four new configs: 48_48_64_62/5+, 64_8_16_32/3+, 80_16_24_40/4+, 96_48_48_48/6+ [mcc] 14:02:00 20 Apr 2004 - file display cosmetics Cosmetic improvements to file display, less cramped folder pane, longer pathlists in marking-libraries pane. Palette adjusted for greater text/background contrast in Dream products. [mcc] 08:40:00 18 Apr 2004 - ped New feature in palette editor: the RGB colour tile to right of sliders is a drop-target. A colour copied from the target palette can be pasted onto the RGB tile. Bugs Fixed Since 19.1.08a ------------------------- [mcc] 17:58:00 22 Apr 2004 - ML parsing errors Some errors incurred by ML3 and ML4 parsers were being lost, import would silently fail with no error message. Other errors were being spuriously introduced. The combination of the two bugs could cause inexplicable failure of ML import. [rls] 12:30:00 22 Apr 2004 - Phase meter Phase meter could stop working after opening a project or after blue-stop. Fixed by re-patching it whenever the buss groups are re-allocated. [mcc] 13:30:00 21 Apr 2004 - WAV import delays Unnecessary delays occurred during search for WAV files. Each file that matched the search filename, but not other criteria, added a 250ms delay. So the more files the search rejected, the longer it took. Also, the search no longer descends into one level of folders when "Folders" is turned off. Release 19.1.08a, April 17th 2004 - Built by Steve Rance ======================================================== Changes Since 19.1.07a ---------------------- [mcc] 13:20:00 17 Apr 2004 - File display Cosmetic improvements to file display. Changed some pixmaps to be full-colour icons, others remain unchanged. Column headings and highlight box width change according to browser level. [rls] 17:30:00 16 Apr 2004 - Phase Meter New Phase Meter display for Constellation. It displays the phase correlation between the main bus left and right channels. The scale is from 180 degrees (out of phase) on the left to 0 degrees (in phase) on the right. At low signal level the display goes to the centre of the scale. Bugs Fixed Since 19.1.07a ------------------------- [mcc] 13:20:00 17 Apr 2004 - gfx bug marking attachments When marking attachments for backup/restore, the pixmaps did not scroll up and down correctly when there are enough files to scroll. [mcc] 10:10:00 16 Apr 2004 - mouse cursor Mouse cursor was drawn at wrong location when positioned close to top of screen. Release 19.1.07a, April 15th 2004 - Built by Steve Rance ======================================================== Changes Since 19.1.06a ---------------------- [mcc] 07:20:00 16 Apr 2004 - file sort The order is toggled by a separate softkey: FWD/REV. The settings are retained in nvram.dat. Bugs Fixed Since 19.1.06a ------------------------- [mcc] 07:20:00 16 Apr 2004 - file filter The displayed filter string is cleared when it should be. The UI is cancelled if editor console reboots while UI active. Release 19.1.06a, April 14th 2004 - Built by Steve Rance ======================================================== Changes Since 19.1.05a ---------------------- [mcc] 17:00:00 14 Apr 2004 - Project find softkey Project menu "find" softkey allows browser view to be filtered. To use, press find, enter filename match string that can include *,=,? wild card characters. The view is filtered as you type to include only file names that match. Press Enter (qwerty or console) to open a file, or one of the other softkeys, as normal. The view filter is cleared when you leave the Project menu. This is a prototype implementation. Some cosmetic work is still required, the nomenclature is not finalised (menu says "find", gfx says "filter"). Bugs Fixed Since 19.1.05a ------------------------- None Release 19.1.05a, April 13th 2004 - Built by Steve Rance ======================================================== Changes Since 19.1.04a ---------------------- [mcc] 07:44:00 13 Apr 2004 - FFS changes File display sorting rule can be defined on Project menu. The rule names are not yet shown correctly. Internal changes to FFS to support view filtering mean that developers who make install this build must reboot the QDC. [mcc] 08:45:00 10 Apr 2004 - V2.07 dfnserver Windows version of dfnserver will accept connections from local clients using localhost address. [mcc] 17:24:00 4 Apr 2004 - track banks Track numbers are displayed as track-within-bank. User-defined views can straddle banks. Only the tracks in the current bank will be displayed at any one time. [mcc] 17:24:00 4 Apr 2004 - 64-track projects disabled 64-track project files no longer exist. Bugs Fixed Since 19.1.04a ------------------------- [mcc] 08:45:00 10 Apr 2004 - mix-follows-edit delay Objectionable delays occurred in mix-follows-edit transactions. Fixed by making all requests asynchronous in V2.00 of FMC XPC interface. This version of QDC is not compatible with earlier versions of FMC (mix-follows-edit will not work at all). A new build of FMC is required. FMC built with V2.00 interface will remain compatible with V1.00 clients, that is, all current builds of 18.2 and 18.3. [mcc] 17:24:00 4 Apr 2004 - 32-track display Fixed new bug that could crash track display when implicit 32-track view selected. Release 19.1.04a, April 2nd 2004 - Built by Steve Rance ======================================================= Changes Since 19.1.03a ---------------------- [mcc] 02:47:00 2 Apr 2004 - track banks When implicit (not user-defined) view is selected the track display will follow the selection of track bank. The Patch and Meter windows also display the track bank. [spr] 02:43:00 2 Apr 2004 - Default Dongle Value changes. The default value for TRACKS is now 96. The default value for FEEDS is now 144. [mcc] 07:55:00 1 Apr 2004 - M1/M2 Pyxis type M1/M2 Type softkey provides selection of Pyxis A or Pyxis B. New M1/M2 status field on track display shows whether the machines are offline or online or armed. Bugs Fixed Since 19.1.03a ------------------------- [rls] 18:15:00 31 Mar 2004 - 9-Pin Follow Slave Fixed bug in follow slave mode when slave stops and locates simultaneously [mcc] 12:09:00 31 Mar 2004 - Track Sel Fixed Track Sel key for more than 48 tracks. [mcc] 18:50:00 30 Mar 2004 - disk freespace updates The freespace on the project disk displayed in Devices window was sometimes incorrect and other times was updated redundantly, introducing objectionable delays. Release 19.1.03a, March 29th 2004 - Built by Steve Rance ======================================================== Changes Since 19.1.02a ---------------------- [rls] 12:55:00 29 Mar 2004 - 9-Pin Follow Slave mode. Various improvements to 9-Pin Bi-directional control (follow slave mode) Now follows slave in FFWD and REW mode. Bug fixed going to STOP mode. Lacing parameters now default to correct values if a Pyxis is online. [mcc] 16:22:00 28 Mar 2004 - 96-track selection Track selection for editing now works correctly with 96-tracks. It is not yet possible to select tracks in more than one bank at the same time. [mcc] 09:58:00 23 Mar 2004 - 19.1.02b: mix-follows-edit Version 1.1 of FMC XPC server interface supports 96 tracks. This is not compatible with V1.0 server (any earlier build of FMC). Mix-follows-edit is thus broken until further notice! Bugs Fixed Since 19.1.02a ------------------------- None Release 19.1.02a, March 23rd 2004 - Built by Steve Rance ======================================================== Changes Since 19.1.01a ---------------------- [mcc] 07:09:00 23 Mar 2004 - 96 tracks Preliminary support for 96-track projects, in Dream family. The machine must have an R&D dongle. [mcc] 15:25:00 20 Mar 2004 - Project/Tracks softkey The "new 24" softkey is replaced by the Tracks softkey. Pressing this key cycles through the available settings for the number of tracks in a project file. The "new" softkey creates a project with the number of tracks specified by "Tracks" softkey. Thus the equivalent of "new 24" would be: press "Tracks" to choose 24, then press "new" The number of tracks specified by the startup script is now merely the initial number of tracks, not the maximum number of tracks. A project with a larger number of tracks may be opened at any time. But of course only the number of tracks specified in the AMAN configuration will be able to playback or record. For example, this will now work (previously it wouldn't): start "sat 24" open 48-track project However only tracks 1-24 will playback or record (because the DSPS remain in a 24-track configuration). [mcc] 10:13:00 19 Mar 2004 - 19.1.01e editor track selection Changes to track selection methods for essentially all edit operations in preparation for more tracks. No intentional change in behaviour. [mcc] 11:33:00 12 Mar 2004 - 19.1.01b preliminary mods for more tracks Various changes to editor console and mdrplay interface in preparation for more tracks. Bugs Fixed Since 19.1.01a ------------------------- [rls] 22:10:00 19 Mar 2004 - Lost mixer packets to TBASE Transport control packets were getting stuck in the queue between MIXINT and TBASE. The queue is now drained properly. [mcc] 16:32:00 19 Mar 2004 - QSIZE TOO BIG Fixed bug introduced in 18.2.25a that could cause DSPS to crash and be restarted due to queue overrun. [mcc] 09:58:00 14 Mar 2004 - Ctrl-Arm Ctrl-Arm (disarm all tracks) did not work in Dream family consoles. [mcc] 17:45:00 12 Mar 2004 - Spurious track selection An ancient bug could cause track selection to be incorrect if track key followed very quickly by an editor supermode or hypermode key. All consoles were affected. ========== 19.1.01a BRANCHED FROM 18.3.03a ============= Release 18.3.03a, March 11th 2004 - Built by Steve Rance ======================================================== Changes Since 18.3.02a ---------------------- [rls] 13:05:00 8 Mar 2004 - sony followslave modifier Removed SONY-FOLLOW-SLAVE variable from tcs_cfg. Replaced by new "followslave" modifier in sony_id file. "followslave" is enabled for Pyxis and Vmotion slave devices. [mcc] 13:52:50 3 Mar 2004 - 18.3.02c Developers that make install this build will need to reboot engine to start new version of FFS, due to minor FFS/console interface mods. [rls] 13:05:00 26 Feb 2004 - 9-pin timecode sense in VITC format Added variable TIMECODE_TYPE to tcs_cfg. This controls the Time Sense command sent to the slave. If TIMECODE_TYPE=2 then VITC will be requested. Default is LTC when TIMECODE_TYPE=1 or is not defined. [mcc] 15:55:21 25 Feb 2004 - MDR support for more than 64 I/O Minor changes to MDR to support more than 64 I/O of one type in Dream mixer configs. Fame & prodigy still limited to 48 I/O. In non-mixer configs the number of outputs is limited to NTRACK and the number of inputs is limited to 64 for Satellite, 48 for Merlin & MFX. [rls] 17:45:00 5 Feb 2004 - New locations for Dream metering Added new "Location" values rtin and rtout for extra metering points. [mcc] 11:32:17 2 Feb 2004 - 64-track Satellite Implemented preliminary support for 64-track project files. Currently available only for Satellite with R&D dongle. Bugs Fixed Since 18.3.02a ------------------------- [rls] 17:45:00 8 March 2004 - Noise in dream 48/48/48/48 Change to dream 48/48/48/48 to fix noise in some EQ stages (removed MSD). [mcc] 10:55:00 8 Mar 2004 - missing 48-track console mods Missing editor console changes to support 48 tracks have been corrected. In Play and Rec menus of MFX/MFK consoles only: When current track was greater than 32, the clip, head, tail & fade softkeys would use the wrong track. In clip EQ menu of all consoles: The initial track selection following EQ copy was wrong, tracks 32-48 were randomly selected, or not. EQ paste could then affect the wrong tracks. [mcc] 13:52:50 3 Mar 2004 - MDR IO limiting At startup, MDR applies certain limits to IO installation to avoid QDC software throwing exception or assert if more than expected number of IO are installed. But the limit only persisted until first blue-stop or Wave menu operation. This has been fixed. [mcc] 13:46:27 27 Feb 2004 - Leading glitch in Serato M The Serato M transform produced anomalous output level for first 40-50ms. Fixed by using 4096-sample output leader for all Serato transforms. [rls] 16:00:00 20 Feb 2004 - New IOGlue firmware. IntGlue increased to revision 47 and IOGlue increased to revision 36 This fixes a bug in IOGlue xilinx firmware which only affected a small number of PXY cards where the PY processor would fail to start. This bug was introduced in 16.5.23, 16.6.15 and 17.1 (Dec 2001). [mcc] 14:11:00 3 Feb 2004 - 48-64tk upconvert The upconversion of 48-track projects to 64-tracks was broken. Any files thus converted must be deleted. [rls] 12:28:00 22 Jan 2004 - Sony Remote track arming Fixed bug in Sony Remote mode. The arm status of tracks above 24 was not being reported correctly to a sony 9-pin master device. Release 18.3.02a, January 19th 2004 - Built by Steve Rance =========================================================== Changes Since 18.3.01a ---------------------- [rls] 13:58:00 16 Jan 2004 - 64 channel IO Increased IO limit for DREAM to 64 inputs and outputs of each type, (AO, AI, AES and MADI). Bugs Fixed Since 18.3.01a ------------------------- [mcc] 12:33:24 17 Jan 2004 - another 64-bit mask bug Fixed fatal bus trap provoked by 64-bit IO masks. [spr] cmask/imask/omask/tmask package corrected for 64 bit masks A bug existed in the mask.c package that was provoked when no unused bits were in the top long word. This has now been resolved. Release 18.3.01a, January 16th 2004 - Built by Steve Rance =========================================================== [steve[ Branched 18.2.31a Here .... 18.3.01a should be identicle to 18.2.31a Release 18.2.31a, January 16th 2004 - Built by Steve Rance =========================================================== Changes Since 18.2.30a ---------------------- None Bugs Fixed Since 18.2.30a ------------------------- [mcc] 16:52:00 15 Jan 2004 - Zeromon restart bug The MediaLink2 Zeromon driver service was not being uninstalled and reinstalled correctly. It was only started reliably once after system restart. If MediaLink2 server was stopped and restarted, the driver service might fail to restart until system was rebooted again. The symptom of this would be delays when writing to server, when expanding the size of a file. [rls] 11:28:00 8 Jan 2004 - 16 track merlin config Added "merlin 16" config which allows 16 tracks at 96k with one QDC card. Also available is "merlin 16 48" which has 16 tracks with 48 track file format. [mcc] 08:16:24 24 Dec 2003 - zeromon blue-screen crashes Windows error handling mode is restored to default when initialisation is complete. This may have been the cause of some blue-screen crashes. Release 18.2.30a, December 22nd 2003 - Built by Steve Rance =========================================================== Changes Since 18.2.29a ---------------------- [mcc] 12:00:00 22 Dec 2003 - zeromon options New -d and -x options allow drives to be explicitly included (-d) or excluded (-x) by the zeromon driver. For example: -d=EFM-P - install driver on E:,F: and M: thru P: only. -x=A-C - do not install driver on A: thru C:. Bugs Fixed Since 18.2.29a ------------------------- [mcc] 13:24:11 18 Dec 2003 - mixer timecode halting Fixed mixer timecode generation so it doesn't halt. In play MX_TIMECODE pkts are generated at 1/4 frame intervals. In stop MX_CURTIME pkts are generated every 250ms. In all other modes MX_CURTIME pkts are generated every 50ms. [mcc] 12:03:01 15 Dec 2003 - post fatal error timeouts Fixed problem discovered while testing new fatal error dialog. The system behaviour after the unexpected exit of a resident task was often worse than it needed to be - delays of several minutes could be incurred waiting for requests to the now-dead task to timeout. Implemented "reserved" semaphores to fix this and deployed them in mdrmain, mdrfs and mdrplay. Now, for example, the project can still be successfully closed even if mdrplay has crashed. Release 18.2.29a, December 12th 2003 - Built by Steve Rance =========================================================== Changes Since 18.2.28a ---------------------- [mcc] 18:09:46 9 Dec 2003 - refresh for Import & B/Up A "refresh" softkey is available in Import and B/Up menus of Dream family consoles. [mcc] 15:14:22 8 Dec 2003 - XPC Revised thread policy for rapidxpc.lib, added XPCAsync*() functions to control asynchronous request stream. Revised xpcserve to read() all available data into local buffer. [rls] 18:00:00 03 Dec 2003 - 96k 48 track mixing Added 96k 48 track config 48/8/20 on 4 QDCs - currently limited to 64 I/O (Analog + Digital) due to TSB slot limitation. 96k mixing configs have no clip based EQ (mdr3h program) Bugs Fixed Since 18.2.28a ------------------------- [mcc] 15:58:22 10 Dec 2003 - Process terminated dialog The pop-up notification of fatal exceptions on PX has been broken for some time (the dialog box that said "A Task has Exited Fatally..." did not appear when it should have). This has been fixed. The appearance of the dialog has changed, it now says PROCESS TERMINATED and shows the process name, RPC and module addresses, as well as the error message text for the exception. To test the new dialog you can use "erx -x 102". [rls] 18:00:00 03 Dec 2003 - SIO1 locking SIO1 must remain locked for 1 second before QDC cards are sent block_reset This may reduce the incidence of QDC cards becoming confused after a sync source change. Release 18.2.28a, December 1st 2003 - Built by Steve Rance ========================================================== Changes Since 18.2.27a ---------------------- [mcc] 12:14:00 1 Dec 2003 - xpcserve V1.04 Test for multiple coefficient update servers if osk X option set. [mcc] 12:14:00 1 Dec 2003 - mfx/mfk M1 & M2 menus Obsolete ViVid code removed to reclaim program space. No change to behaviour is intended. Bugs Fixed Since 18.2.27a ------------------------- [rls] 18:30:00 25 Nov 2003 - AES2 input sync not locking bug. If an AES2 channel was used for an input sync source, the system had trouble acquiring sync. This was because the AES2 channels are reset after sync is acquired. This could result in the sio1 losing lock causing sync to be re-acquired ad infinitum. Fixed by skipping the reset of the AES2 channel used for input sync. Release 18.2.27a, November 24th 2003 - Built by Steve Rance =========================================================== Changes Since 18.2.26a ---------------------- None Bugs Fixed Since 18.2.26a ------------------------- [mcc] 08:00:00 24 Nov 2003 - netio bugs Bugs fixed in Windows version of netio.lib used by FMC and dfn_backup: Socket not closed by net_close(). Potential IO buffer overrun due to malformed struct. Memory leak in net_makdir() (>256KB per dir created). Release 18.2.26a, November 21st 2003 - Built by Steve Rance =========================================================== Changes Since 18.2.25a ---------------------- [mcc] 15:21:42 17 Nov 2003 - os9 ANetD V2.05 Read/write for system paths (files on /dd) is done at high priority (806). Open/create/close/delete/makdir/setsize are always done at high priority. New options: -h=N set high priority (default 806) -p=N set normal priority (default 128) [mcc] 13:59:22 15 Nov 2003 - netio timeout In netio.lib the default timeout for open,create,close,remove is four seconds (was two). New function net_set_file_timeout() allows application to adjust timeout (in seconds). [mcc] 12:11:22 15 Nov 2003 - object handles Changed embedded object handle allocation to be cyclic. Previously it allocated the highest numbered free handle, almost always #15, thus making detection of stale handle impossible. Now handle numbers are allocated cyclically. Handle #zero is not used, to avoid any confusion. [mcc] 11:40:19 14 Nov 2003 - flfscheck V3.05 The -c (copy) option is fixed. The -f and -n options have been disabled and the requirement for R&D dongle has been lifted. Bugs Fixed Since 18.2.25a ------------------------- [mcc] 10:33:20 20 Nov 2003 - 24-bit WAV channel mask The channel/speaker-position mask in exported 24-bit WAV files was random garbage. Depending on the value, some WAV players might refuse to play a file. Now the channel mask is set to front-left/front-right(3) for stereo, or front-center(4) for mono. [rls] 17:40:42 19 Nov 2003 - dream 48/48/48/48 The DSP programs for station/dream 48/48/48/48 are now working. Fixed a bug in the TSB input buffer size which caused a DSP program to fail in this config. [mcc] 15:21:42 17 Nov 2003 - netio unexpected replies Unexpected replies were not handled correctly by some netio functions. Release 18.2.25a, November 13th 2003 - Built by Steve Rance =========================================================== Changes Since 18.2.24a ---------------------- [mcc] 12:42:19 12 Nov 2003 - MediaLink app args Command line arguments for MediaLink applications can be defined in the dfnserver.ini file. The [appargs] section allows options to be specified for each application. For example, to set dfnserver debug to level 2: [appargs] dfnserver.exe = -z=2 [mcc] 12:42:19 12 Nov 2003 - ANetD (Windows) The -x option inhibits TX critical section, for testing only. [mcc] 17:44:23 6 Nov 2003 - channel sorting Channel sorting for playback has been revised to be more efficient. Bugs Fixed Since 18.2.24a ------------------------- [mcc] 12:42:19 12 Nov 2003 - playback buffering problems Two very old problems prevented the available WRAM from being consumed effectively in some cases. 16-bit clips were typically loading only 2/3 of the available WRAM ahead of cursor. Also, other tracks stopped advancing while any track was "satisfied" (WRAM fully loaded ahead). [spr] 17:39 10th November 2003 - FLFS and Optical Media Optical media with large (2048 byte) sectors can be used with the FLFS file format. [rls] 17:00:00 7 Nov 2003 - mixer transport bug A zero length record drop-in occurred when pressing play on the Fame/Softmix. This bug was introduced in 18.1. [mcc] 17:44:23 6 Nov 2003 - error propagation by new/open project Errors during FFS phase of creating or opening a project were not propagated correctly. Symptoms include the console inappropriately entering the "Select Destination - Press Okay" state following an error. Release 18.2.24a, November 6th 2003 - Built by Steve Rance ========================================================== Changes Since 18.2.23a ---------------------- [rls] 11:38:00 6 Nov 2003 - 16 track config Added "sat 16" config which allows 16 tracks at 96k with one QDC card. Also available is "sat 16 48" which has 16 tracks with 48 track file format. Bugs Fixed Since 18.2.23a ------------------------- [mcc] 18:13 10 5 Nov 2003 - keylock bug Fixed bug in keylock method for os9. [mcc] 15:20:00 5 Nov 2003 - stereo WAV crossfade bug A fatal software error (stack overflow) was provoked by crossfades between stereo clips, when both waveforms are interleaved stereo, eg, stereo WAV files. System became unusable, reboot was required. [mcc] 09:13:55 31 Oct 2003 - races in clip EQ Race conditions in the monitoring while editing clip EQ could sometimes cause missed updates or transient updates with wrong parameters, resulting in audible artifacts. [rls] 18:13:00 30 Oct 2003 - fame 24/24/24 some dynamics not working The fame 24/24/24 config had too many rmschans per section and the dynamics didn't work on lives 21-24 and tracks 21-24. Rearranged fame24_24_24 config so that there are only 18 rmschans per section. Audioman doesn't work with more than 32 and previously there were 18 on the primary and 18 on the secondary in the same section. Added ASSERT to check if there are more than 32 rmschan resources. Release 18.2.23a, October 29th 2003 - Built by Steve Rance ========================================================== Changes Since 18.2.22a ---------------------- [mcc] 18:54:42 28 Oct 2003 - track views in MFX/MFK The user-defined track views (Blue 3,6,9) are now supported by MFX/MFK console. [mcc] 16:47:51 28 Oct 2003 - track view menu latching If any track keys are depressed when Blue-digit key is released, the menu persists until all track keys are released. This allows two hands to be used to select tracks. [mcc] 17:03:21 27 Oct 2003 - MediaLink changes MediaLink1 compatibility has been reinstated in this build. The server software must be upgraded to 18.2.23+ (or 18.1.14+). Engines running the following software versions are not fully compatible with this release and need to be upgraded as shown: not compatible upgrade to 15.9.01 15.9.02 18.1.01-18.1.13 18.1.14 18.2.01-18.2.22 18.2.23 Earlier versions (15.6, 17.3) are fully compatible with this release. There is a subtle change to licensing that may affect sites with more than one MediaLink server. For each server, the node limit (number of licensed connections) must now include the other servers. For example, a site with 8 engines and 3 servers requires each server to be licensed for 10 nodes. [mcc] 14:32:00 25 Oct 2003 - MediaLink2 process mgt MediaLink2 server process management is improved. If an application fails to start or exits unexpectedly, a dialog box pops up notifying the user. If a licence upgrade is performed while applications are running the user is informed that licence will take effect next time apps are started. [rls] 22/10/2003 TSB capacity test. Insufficient TSB capacity is now fatal during startup. This is caused by running a large DREAM config on a QDC card with 25ns DPRAM chips fitted. To find the problem QDC card run "qdcinit -v" and look for "HLev 4-25" Bugs Fixed Since 18.2.22a ------------------------- None Release 18.2.22a, October 21st 2003 - Built by Steve Rance ========================================================== Changes Since 18.2.21a ---------------------- [mcc] 17:31:20 10 Oct 2003 - track views User-defined track views are saved in /nvr/trkview and restored at the start of the next session. [mcc] 10:54:44 9 Oct 2003 - feed New unix tool "feed" for developers in $MFX4/MFX/XPC/TEST/FEED. Allows feed parameters on $DEVTARGET to be set directly. Must "rehash" after make install for first time. See feed -\? for help. Bugs Fixed Since 18.2.21a ------------------------- [mcc] 14:51:25 17 Oct 2003 - return feed EQ Various problems with EQ on return feeds fixed. [mcc] 08:51:56 13 Oct 2003 - track views, detail filter Clip details are now filtered correctly for all views: Stereo centre lines and fades are inhibited for more than 24 tracks. Clip info text is inhibited for more than 16 tracks. [mcc] 13:32:19 10 Oct 2003 - refresh empty folder Refresh failed for empty folders or units when browser focus in Files pane, positioned to a blank file. Release 18.2.21a, October 8th 2003 - Built by Steve Rance ========================================================== ===================================================================== 18.2.07a BRANCH CREATED - THIS BRANCH IS PRE USER DEFINED TRACK VIEWS THE NEW HEAD IS NOW 18.2.20a ===================================================================== Changes Since 18.2.06a ---------------------- [mcc] 16:22:01 8 Oct 2003 - user-defined track views There are three user-defined track views, selected by pressing Blue-9, Blue-6 and Blue-3. The views are initially blank (contain no tracks). To create a view, select tracks while still holding down the number pad key. There are four softkeys: clear - clears the view armed - view the armed tracks selected - view the tracks selected for editing Sorted YES - views are ordered by track no. NO - views are ordered as selected, new tracks are added to bottom of view. Caveats: Views are not yet retained, they persist only for the current session. There is not yet any way to "latch" the menu, you must continue to hold down the number pad key while adjusting the view. [mcc] 13:00:04 4 Oct 2003 - mix-follows enabled by default Mix-follows-clip editing is enabled by default, the MIXFOLLOWS environment variable is no longer necessary. However it can be used to disable the feature if desired: setenv MIXFOLLOWS 0 Bugs Fixed Since 18.2.06a ------------------------- None Release 18.2.06a, October 3rd 2003 - Built by Steve Rance ========================================================== Changes Since 18.2.05a ---------------------- [mcc] 09:17:45 1 Oct 2003 - priority reductions Reduced the priorities of some "exclusive" priority apps: audioman,mixint,xpcserve,consoled,tcs_console,gfxkey,gfxmouse. [mcc] 16:01:41 29 Sep 2003 - mixint sender Revised mixint sender to reduce CPU load. Testing of fame with HSSL transport is required to confirm that there are no undesireable side effects (none are expected). The change allows multiple packets to be sent with a single write(). Previously for HSSL, packets were sent one at a time. [rls] 17:15:37 26 Sep 2003 - dream configs with short fader paths Added start configs with short fader paths, 48/48/48/48 and 48/8/8/24 In aman_cfg changed dreamplus to dream48_48_48_48 Bugs Fixed Since 18.2.05a ------------------------- [mcc] 12:51:19 3 Oct 2003 - aman initialisation Uninitialised class members caused spurious behaviour, mostly after quitting and restarting application (because all members happened to be zero following system restart). [mcc] 15:17:00 1 Oct 2003 - aman debug sent to XPC client When aman mxco debug output was enabled it was written to XPC client on the connected socket, causing potentially fatal havoc for the client. Release 18.2.05a, September 26th 2003 - Built by Steve Rance ============================================================ Changes Since 18.2.04a ---------------------- [mcc] 09:41:37 26 Sep 2003 - XPC input patching Added qdc_input_source() to patch feed input. Bugs Fixed Since 18.2.04a ------------------------- [rls] 2003/09/25 07:47:16 SONY-FOLLOW-SLAVE machine control mode. Fixed the PLAY not locking bug with "SONY-FOLLOW-SLAVE" mode enabled. If the slave goes into PLAY when we are stopped, we will also go into PLAY To enable, set the environment variable SONY-FOLLOW-SLAVE in tcs_cfg Also, if the slave relocates while we (the master) are in STOP mode, then then we will relocate to follow it. Release 18.2.04a, September 25th 2003 - Built by Steve Rance ============================================================ Changes Since 18.2.03a ---------------------- [mcc] 11:06:24 25 Sep 2003 - mix follows undo MDR client support for stage 1 mix follows undo. [mcc] 11:41:04 24 Sep 2003 - timecode transmission FMC can disable the 500ms timecode transmission timeout by setting bit 5 (0x020) in byte[1] of timecode tally request. Once set, further timecode tally requests are not required as timecode will be sent automatically, forever. Bugs Fixed Since 18.2.03a ------------------------- [rls] 18:13:20 19 Sep 2003 - clicks in MRT Added nops to MDR trim function and removed some debug from MDR and MRT. This almost fixes a clicking problem in mixer configs. The clicks were caused by too frequent reads or writes to DRAM by the MDR trim function. Enough DRAM bandwidth needs to be available for MRT. [mcc] 14:43:20 19 Sep 2003 - config mismatch crash When FMC and QDC configurations do not match it was possible for Audioman on QDC to crash, causing the system to hang. The crash would follow Blue-Stop or Wave menu operations. The fault was due to the failure of Audioman to reject operations on non-existent feeds. Now it will return error #073:030 "Mixer Bad Feed" for such operations. Release 18.2.03a, September 18th 2003 - Built by Steve Rance ============================================================ Changes Since 18.2.02a ---------------------- [mcc] 09:22:51 18 Sep 2003 - connect to FMC/XPC MDR connects to FMC/XPC server when the first project is opened or created. This allows interface to be tested without invoking any edit functions. A message is printed to tvtlog: connecting to FMC/XPC server at Any error is ignored but reported to tvtlog. [mcc] 14:44:22 17 Sep 2003 - mix follows clips Initial version of mix follows clips editing. MDR invokes FMC XPC interface for binnacle cut,copy,erase and paste. By default this feature is inhibited, pending further development. To enable for testing you must: setenv MIXFOLLOWS 1 before starting QDC application software. The setenv can be put in /nvr/profile for convenience. [steve] Wednesday September 10 13:44:57 - Pre-Release Flag Implemented a global flag to distinguish a pre-release build. This can be used in C by including and using the MACRO is_prerelease(). For assembler include and use the macro is_prerelease. The State of the pre-release flag can bee seen with the command "globals -s". [mcc] 12;01:21 10 Sep 2003 - prohibited filename characters Six more punctuation characters are not allowed when entering filenames: ! - ; , ( ) The total set of characters prohibited from filenames is now: space * = ? | & / \ ^ ~ @ < > : \ " ' ` ! - ; , ( ) [mcc] 19:16:40 28 Aug 2003 - satmix/mfxmix 24 with one QDC With one QDC card, satmix 24 and mfxmix 24 use 24/8/16 configuration. This was changed to 24/8/12 in 18.2.02a and has now been reverted. If necessary, 24/8/12 can still be explicitly selected. [mcc] 08:07:24 29 Aug 2003 - Browse up/down repeat On Dream family consoles, the browse up/down arrow keys now repeat if held down. [mcc] 09:03:00 23 Aug 2003 - mfxquit New unix tool for QDC developers to perform "quit -y" on $DEVTARGET or any QDC host[s]. See "mfxquit -\?" for help. Bugs Fixed Since 18.2.02a ------------------------- [mcc] 12:01:21 10 Sep 2003 - delete folder on MediaLink server Folders could not be deleted from MediaLink server using the QDC console UI. The bug was in the QDC, not the server software. [rls] 18:12:10 5 Sept 2003 - SIO1 locking bug Fixed a bug in the SIO1 "retry locking" logic which was causing problems with locking to run-up and run-down rates. [rls] 11:46:10 4 Sept 2003 - clicks/static bug in dream configs Decoupled TSB DMA I/O for all dream family mdr:mr DSP configs. This seems to fix the "clicks/static when many tracks are playing" bug. [mcc] 12:57:21 3 Sep 2003 - qdclog Changing qdclog options from the ctrl-C ui no longer crashes the system! The fault was caused by inadvertent expansion of "opt" macro from ESP/options.h. [mcc] 17:44:29 31 Aug 2003 - stalled waveform display The waveform drawing could become temporarily stalled after locating or high-speed jogging to a point that already had waveform loaded. [mcc] 17:45:51 31 Aug 2003 - track display covers ABase list If ABase window is active when track display window is started, the track display width was not set correctly and covered the ABase search list. [mcc] 13:51:22 26 Aug 2003 - Import 48tk ML4 into 24tk project 48tk ML4 files could not be imported into 24tk project. The import silently failed with no explanation. Now the popup option to restrict the range of tracks is presented, the same as for ML3. [mcc] 10:32:51 26 Aug 2003 - 24/8/16 on one QDC The mr003 DSP module for 24/8/16 on one QDC would immediately crash. Now it doesn't, but testing is required to confirm that all level coefficients are being updated correctly. No other DSP module is affected by the fix. Release 18.2.02a, August 25th 2003 - Built by Steve Rance ========================================================= Changes Since 18.2.01a ---------------------- [steve] Monday August 25 19:58:05 EST 2003 - License File The license file for 18.2 is now called license.182 [mcc] 11:41:19 25 Aug 2003 - ESPMsg V12 Various improvements in the ESPMsg IPC system to support increased demands by XPC, especially asynchronous messages. Senders no longer poll (sleep 40ms & retry) when the queue is full. Instead the queue length is controlled by a semaphore. Senders wait on the semaphore with the caller's send timeout for every message. This will block if the queue is full. The receiver signals whenever the queue tail is advanced. Thus there is no retrying, the sender is only woken when the queue is no longer full, or the timeout has expired, or the receiver has shutdown. The new error #254:19 (0xfe13) "ESPMSG Queue Full" indicates a bug in the semaphore protocol, it should never occur. Note: when installed with "make install" the system will have to be re-booted. [mcc] 14:41:22 12 Aug 2003 - development menus removed Removed menus that were only for development purposes from Dream consoles. Bugs Fixed Since 18.2.01a ------------------------- [mcc] 14:45:24 22 Aug 2003 - 24/8/16 configs on 1 QDC The 24/8/16 audio configuration for 1 QDC does not work correctly so the 24/8/12 config has been reinstated for use in the meantime. The following commands may be used to select 24/8/12: satmix 24 mfxmix 24 stn 24/8/12 dream 24/8/12 [mcc] 12:37:19 16 Aug 2003 - ODIF LE input formats ODIF little-endian input format conversion was broken by changes to support 24bit WAV files. All Wave menu operations on little-endian (WAV) clips were affected by this bug. The output was a mostly silent clip with short bursts of audio every 64K samples. WAV export of little-endian clips could also be affected in some cases. ------------------------------------------------------------------------ ========================= 18.2.01a BRANCHED ============================ ------------------------------------------------------------------------ Release 18.1.06a, August 11th 2003 - Built by Steve Rance ========================================================= Changes Since 18.1.05a ---------------------- [tf] 11 Aug 2003 - MediaLink2 Extended the demo version of MediaLink2 to run 30 days. Minor enhancements re usage of message-boxes. Upgrading the licence does not shut down the server Bugs Fixed Since 18.1.05a ------------------------- [mcc] 10:31:01 9 Aug 2003 - buffer overflow in mfx console The buffer to receive changes to armed track selection was too small. Overflowed into upper LCD buffer of INPUT patch menu with unknown side effects. Only affected mfx/mfk consoles. [mcc] 16:15:22 5 Aug 2003 - removable drives locked Removable drives sometimes became locked forever, requiring drive to be switched off or system rebooted before media could be ejected. Affected QDC only, not server. [mcc] 16:15:22 5 Aug 2003 - tape cache after backup-verify Backup with verify did not purge the cache of the previous contents of the tape. [mcc] 16:15:22 5 Aug 2003 - crash after too many ejects The system locked up and became unusable after too many Backup/eject operations - about 40 or 50. [mcc] 09:52:11 1 Aug 2003 - tvt bug The tvt driver lost control when called by process ID > 99. This was partly to blame for "crash after too many ejects". Release 18.1.05a, July 25th 2003 - Built by Steve Rance ======================================================= Changes Since 18.1.04a ---------------------- None Bugs Fixed Since 18.1.04a ------------------------- [steve] Friday July 25 18:50:13 EST 2003 - Build Process The build process for 18.1.03 and 18.1.04 was broken because the start up scripts (stn fame sat etc ...) where not being uploaded to the target at all. Release 18.1.04a, July 25th 2003 - Built by Steve Rance ======================================================= Changes Since 18.1.03a ---------------------- [mcc] 14:41:53 25 Jul 2003 - MediaLink for QDC Machines with preview dongle enabled may optionally use MediaLink1 protocol instead of MediaLink2. To do this run setup and add or change the MEDIALINK item in [Web] section to be: MEDIALINK = 1 This is only possible with beta software releases. [mcc] 13:58:21 22 Jul 2003 - 18.1.03b, SysStartup Revised SysStartup handling by QDC, sequence number 63 is only returned when MDR & AMAN are ready ("All clients ready"). Bugs Fixed Since 18.1.03a ------------------------- [mcc] 11:47:19 23 Jul 2003 - 18.1.03c, fx errors at startup Before first project was opened or created, all FMC->QDC requests that target FX groups were silently failing. Now the audio system is configured at startup, rather than waiting until first project. Also, no requests will silently fail, AMAN reports all errors from service functions to tvtlog. [mcc] 17:07:22 21 Jul 2003 - The Unbuffered A bug in playback buffer mgt could cause track(s) to stop buffering for certain EDLs (it was very timing specific). It was probably introduced in 16.4.01. Release 18.1.03a, July 18 2003 - Built by Steve Rance ===================================================== Changes Since 18.1.02a ---------------------- [mcc] 15:36:25 16 July 2003 - MediaLink for Windows Tidied up dfnserver.ini. Fixed bug when all volumes are library volumes. Removed MEDIALINK project from m4 workspace. Bugs Fixed Since 18.1.02a ------------------------- [mcc] 15:53:29 18 Jul 2003 - MediaLink mangled file lock list When a node goes offline, the dfnserver on other nodes is supposed to remove any file locks owned by that node. If a file was also locked by one or more other nodes then the lock list could become mangled with potentially fatal results, either immediately or sometime later. Release 18.1.02a, July 15 2003 - Built by Steve Rance ===================================================== Changes Since 18.1.01a ---------------------- [steve] Friday July 11 14:49:24 EST 2003 - Dongle The dongle system used in revision 18.1+ is no longer compatible with the revision 16/17 versions. The file name has been changed from "features" to "license" so the two different systems are not accidentally intermixed. All systems that are going to run revision 18 code must be registered with FairlightAU or else the system will fail to run as expected. [mcc] 18:43:17 9 Jul 2003 - MADI IO for mixer Mixer protocol extended to support MADI IO. All packets use 8-bit IO selector for inputs & outputs. Bits 6-7 of the selector distinguish the type: 0x00 - analogue 0x40 - digital (AES) 0x80 - buss 0xC0 - MADI [mcc] 7:18:00 5 Jul 2003 - disabled OMF,WVF,ED2 open OMF, WVF and ED2 files can no longer be opened on Project menu. [mcc] 16:57:00 3 Jul 2003 - testpat Type the command "testpat" to see a test pattern that is useful for adjusting monitor contrast & brightness. Press F1 to toggle gamma comparison level between 25% and 50%. Press any other key to exit testpat. The window mgr must be running to use testpat (use "gfx" command to start window mgr without starting product apps). [mcc] 11:46:19 1 Jul 2003 - disabled 23.98 23.98 timecode format is disabled pending further R&D. [mcc] 12:17:33 27 Jun 2003 - fallback configs removed The "fallback" mixer configurations have been removed. Previously, if there were not enough QDCs for the selected configuration you would get some smaller configuration. Now there must be enough QDCs or audioman will not start, and will report "Not Enough QDC DSP Resources". If this occurs you must specify a smaller configuration. For example, to start fame or prodigy with only one QDC the following commands must be used: fame 24/12/12 prodigy 24/12/12 The minimum number of QDCs required for each configuration is documented in the startup scripts. [mcc] 12:31:19 23 Jun 2003 - startup scripts The startup scripts for all products have been modified to accept two parameters, thus allowing audio configuration and project format to be specified independently if desired. The syntax for all scripts is: scriptname [ config [ project ]] The config parameter defaults to 48 if unspecified. The project parameter default to the config if unspecified. The available config parameters for a script are displayed by scriptname -? Two options are also supported: -e Do not execute: process and report parameters but do not lauch apps. -s Do not display splash image: show launch script execution. The *plus scripts have been removed. The new scripts allow users that customise the colour palette file to use their own file. To do this set the PALETTE environment variable in /nvr/profile: setenv PALETTE /dd/usr/sys/mypalette.pal [mcc] 10:58:57 17 Jun 2003 - 18.1.01c Disabled MediaBase interface. [mcc] 16:47:11 16 Jun 2003 - Display no. of tracks The file browser window displays the no. of tracks for .MT files. [mcc] 11:46:19 13 Jun 2003 - 18.1.01b [mcc] 06:45:51 6 Jun 2003 - XPC New function qdc_fmc_compat() verifies FMC/QDC version compatibility. Server now reports transport and IO errors to the user. [mcc] 19:34:17 2 Jun 2003 - hosts file There are only two entries in the hosts file generated by setup, localhost and the hostname/IP as defined in /nvr/setup. The "devhosts" file no longer exists. Bugs Fixed Since 18.1.01a ------------------------- [mcc] 15:42:00 15 Jul 2003 - bg level update failure Background level updates ceased after 2^31-1 samples (~12.5 hrs @48K). [mcc] 10:25:44 13 Jul 2003 - mfxmix/satmix patching In mfxmix and satmix, MDR was not correctly giving FMC exclusive control of input patching as it does for all other mixer products. [mcc] 11:32:14 28 Jun 2003 - 96K silence overflow Fixed bug caused by neglected 96K modification to support silent segment durations >= 2^32 samples. The reported symptom was tvt messages like: "WARNING(fabs(ds) < 2147483648.0)...in sync.c". The bug occurred only when playing deep in the silent gap between end and start of project, around 12hrs 26mins before first clip on a track. [mcc] 09:57:24 13 Jun 2003 - ped Undo label display The palette editor sometimes displayed the Undo button label in the wrong location. Release 18.1.01a, June 2 2003 - Built by Michael Carlos ======================================================== Changes Since 17.4.22a ---------------------- [mcc] 18:42:17 29 May 2003 - 23.98 Alpha version of 23.98 frame rate software. SIO1 changes are still under development. Setup mode: The "NTSC" softkey label has changed to "Rate". When the Format is 24 the Rate key toggles between 24.00 and 23.98. ML import/export: ML3 syntax is "framerate = 23.98" ML4 syntax is "framerate = FR23.98" RAPID: Timecode format QDC_TCF_2398 (4) selects 23.98. [mcc] 18:42:17 29 May 2003 - MediaLink file gopher Various changes/fixes to os9 and Windows gophers: An unexpected error when caching a file no longer causes the file to be excluded from the cache, problem files are not invisible. In general all errors are reported to log and then ignored, previously whole folders or even drives could be excluded due to some error encountered by gopher. [mcc] 10:34:00 13 Mar 2003 - login Reconnect field When MediaBase server fails to respond, dfnserver will not try to talk to it again, because the excessive delay causes DFN RPC timeouts. The Reconnect field in login GUI overrides this and forces a reconnection attempt. Hopefully this is a temporary measure. Bugs Fixed Since 17.4.22a ------------------------- [mcc] 19:31:00 12 Mar 2003 - mfx can't tab in login window In mfx/mfk consoles, macro keys are now turned off by popup dialog boxes that seize keyboard focus. The previous state is restored when the dialog exits. This allows Next/Prev fn keys to be used in the dialog. Release 17.4.22a, March 11th 2003 - Built by Steve Rance ======================================================== Changes Since 17.4.21a ---------------------- [mcc] 14:29:07 10 Mar 2003 - mixer tcf/samplerate pkts The enum parameter for MX_TimecodeFormat and MX_Samplerate pkts is no longer ANDed with 3. See Mixer_Protocol.doc. [mcc] 10:56:23 10 Mar 2003 - login/logout MediaBase login/logout are implemented. To use, type "login" or "logout" in GFX command line. The server must be configured to use MediaBase: [options] mediabase = 1 [mcc] 12:57:00 9 Mar 2003 - preliminary 23.98 changes 23.98 HD timecode format is defined but not yet selectable. Bugs Fixed Since 17.4.21a ------------------------- [spr] 13:05:00 11 March 2003 - Dynamic ROM The Dynamic ROM has been fixed so that when the system debugger is enabled, the system can be resumedd by the "g" command. The serial I/O chip's state was not being reverted to the same as when the debugger was entered. [mcc] 08:24:52 11 Mar 2003 - distinguish-crossfade bug When distinguish-crossfade option enabled, the fade was sometimes displayed as the wrong type. [mcc] 18:23:12 10 Mar 2003 - mfx3+ incompatibility Changed the way MediaLink2 denies access to MediaLink1. The old method caused 15.6 dfnserver to crash! Release 17.4.21a, March 7th 2003 - Built by Steve Rance ======================================================= Changes Since 17.4.20a ---------------------- [mcc] 17:31:21 6 Mar 2003 - Export WAV files Browser is invoked by Export WAV files in Merlin & mfx consoles. Fixed bug when browser is cancelled by a mode key (resulted in erratic behaviour). [mcc] 17:27:00 6 Mar 2003 - dfnserver.ini The dfn_cfg file is obsolete for Windows DFNServer. It is superceded by dfnserver.ini. [awk] 11:21:51 6 Mar 2003 - MediaLink Console App. Removed the "l" option from the upgrade wizard. User code 1 is now stored until it is used once on a successful upgrade, after which a new user code 1 will be stored. [mcc] 09:49:51 6 Mar 2003 - StarGate debug StarGate server debug can be enabled by setting the SGDEBUG environment variable to desired level 1-4, for example: setenv SGDEBUG 2 This must be done in os9 shell, before starting fame or prodigy, or in /nvr/profile. WARNING: Stargate debug can be very verbose. Excessive debug output will impact the performance of your system. Level 4 should be used strictly for controlled diagnostic purposes, never in real-life sessions. [mcc] 12:58:22 27 Feb 2003 - Postponed SysStartup The SysStartup signal to enable the mixer is now postponed until after the editor console startup is complete. If should prove to cause a problem then the old behaviour can be temporarily reinstated by "setenv MFXEARLYMIX 1" before starting the QDC application (or in /nvr/profile). This change prevents initial requests from the mixer from interleaving with those from the console at startup. [steve] Feb 27th 2003 - System Changes Microware Kernel is now Edition 375 (was 372) Microware CSL is now Edition 25 (was 24) [gja] 11:39:39 26 Feb 2003 - Machine Status packet Implemented code for creating mixer packet when machine status changes. [awk] 14:41:00 19 Feb 2003 - MediaLink Licence File Renamed the licence file from dfnserver.ini to MediaLink.dat. After installing this version you can remove dfnserver.ini if you wish. Although leaving it there will not hurt. Bugs Fixed Since 17.4.20a ------------------------- [ajc] 15:47:18 6 Mar 2003 - Record ("Bounce") Sync Bug An error in the timestamping of dropin/dropout events was causing recorded data to be out of sync when using video house sync. [rls] 17:08:19 4 Mar 2003 - SIO1 resync bug When the selected sync source is out of lock for 4 seconds TBASE switches to internal sync until lock is achieved, but sometimes it would immediately switch back to the original source. TBASE now selects internal sync for a minimum of 2 seconds before switching back to the original source. [mcc] 20:04:19 27 Feb 2003 - project load aborted bug A bug has been fixed that only manifested when project load aborted due to the detection of a corrupted project file. The software incorrectly closed garbage path no.s for non-existent file attachments, potentially hanging-up the system. [mcc] 18:55:12 27 Feb 2003 - tape backup bug A bug that prevented Backup to tape was introduced in 17.4.14a. This has been fixed, pending testing. [steve] Feb 18th 2003 - TVT Device Corrected IRQ handling for /tvt and /tvtctrl device. [steve] Feb 18th 2003 - QDCInit/Xilinx Loading Make Xilinx loading from cold start the same as if option x is specified. This corrected a Xilinx loading problem seen on my system that was fixed by specifying -x. I hope this is compatible with all the other QDC's around the world. I have tried it on about 6 QDC's now and it seems to be OK. Release 17.4.20a, February 17th 2003 - Built by Steve Rance =========================================================== Changes Since 17.4.19a ---------------------- [mcc] 17:30:00 14 Feb 2003 - 24-bit WAV support 24-bit WAV files are supported for import and export. Import WAV files SEARCH IN pathlist is inherited from browser position. BIT DEPTH field selects 16-bit, 24-bit, or don't care(neither). Search result list includes BITS and BWF DESCRIPTION columns. Export WAV files Invokes browser to select destination directory and make new folder. Single DESTINATION FOLDER field instead of separate UNIT & FOLDER. BIT DEPTH field selects output format, Auto will use 24-bits for anything other than 16-bit source. Exported 24-bit files are in Microsoft-compliant format. Import/keep, Space/keep-borrowed, Blue-paste and AudioBase blue-paste no longer convert little-endian or interleaved audio formats to MFX native format. The original format is retained. [awk] 18:30:00 11 Feb 2003 - dfnserver for WIN32 Removed the feature added on the 7 Feb 2003 where, if the dfg_cfg file was not found in the current directory then dfnserver would search for the file using the "PATH" environmental variable. Bugs Fixed Since 17.4.19a ------------------------- None Release 17.4.19a, February 11th 2003 - Built by Steve Rance =========================================================== Changes Since 17.4.18b ---------------------- [awk] 12:30:00 10 Feb 2003 - Medialink for WIN32 Added the new licensing features as documented in the Medialink 2 Function Spec. Added the Medialink.exe console app as documented in the Medialink 2 functional Spec. This app will allow a user to start and stop the dfnserver applications. It will also allow a user to configure their licence. The software will run by default with a demo licence. This licence is valid for 7 days from the time of installation and will allow connection to only 1 engine at a time. After 7 days the licence will expire and hence the software will not run. This 7 day period is a beta test period only, the final release the period will be set at 30 days. To reactivate your expired demo licence or to upgrade your licence you should use the upgrade wizard provided by the UI in the medialink.exe app. The wizard will display 2 numbers, your user code 1 and user code 2. Contact Fairlight with these 2 numbers to which Fairlight will respond with a register key number. Enter this register key number into the wizard to upgrade your software. For now, during the testing period, Allan Kirk will act as the Fairlight contact. [mcc] 13:07:39 10 Feb 2003 - Export WAV files Export WAV files softkey now requests user to browse to destination directory and press Okay softkey. A new folder softkey is also provided. This is only in Dream family consoles so far. [ajc] 01:48:08 10 Feb 2003 - option.h (OSK & Unix) Added macro opt_argfile & library fn to append args to argv from a file. [awk] 17:10:00 7 Feb 2003 - Medialink for WIN32 Added software licensing to dfnserver. dfnserver will now require the appropriate licence to run. Added a new application called "medialink.exe" that allows a user to configure their dfnserver licence [awk] 17:10:00 7 Feb 2003 - dfnserver for WIN32 If the dfg_cfg file is not found in the current directory then dfnserver will search for this file using the "PATH" environmental variable. [mcc] 11:09:52 7 Feb 2003 - Embedded objects for satmix The embedded object server is now enabled for satmix. [mcc] 10:06:17 7 Feb 2003 - Monitor softkey The Arm Monitor softkey behaviour has been revised for consoles other than Merlin. The Monitor key is once again a toggle, it alternates between INPUT and one of the other modes: AUTO,RECORD,MUTE or REPRO. By default the other mode is AUTO, but this can be changed by pressing Blue-Monitor, selecting a mode, and pressing enter. The Monitor setting is retained across sessions. Note: in macros and mfx3/48 serial remote control, number-pad digits can be used to select mode in Blue-Monitor: 0 = AUTO 1 = RECORD 2 = MUTE 3 = REPRO Bugs Fixed Since 17.4.18b ------------------------- [awk] 17:10:00 7 Feb 2003 - dfnserver for WIN32 fixed a minor option z problem [awk] 19:30:00 06 Feb 2003 - dfnserver validating domains fid_valid(), when validating a given domain, would validate the domain only. It should have validated the given domain AND the given node. Release 17.4.18b, February 6th 2003 - Built by Steve Rance ========================================================== Changes Since 17.4.18a ---------------------- None Bugs Fixed Since 17.4.18a ------------------------- [steve] 11:14:07 6th Feb 2003 - Features Bug in features fixed. See CVS history for details. [mcc] 17:52:02 5 Feb 2003 - 16bit LE waveform display Fixed 16bit LE waveform display, newly broken in 17.4.18a. [mcc] 13:47:22 5 Feb 2003 - Undo after ML import Fixed ancient bug: pressing Undo after ML import would undo the entire EDL. Release 17.4.18a, February 5th 2003 - Built by Steve Rance ========================================================== Changes Since 17.4.17a ---------------------- [mcc] 09:02:23 5 Feb 2003 - MediaLink2 The new MediaLink protocol is now enabled, there is no compatibility with earlier releases. The node registration policy is active, fixed at 3 (THREE) nodes until dongling integrated. Support for write-protected files is enabled. [mcc] 11:18:11 2 Feb 2003 - Library volumes Windows dfnserver will hide drives from the file browser that are listed as LIBRARY_VOLUMES in dfn_cfg file (same syntax as MEDIA_VOLUMES). Library volumes are handled exactly the same as media volumes except they are invisible to the browser. [awk] 13:45:00 30 Jan 2003 - ONCRPC library Activated the multithread safe version of the library [awk] 15:30:00 29 Jan 2003 - ONCRPC library Changed the type from a DLL to a static library [ajc] 16:07:05 27 Jan 2003 - ODIF Added 24 bit little-endian modes to ODIF interface [mcc] 12:59:00 25 Jan 2003 - ML export buffer Omdlexport uses 64K output buffer instead of 512 bytes, runs a bit faster as result, around 20% for large files. [mcc] 18:27:12 21 Jan 2003 - mixint in Satellite R&D dongle is no longer required for mixint in Satellite. Bugs Fixed Since 17.4.17a ------------------------- [ajc] 01:27:27 4 Feb 2003 - missing oscillator/talkback feeds Corrected bug in translation of mixint feed numbers which caused talkback feeds (TB + Osc) to not work. [mcc] 10:17:33 30 Jan 2003 - import error disabled tracks Fixed bug that spuriously disabled tracks after an error opening a project in the Import menu. The first time this occurred after startup, ALL tracks were left disabled. This appeared to be unrecoverable so users would restart, but the simple work-around is to go to Disable menu and press ON/OFF softkey to enable all tracks. [mcc] 16:16:00 28 Jan 2003 - interleaved stereo playback Fixed bug in playback of interleaved stereo WAV files, introduced in 17.4.17a [mcc] 12:33:51 23 Jan 2003 - ML4 too many segments Exported ML4 files (correctly) contained segment lists for embedded objects. However these cannot be imported, the ML4 parser fails with error "too many segments". To avoid this, object segment lists are now omitted from exported ML4 files. [steve] 19:29:00 22nd January 2002 - ANETD Fix ANETD so that a malformed pathlist will not cause ANETD to bus trap. Release 17.4.17a, January 21st 2003 - Built by Steve Rance =========================================================== Changes Since 17.4.16a ---------------------- [ajc] 00:55:10 21 Jan 2003 - MFXAUTO Functionality Implemented mixer packet MX_DEFREC (set dropin/dropout times) for autoconform functionality via RAPID i/f. [mcc] 15:42:19 17 Jan 2003 - Export menu The Dream-family Export menu has been migrated to Merlin and MFX. See 17.4.02a release notes. [mcc] 13:00:00 13 Jan 2003 - Import/WAV Files The SEARCH IN field is initialised to the file browser position. Files with pathlists longer than 80 chars can now be imported. [awk] 19:54:23 8 Jan 2003 - ONCRPC DLL build environment Modified the ONCRPC DLL to export data using __declspec(dllexport) rather than the oncrpc.def file. This will make it easier in the future to change the library into a static library. [mcc] 19:54:23 3 Jan 2003 - Prepare for MediaLink node dongle Preparatory changes have been made to support MediaLink node dongle. No change in behaviour is intended or expected yet, but the new behaviour can be enabled (at compile-time) by undefining DFN_V2_COMPATIBLE in DEFS/DFN/dfn.h. See Log 1.14 in that file. [rls] 17:37:17 18 Dec 2002 - tcs_cfg option Added environment variable SONY-FOLLOW-SLAVE to tcs_cfg If enabled then, if the slave relocates while we (the master) are in STOP mode, then then we will relocate to follow it. Also, if the slave goes into PLAY when we are stopped, we will also go into PLAY. This feature needs further debugging, since the master and slave are not tracking correctly. [rls] 17:11:17 18 Dec 2002 - tcs_cfg option Added environment variable SONY-CUE-COMPLETE to tcs_cfg If enabled then, if a slave sets the "CUE-UP COMPLETE" status bit it will override the "CUE-UP" status bit (if it is also set). This is needed for compatability with a SSL slave. The symptom is that the STOP LED remains flashing after a locate operation. [mcc] 11:26:17 17 Dec 2002 - backup/restore popup When marking a file for backup/restore/copy the destination file is checked not only for existence but also write-permission. If the file is read-only (and hence cannot be overwritten) the user is notified with a popup. Bugs Fixed Since 17.4.16a ------------------------- [steve] 15:58:00 20th January 2003 - Path table extensions. Fixed path table extension bug which caused only 32 paths per task to be open at the same time. This was introduced by added flight into the PREIO section of modules to be initialized. As IOMAN is inited after this, all I$ called were not being intercepted as expected. This mod was only introduced to enable "named" mutxes in the SCSI driver. As this method is no longer used inside the SCSI driver, flight has been reverted to be initialized in the startup file. [mcc] 16:03:22 15 Jan 2003 - anetd dongle Without MediaLink dongle, os9 anetd would not run and could not service mixer object file requests from FMC. Now the dongle is not required. [mcc] 11:28:37 15 Jan 2003 - ML parser Lines longer than 256 chars in ML file were not read correctly, the input line buffer was too small to handle long pathlists. ML4 WAVE birthdays were not imported correctly. [mcc] 17:00:00 8 Jan 2003 - crossfade priming and display bugs When locating to a crossfade the prime buffers were not ordered correctly. This could sometimes cause the track to become "starved" later, when the initial buffers need to be reused. It also provoked another bug, where the waveform display would "stall" at the cursor after a crossfade, and only render as the cursor passed over the clip. Both problems have been fixed. [ajc] 11:44:38 19 Dec 2002 - missing feeds after stereo unlink in Fame Fixed bug in group allocation which would sometimes allocate too many feeds to a DCC; this caused the right feed of a stereo pair to go dead after an unlink. [ajc] 01:07:45 18 Dec 2002 - bounce/StereoMixedOut track sync bug Fixed another error in mix-chain setup which meant that tracks from different DSPs were not exactly in sync in stereo mixedout mode & multitrack bounce. [mcc] 11:26:17 17 Dec 2002 - ANETD bugs Windows and os9 anetd both had the potential to confuse two files with the same "inode". At worst, this could result in the application reading or writing the wrong file. That would cause symptoms that resembled corruption of the file system. Windows anetd has been simplified by taking advantage of the fact that Windows will not allow an open file or any of it's parent folders to be renamed or moved. [ajc] 21:34:01 16 Dec 2002 - locate bug Fixed error in subDDfrac() in fractional computation library which caused TBASE to locate to the wrong position under certain circumstances. Error would depend on exact alignment of TU and WSN before and after the jump, and on the current sample rate, frame rate and NTSC setting. [ajc] 01:08:11 16 Dec 2002 - bounce sync bug Fixed error in loading of DCC parameters at startup which caused sync offset for bounce operations to be wrong. Release 17.4.16a, December 11th 2002 - Built by Steve Rance =========================================================== Changes Since 17.4.15a ---------------------- [mcc] 06:42:11 5 Dec 2002 - 24-track GFX priority The same GFX priority is used for all configurations. Previously it was slightly lower if MDR started with 24 tracks. Bugs Fixed Since 17.4.15a ------------------------- [mcc] 09:48:22 11 Dec 2002 - edit range display bug The range was displayed incorrectly when the field of view of the track display included the point(s) 12 hours away from range head tail. The range would then be displayed as though it completely enclosed the current view. [mcc] 18:35:51 8 Dec 2002 - FLFS re-entrancy bug A serious bug in FLFS file mgr has been fixed. The file system was in jeopardy when two or more processes were updating the file system simultaneously. E.g., one process deleting a file while another process is creating a file. Damage to the file system was most likely with project files on the boot drive of mixer systems, due to anetd transactions from mixer. IMPORTANT: flfscheck all FLFS drives before installing this build. This bug fix cannot repair disks that are already damaged, it can only prevent future damage. [mcc] 15:48:22 6 Dec 2002 - stereo clip on track N Fixed bug provoked by stereo clip on track N of an N-track system. Release 17.4.15a, December 4th 2002 - Built by Steve Rance =========================================================== Changes Since 17.4.14a ---------------------- [ajc] 21:03:50 2 Dec 2002 - New product Configurations - Added fameplus & prodigyplus cfgs & support; these cfgs have an additional 8 live feeds and 8 busses over the "normal" cfgs. - Added dreamplus cfg & support; this cfg has an additional 48 short return feeds which have 2 band EQ and no dynamics. [ajc] 21:07:19 2 Dec 2002 - Audioman Update - Option to patch monitor matrix output pre-fader. - Some audioman fixes ported from 17.3 - Increased max number of Fame Multitrack Busses to 16 [ajc] 01:00:29 29 Nov 2002 - SCSI Driver Reverted last change. Seems to provoke a hang-up bug in scsi driver which has already been reported but which is much worse with this change. Real cause of hang-up still being investigated. [ajc] 01:17:17 27 Nov 2002 - SCSI Driver Changed to use Flight semaphore/mutex facilities whenever these are available instead of polling mutex. Steve's change below should ensure that Flight is available when the driver starts, but if it isn't for some reason then the old polling method will be used until Flight is loaded. This change should ensure that processes are not locked out of the scsi driver by "unlucky" timing of the polling operations. The mutex ensures that requests are serviced strictly in order of arrival. [steve] November 26 14:52:12 2002 - Flight flight is now initialised during the kernels P2 init stage. This is so flights extension facilities can be used as soon as possible. You should now see "flight: v3:16 - Fairlight OS9 extensions" being printed immediately below "A valid OS9 bootblock was found" when booting to a hard disk. Bugs Fixed Since 17.4.14a ------------------------- [mcc] 11:30:54 3 Dec 2002 - FLFS file names FLFS file mgr now ensures that the first character of a new file name is a legal character. An application (FFS?) bug is creating files with names like "/sd01". This mod is intended to make that impossible, and will return Error #235 Bad Name instead of creating the illegally named file. These characters are now prohibited as first character of a filename: / \ * & ; - > < # ^ and space. [mcc] 07:52:25 29 Nov 2002 - bus trap in close Fixed mdrmain bus trap when closing a remote file on a host that has been shutdown or disconnected. It still won't close the file properly but it shouldn't crash now. [steve] November 27 10:10:12 2002 - SCSI port addresses SCSI port addresses as reported by dvcs, scct, irqs are now consistent with each other. The Encoding is 0xABCCCCDE where :- A = The top nibble of the port address. This is 7 for a QDC engine. B = The PCI Slot Number CCCC = Zeros for a QDC engine D = The SCSI Target ID E = The partition Number (for old OS9 and FAT32 disks) [steve] November 27 14:44:11 2002 - mediad - Spin up Hard Disks Spin up fixed disk drives that are installed after the system has booted. Disks on the SCSI chain at bootup would be spun up by the ROM. This would only effect hard drives which are optioned not to spin up when power is applied. (Kingston type drives) Release 17.4.14a, November 26th 2002 - Built by Steve Rance =========================================================== Changes Since 17.4.13a ---------------------- [steve] November 19 16:48:04 EST 2002 - ROM startup The ROM startup speed has been improved. There is a new NVRAM variable that controls the select timeout for non-existent devices. After an NVRAM re-configuration, the scanning of SCSI devices will be much faster. Existing systems will be slow untill this a re-coniguration is done. [steve] November 19 16:48:04 EST 2002 - ROM "press a key to pause" The ROM can be made to pause in a number of places to examine the screen. It is possible that a stray key from an attached console could trigger this feature and pause the ROM when not required. Although this condition could happen, there are two places in the ROM already that would pause the system and neither of these have been reported as a problem. So I assume stray keys dont happen. (at least not during power-up or restart) During the beta stage please report to me (steve) if you do see the ROM paused when it should not have been. Bugs Fixed Since 17.4.13a ------------------------- [mcc] 11:23:51 26 Nov 2002 - new/open 24 in fame,prodigy In 48-track configurations, the project format can be changed to 24 in fame and prodigy. That is, the new24 key and the opening of 24-track projects should work exactly as for all other products. [mcc] 10:47:21 25 Nov 2002 - Backup use connection cache The dfn_backup task now uses the RPC connection cache again, but limited to 4 connections. The cache became disabled by default in 17.4.11a and the behaviour of inter-node backup/retore/copy would have been affected. [mcc] 10:00:00 25 Nov 2002 - Refresh bugs There was a bug in dfnserver that was provoked by client trying to add to the cache a file that is not a recognised MediaLink file type, for example a .zip file. The only software that does this is the Refresh procedure. The bug could cause dfnserver to exit due to fatal exception (access violation in Windows). The Refresh procedure consumed too much stack and too many paths. When recursively refreshing a deep tree, FFS could crash from stack overflow, or the operation could fail with Path Table Full error. It also bailed out at the first error and didn't keep going. Now it ignores non-critical errors and attempts to continue. Discovered that ctrl-Q causes refresh to terminate. This may not be well known. If you mistakenly start refresh at wrong location, for example, then press ctrl-Q to stop it. [gja] November 25 09:20:40 EST 2002 - EQ smoothing A bug in EQ smoothing was fixed whereby after an extended period of EQ changes, the smoothing would inadvertently stop working. The problem was due to the use of a timeout counter as a signed integer rather than an unsigned long. [awk] November 21 13:00:00 EST 2002 - ONCRPC DLL Currently the ONCRPC DLL is safe to be accessed concurrently by multiple processes. But it is not safe to be accessed concurrently but multiple threads running from within in the one process. Define the preprocessor symbol "MULTITHREAD_SAFE" (in RPC.H) and recompile the DLL to create a multiple thread safe version of the DLL. Please note that the interface of the multiple thread safe DLL had to be changed and hence dfnserver will also have to be recompiled. The MULTITHREAD_SAFE preprocessor symbol is undefined by default and hence the DLL is unchanged by default. Release 17.4.13a, November 18th 2002 - Built by Steve Rance =========================================================== Changes Since 17.4.12a ---------------------- [steve] November 15 08:38:53 - SelectablePCI/SCSI Boot slot The PCI/SCSI boot slot can now be selected by the sames means as selecting the SCSI target in the ROM. Before this change, you issued the command "scsi 4" to boot to SCSI device 4. You can now insert an extra number which will control the PCI slot to boot to. For example "scsi 23" will boot to Slot 2 device 3. The same method of using two numbers is also used when setting NVRAM boot order. [mcc] 13:13:51 14 Nov 2002 - RAPID MDR V1.2: track numbers The RAPID XPC interface to MDR now requires track numbers to range from 0..N-1. Previously they were 1..N. [steve] November 13 14:58:53 - Optional second argument for FAME and Prodigy The FAME and Prodigy 24 track file format start up commands can now take a second argument of "48". This will start them as before but with a 48 track file format by default. [steve] November 13 14:58:53 - "New 24" in FAME and Prodigy When the system has been started with the 48 track file format, FAME and Prodigy now have the "New 24" button on the project menu so 24 track projects can be created. [mcc] 09:14:33 13 Nov 2002 - Input monitor settings in NVRAM The input monitor settings are now stored in console NVRAM and will be restored at next restart. Both the main and alternate settings are stored to preserve the toggling behaviour of double-clicking Monitor softkey (MFX or Dream family) or pressing Input Mon key (Merlin). [steve] November 12 20:21:00 - Backup Buffer overflow A possible buffer overflow was possible in backup when the suffix of the file being backed up was greater than 2 (i.e. .WAV). This has now been fixed so any suffix size upto 4 can now be backed up. [steve] November 11 17:14:23 - Hide Boot Device A new option in the /nvr/setup file can hide the boot device from the QDC software. The new section is called [Options] and the new member variable to control this is "HIDE_BOOT". The arguments for this can be YES or NO. When set to YES, the device that is the same as "/dd" is not sent to DFN or file_gopher to be included either by the local system or any remote system. [steve] November 11 16:16:07 - tapes can now use all 28 chars Backing up and restoring will now preserve all characters of a file limited only by OS9. [mcc] 13:54:29 11 Nov 2002 - 28-character file names File names displayed and entered on the console can be up to 28 characters long. That is the total length, including suffix. Exported WAV file names can also be this length. [rls] startup Enabled input sample rate conversion (SRC) for all AES2 cards by adding "-c" option to qdcinit in /dd/startup [grega/-steve] 8 Nov 2002 - New FMC FMC V2.0.207 is now included Bugs Fixed Since 17.4.12a ------------------------- [steve] - 17/11/2002 23:40:41 - Booting to the ROM. Fixed long standing bug where booting to the ROM after scanning the PCI SCSI bus or attempting to boot to a PCI SCSI device caused a ROM boot to fail. This was because the /dd device was set up for a real disk and not a RAM disk that the ROM boot block requires. Booting to the ROM as the first operation was OK. [mcc] 08:41:22 15 Nov 2002 - WAV export safe tracks WAV export was mistakenly prohibited for safe tracks. [mcc] 07:20:48 10 Nov 2002 - Filename buffer overflow The UI for editing filenames did not correctly handle the case where the length of the name is longer than the maximum allowed. This can happen if the file was created or renamed outside of the console UI. The bug allowed any number of characters to be entered until the console finally crashed. Introduced in 16.6.13. [mcc] 13:24:09 7 Nov 2002 - FFS Known Module Sometimes in response to dfn errors, FFS would not unlink browser data modules correctly, causing the curious error "Known Module". This has been fixed by adding a garbage collection mechanism to unlink a module that is created but then not used due to dfn or other error. Release 17.4.12a, November 5th 2002 - Built by Steve Rance ========================================================== Changes Since 17.4.11a ---------------------- None Bugs Fixed Since 17.4.11a ------------------------- [mcc] 09:52:23 6 Nov 2002 - new FFS bug With every quit and application restart, FFS opened an additional path to the browser position file, eventually it would run out of paths. To work around, if you quit then always reboot, don't just restart the application. Introduced in 17.4.09a. [mcc] 08:19:56 6 Nov 2002 - Edit List Mismatch after changing frame rate After changing project frame rate, the clipboard was sometimes left in a state that caused "Edit List Mismatch" error to occur when the project was closed. The problem was that the changes to the clip times on the clipboard were not flushed to the file. This did not matter (because the clipboard is discarded when a file is opened), but it caused this alarming error to be reported. [mcc] 18:36:07 5 Nov 2002 - mangled hooks after changing frame rate When changing the frame rate of projects that straddle "midnight", there was a bug with the potential to mangle the hook times of clips starting before midnight. Clips affected by the bug would not play correctly, displayed hashed waveform, caused error End of Waveform and warnings about TAIL GT END when opening the project. This bug has been fixed. [mcc] 14:54:25 5 Nov 2002 - ANETD/netman bugs Windows ANETD did not return EOF error correctly when reading directory. The directory read function was not multi-thread safe. File modification/creation dates were not being returned correctly from the server, nor were they being received correctly by the client. Release 17.4.11a, November 4th 2002 - Built by Steve Rance ========================================================== Changes Since 17.4.10a ---------------------- [mcc] 13:06:51 3 Nov 2002 - MediaLink Various changes to reduce the number of open RPC connections. 1) The dfnserver RPC service port numbers are cached by dfnlib. These persist for the lifetime of the server and only need to be obtained once from the portmapper. This avoids redundant UDP request to portmapper for every connection. 2) Portmap request timeout reduced from 80 to 15 seconds. 3) Only V2 DFN protocol is supported for outgoing connections. This avoids node_info request to get version for every connection. 4) Dfnlib provides optional notification of errors to node daemon. 5) Noded V3.06: Uses new strategy of delaying between nodes (-b option). Accepts requests from dfnlib user to re-scan a node immediately. This prevents UI from waiting for node deamon to notice that a node has gone offline. 6) FFS: The RPC connection cache is enabled but limited to 5 connections. Noded error notification is enabled so that noded is notified immediately when the node selected by the browser goes offline. Two ancient bugs fixed in the handling of that case. There should now be no more than ten or fifteen second delay. All remote connections are closed by FFS_Stop (when MDR quits). [rls] 15:48:23 31 Oct 2002 - startup Start SCSI Caching before JTAG, it saves a few seconds. Added printout "... JTAG error" if qdcinit is not run due to missing tapcfg [mcc] 10:58:23 25 Oct 2002 - tvtsave The "long" output format is now the default; the -l option is not required. The new -s option will force "short" format if desired. Never use short format for tvtlogs to be sent to Fairlight R&D; we always want the long format, hence this change to tvtsave. [steve] October 24 16:26:47 EST 2002 - ROM Power Off A new boot driver is in the ROM. This driver will switch off a system with a software controlled PSU. Its main use is really for diagnostics. The author cant think of any reason why a user should need this feature. Possibly only to power off after attempting a disk boot maybe????? [mcc] 13:38:45 22 Oct 2002 - separate tape gopher A separate instance of file gopher is forked for use with tape devices. This prevents dfnserver from blocking for a long time due to file gopher rewinding tape. Delays will now occur only if there are two or more tape drives mounted. [steve] October 22 13:35:51 EST 2002 - SCSI driver Timeout Revert SCSI driver timeout to 100 seocnds. This will really make the driver timeout anytime between 100 and 200 seconds. Bugs Fixed Since 17.4.10a ------------------------- [mcc] 09:54:22 22 Oct 2002 - EMU 422 receiver hogged cpu The 422 receiver task could hog the (virtual) cpu in extreme cases. This prevented input key task from running and responding to input. [ajc] 23:00:20 30 Oct 2002 - DSP code bug fix Fixed buffer addressing bug in non-BGUPDATE mixer code; was not causing any problem in 17.4 as BGUPDATE is now always enabled. Release 17.4.10a, October 21st 2002 - Built by Steve Rance ========================================================== Changes Since 17.4.09a ---------------------- [steve] October 21 13:39:25 EST 2000 - Remove Debug Removed uncondditional debug from the SCSI timeout fixes made in 17.3.12a [steve] October 21 13:39:25 EST 2000 - Mediad Only ignore BUSY error if device has already been made. [mcc] 12:43:09 20 Oct 2002 - File gopher The Windows and os9 file gophers no longer keep dfnserver blocked for unacceptable periods. This was causing client requests to dfnserver to timeout, thus making that node appear to be offline. Bugs Fixed Since 17.4.09a ------------------------- [ajc] - 23:12:28 17 Oct 2002: mdr3 DSP program - "DC Noise" bug Fixed uninitialised register bug in fwd shuttle with linear interp for NCHAN==3 or 4. This only manifested in mdr3 since mdr4 uses no interpolation. Bug was introduced during mods for varispeed. Release 17.4.09a, October 17th 2002 - Built by Steve Rance ========================================================== Changes Since 17.4.08a ---------------------- [gja] 12:28:16 14 Oct 2002 - EQ De-glitching code addition Added code to reduce clicking in audio during EQ parameter changes. As part of this, and aman message was introduced to allow a ramp time parameter to be changed at runtime. This message is invoked via aman eqr where is the number of audioman service loop timeouts an EQ parameter change will be ramped for (each timeout is about 20msec), and is currently not used (and should be set to 0), but will eventually allow the ramping time to be set on a per-feed basis. In parallel with the EQ de-glitching code additions, a function to convert from MDR filter descriptor to biquad coefficients directly was introduced - this is called f2b. [mcc] 15:14:22 12 Oct 2002 - V2.00 zeromon Zeromon now has some options and a help sheet (see zeromon -?). An incompatibility with mapped network drives has been confirmed in Windows 2000. If zeromon is enabled for any mapped network drive then all mapped network drives become inaccessible to ANET and to most of the PC itself. Therefore the rules for selecting the drives on which to install the driver have changed. In V1, the driver was installed on ALL drives, including inactive drives and mappped network drives. Now zeromon is only installed on ACTIVE drives, but NOT the boot drive or any mapped network drives. New options allow this behaviour to be modified: -i -- install on inactive drives -b -- install on boot drive (not recommended) -n -- install on mapped network drives (not recommended) The V1 behaviour can be obtained with all three options: zeromon -ibn. Zeromon is never installed on A: or B: drives. At startup, zeromon now prints version and build date, and a line showing the drives on which the driver has been installed. [mcc] 14:32:11 9 Oct 2002 - QDC noded V3.04 The -R option enables the RPC ping of known nodes. RPC ping error/timeout tvt messages are suppressed for unknown nodes. Bugs Fixed Since 17.4.08a ------------------------- [steve] 14:21 17 Oct 2002 - Backup/Verify Problems A probelm related to backup/verify has been fixed. The problem was tape drives that take a long time to rewind caused mediad to unload the device descriptor for the tape device. This in turn caused the verify operation to fail as there was no longer a tape device in the system. [mcc] 09:04:53 17 Oct 2002 - Windows node and media daemons If an error occurred when mounting a node or drive, the operation was not being retried. Now node and mediad daemons will retry until all nodes/drives are mounted. [mcc] 17:57:22 11 Oct 2002 - Set file size on server V2.01 anetd correctly returns errors incurred when setting file size. The mapping of Windows to os9 errors is more complete and will result in more meaningful errors from anetd on the server. [mcc] 08:33:51 9 Oct 2002 - Incompatible stereo alignment A solution has been found for stereo WAV files in which the audio is not aligned to a multiple of four bytes. Previously, attempting to import such files would fail with the error: "Incompatible stereo alignment". This has been fixed. Borrow and keep will now work correctly for these files, as will import of ML files that describe such WAV files. Note that ML import used to silently accept such files, but they would not play correctly. That is, the "Incompatible stereo alignment" error was not reported, but the alignment was in fact incompatible. To be precise, a WAVE with 16bit interleaved stereo format and a segment start address that is not a multiple of four would not play correctly. Now it will. Release 17.4.08a, October 8th 2002 - Built by Steve Rance ========================================================= Changes Since 17.4.07a ---------------------- This build was done when 11 QDC systems were avalible for a week. Noded has some new options and also a new timeout/polling method. Bugs Fixed Since 17.4.07a ------------------------- Nodes no longer drop on and off line. Release 17.4.07a, October 3rd 2002 - Built by Steve Rance ========================================================= Changes Since 17.4.06a ---------------------- [mcc] 10:29:17 3 Oct 2002 - MediaLink QDC noded V3.01 uses RPC connection cache by default The connection is purged when node taken offline. New -x option disables the connection cache. Dfnlib returns new specific errors instead of -1 in most cases: 150:047 No DFN Connection - Null RPC handle for request 150:048 Stale DFN Connection - Cached connection is stale 150:049 DFN Request Failed - No reply from server, no other error obtained. 150:050 Can't Open DFN Connection - can't connect, no other error obtained. 150:052 Can't Verify DFN Version - initial request to get version failed Bugs Fixed Since 17.4.06a ------------------------- [mcc] 18:15:23 2 Oct 2002 - new source of bus trap in dfnlib New dfnlib created potential for bus trap. Fixed by making send_rpc() return error EHOSTDOWN if CLIENT handle is null. Release 17.4.06a, October 2nd 2002 - Built by Steve Rance ========================================================= Changes Since 17.4.05a ---------------------- [mcc] 12:38:22 2 Oct 2002 - MediaLink QDC noded V3.0: -c option enables RPC connecton cache. QDC dfnlib (all clients): osk U option disables UDP probe connection, additional diagnostic and execution timing messages. Bugs Fixed Since 17.4.05a ------------------------- [mcc] 16:30:22 1 Oct 2002 - ANETD V2.00 Windows ANETD was causing various kinds of file corruption by opening two handles for every file. It now opens only one handle and the corruption does not occur. Two handles were being opened in an attempt to maximise throughput, but this no longer seems to be necessary on modern PCs. Indeed, the throughput now is somewhat faster that it was before. This bug was reported under the guise of several different faults. The recipe that led to the fix was to create new project on server, then borrow one clip from another project on server. The result was "hashed" waveform and Non Existent Waveform error. This no longer happens. Release 17.4.05a, September 30th 2002 - Built by Steve Rance ============================================================ Changes Since 17.4.04a ---------------------- [mcc] 14:30:19 29 Sep 2002 - MediaLink The Windows client code generated by ONCRPC rpcgen was not multi-threadable. Modified rpcgen to emit code that uses thread storage for static return values/structs if MULTITHREAD is defined. T The DFN connection cache is no longer shared by all threads. The cache method has been revised so that CLIENT handles are only destroyed by dfn_close(). [ajc] 00:26:36 29 Sep 2002 - StationPlus . Added StationPlus cfg, which has 24 busses instead of 16. This makes use of the new feature Background Coefficient Update in the DSP code. As such it is still experimental until listening tests have confirmed that there is no loss of audio quality using BGupdate mode. . New startup command "stnplus" takes parameter 48 (default) or 32. There is no "stnplus 24" as this seems to imply a single QDC configuration but it is not possible to get 24 busses with 1 QDC. . Single QDC Station (stn 24) now has 16 busses, also using BGupdate mode. The current 12 buss cfg is still available as "stn 24/8/12". [mcc] 14:12:33 27 Sep 2002 - MediaLink The "DFN request failed" message from dfnlib at last includes the function name, reliable server IP address, and thread name (Windows). Windows dfnlib now uses more multithreaded storage. Os9 dfnserver default _mallocmin() reduced from 1Mb(!) to zero. Windows dfnserver -z option has four levels, more z's produces more debug output. Bugs Fixed Since 17.4.04a ------------------------- None Release 17.4.04a, September 27th 2002 - Built by Steve Rance ============================================================ Changes Since 17.4.03a ---------------------- [ajc] 03:18:54 27 Sep 2002 - Background update of mixer coefs Experimental: update mixer coefficients at background level; this means ramp will not be updated on every sample and there is a possibility of zipper noise effects. DSP requirements for a given size mix matrix are reduced. This change is initially for evaluating the feasability of using this technique in 17.4. Bugs Fixed Since 17.4.03a ------------------------- None Release 17.4.03a, September 25th 2002 - Built by Steve Rance ============================================================ Changes Since 17.4.02a ---------------------- [mcc] 15:05:22 25 Sep 2002 - Windows ANETD Experimental V6 of ANETD that uses only one path submitted for testing. See anetd -? for description of new options. [mcc] 14:18:12 25 Sep 2002 - V2.2 os9 node daemon V2.2 os9 node daemon - revised for greater tolerance. Introduced three-state ping status with NodeState: offline, online, or unknown. The status now classified as unknown used to cause nodes to be considered offline. Tvtline debug screen can be enabled with osk N option as well as -t option. Added r,i,k,u,n options for more control. Changed meaning of -a option to a count, the number of iterations of the polling cycle. The -l and -h options now override the SCANHOSTS setting and have no default values. If both SCANHOSTS and -l,-h options are undefined then the scan range is 1-31. A warning is now printed if the node name does not match the name of the constructed device descriptor. This will occur if the node name contains illegal characters like "-". Bugs Fixed Since 17.4.02a ------------------------- [steve] September 25 11:05:18 EST 2002 - tvtline Fixed bug in signal handling of the debugging tool "tvtline". It was possible for updates not be displayed because of the way the OS9 "wakeup" signal is handled when a task is not sleeping. This has been fixed. [mcc] 14:31:24 14 Sep 2002 - MediaLink bugs Three bugs fixed in dfnserver & node daemon: Node daemon pinged local host If the server IP address was included in a network scan range, then node daemon was pinging and notifying the server of itself. (Windows version only). Insertion of 1st node in local domain The node list was not always correctly grouped by domain. This may have confused the browser. (Windows and os9 versions). Mishandled hostent Mishandling of the hostent structure caused the server to think it had more IP addresses than it really did. The extra addresses were all bogus. (Windows version only). Individually, none of these were shown to have any serious side effects. But they are all related to the detection and handling of nodes going on and off line. In combination they may beeen contributing to general misbehaviour of nodes that clients are reporting, but which we cannot reproduce. Release 17.4.02a, September 11th 2002 - Built by Steve Rance ============================================================ Changes Since 17.4.01a ---------------------- [mcc] 16:30:12 11 Sep 2002 - Export menu The Export menu in Dream consoles has been revised to allow selection of ML format. The fields have been renamed for more consistency: DL file - export DaD DL file ML file - export ML file in selected format ML Format - select ML format (retained in NVRAM): V3 Flat Original flat ML3 V3 All All clips ML3 V3 mfx3+ All clips and incomplete pathlists for mfx3+ V4 ML4 WAV files - export WAV files [mcc] 19:22:58 10 Sep 2002 - filex -i option Rev 15.6 will not accept ML3 files with complete pathlists if the pathlists are not current. The MediaLink search engine is not invoked and "some of the audio..cannot be found" error occurs. To work around this the filex -i option has been implemented. This causes omdlexport to use "incomplete" pathlists, where the leading /node/dev/ elements are omitted. [mcc] 11:41:27 7 Sep 2002 - clip selection The selection of clips for editing has been substantially revised to support the features of the RAPID/XPC interface. The RAPID client now has more power and selectivity than the binnacle user. The binnacle Cut, Copy and Erase menus and Merlin/MFK Edit menu have been implicitly affected by these mods, but no change to their behaviour is intended. Bugs Fixed Since 17.4.01a ------------------------- [mcc] 12:27:33 11 Sep 2002 - StarGate online detection The detection of StarGate status in fame/prodigy was inadvertently broken in 17.1.05a, and has now been repaired. [mcc] 13:26:19 10 Sep 2002 - Faulty AudioBase clean-up On exit from AudioBase menu, waveforms were sometimes disposed when they shouldn't have been. This caused "Non-Existent Waveform" or "Non-Existent External File" errors following Undo or Redo. Release 17.3.08a, September 5th 2002 - Branched 17.4 Here (17.4.01a) ====================================================================== Changes Since 17.3.07a ---------------------- [ajc] 12:09:57 3 Sep 2002 - Satellite with Mixint Separated aman_cfg enable flags for mixer comms (mixintenable) and mixer functionality (mixenable). Mixer comms now enabled for Satellite configurations (only with Developer Dongle for the time being) [ajc] 12:11:11 3 Sep 2002 - Varispeed Varispeed SRC filters now switched according to current sample rate; filter characteristics now optimised for each DSP config/sample rate combination. Filter description files (*.fdp) in each directory can be viewed using 'fdp' utility. Bugs Fixed Since 17.3.07a ------------------------- [ajc] 12:09:15 3 Sep 2002 - Mixer Meters Fixed bug which could cause spurious data to be sent if meter queues were scanned more often than expected. [mcc] 14:19:11 2 Sep 2002 - Blue split Fixed old bug in Blue split (split-clip/all-layers in Dream family). The operation halted at the first selected track on which no clip was split. Higher numbered tracks beyond that were not split. Release 17.3.07a, September 2nd 2002 ====================================================================== Changes Since 17.3.06a ---------------------- [mcc] 17:25:10 29 Aug 2002 - Max varispeed The upper limit for varispeed is 12.5%, 2.04 semitones. The limit only applies to the console UI, the vari command is unaffected. Bugs Fixed Since 17.3.06a ------------------------- [mcc] 10:41:44 2 Sep 2002 - Merlin InputMon bug The new Merlin InputMon menu introduced in 17.3.03a had a serious bug that caused a whole zoo of faults in the console. [mcc] 14:44:19 1 Sep 2002 - dotted last buffer Fixed old bug in handling of read beyond end of waveform that in certain conditions caused last buffer of clip to fail to load, yielding "dotted" waveform. Release 17.3.06a, August 28th 2002 ====================================================================== Changes Since 17.3.05a ---------------------- [mcc] 18:30:00 26 Aug 2002 - AudioBase changes/fixes Various changes/fixes to both AB1 and AB2. Form maintenance bug: The state of the search form was not being saved & restored correctly when the domain was changed to/from search-project. This caused the software to become confused about the values and states of the fields in the form. This bug was introduced with search-project, it has always been there. The search forms for both domains (project & database) are now correctly and separately maintained. There are two forms. Changes to the search-project form will not affect the search-database form, for example. The forms are saved on the boot disk, and so are remembered across system restart. LINE field: The interpretation of the LINE field has been changed. The meanings of the inactive and ANY states have been reversed. When the field is inactive (ie, blank, green), the files referenced in the search results are NOT tested to see if they are online or not. This yields the fastest search. The results of the search will show the LINE value as "?", meaning "don't know". When the field is set to ON, OFF or ANY, each file is tested to see if it is online, yeilding a much slower search. If set to ANY, then all files returned by the search are included in the results, online or offline. Existence cache not cleared, larger: AudioBase keeps a cache of the status of files tested for existence. This is no longer cleared (forgotten) at the start of every search as it was in previous releases. Instead it is only cleared when the AudioBase window is invoked. This is much more efficient, but can get out of date. For example, if you remain in AudioBase menu while putting a disk online then files on the disk may still be noted as offline in the cache, and thus appear to be offline in subsequent searches. In such a case you need to clear the cache by leaving and returning to the AudioBase menu. The cache is larger and circular instead of finite. It now remembers the status of the last 8192 files tested (instead of the first 1024 files). Search for FILENAME in project: In the search-project domain, the FILENAME field is now a search criterion. You can, for example, search for an attached file by name. Or search for only borrowed clips by setting FILENAME to not-the-project: "~@". ("@" is the shortcut for the current project name, and "~" means NOT). [ajc] 00:50:06 23 Aug 2002 - Varispeed 96k Made a separate program for use at 96k which implements a different filter for varispeed SRC. For test purposes this is always installed by Merlin at the moment since audioman is not yet able to switch DSP programs when the sample rate changes. [ajc] 21:44:11 25 Aug 2002 - Varispeed mods Fixed bug in generation of cubic-spline interpolation table. Implemented PMC design method in genfilt. Varispeed SRC filters are now configured as follows: mdr2/mdr2n (@48k): 0-20000 flat +/- 0.25dB stopband rejection -100dB; cubic-spline interp mdr2n96 (@96k): 0-30000 flag +/- 0.2dB stopband rejection -100dB; linear interp mdr3/mdr3n (@48k): 0-20000 flat +/- 0.25dB stopband rejection -100dB; linear interp Note that the response at 44k1 will be scaled proportionally, hence flat to 20000*44.1/48 = 18375Hz, after which it rolls off steeply. . Fixed tbase to update pitch when varispeed turned off in jog/shuttle Bugs Fixed Since 17.3.05a ------------------------- [ajc] 00:48:30 23 Aug 2002 - Varispeed Fixed bug in varispeed input SRC which caused severe clicks & distortion at certain varispeed ratios. Also fixed bug in SRC for reverse shuttle. Release 17.3.05a, August 22nd 2002 ====================================================================== Changes Since 17.3.04a ---------------------- [ajc] 00:27:57 22 Aug 2002 - temporary debug Mixer Transport-Mode Notifications are echoed to TVT log. [ajc] 00:20:49 22 Aug 2002 - interpolation on SRC Added cubic-spline interpolation to SRC processing for varispeed. This is to reduce distortion at high frequencies resulting from jitter when no interp- olation is used. However DSP load is increased significantly. As a result current version does not work at 96k at all with varispeed ON. This version implements spline interp. on both record and playback SRC with high-quality (32 tap) filter. More research needs to be done to determine: a) minimum no. of filter taps required for adequate quality b) whether linear interp would also give satisfactory distortion at lower DSP cost. c) whether reduced quality SRC can be used for playback. Bugs Fixed Since 17.3.04a ------------------------- [ajc] 00:17:27 22 Aug 2002 - varispeed restore Corrected bug in TBASE which caused varispeed ON cmd to always be sent to DSPs after blue-stop. Correct state would be restored by switching varispeed on and off again. Release 17.3.04a, August 21st 2002 ====================================================================== [rod] Aug 21st 2002 - New Sony 9 pin machine Added "Sony" "Video Disk Recorder" "MSW A2000P" to sony_id Bugs Fixed Since 17.3.03a ------------------------- [ajc] Aug 21st 2002 - Meter Patching Fixed meter patching packet. [mcc] 01:06:22 Aug 16 2002 - Import ML with wrong pathlists ML files containing complete pathlists for audio files did not import correctly unless the pathlists were current and correct. That is, the files had to exist (and be online) at the pathlists defined in the ML file. Otherwise "Some of the audio...cannot be found". This has been fixed. It will now search elsewhere for a file if the ML pathlist is not correct, as it always did prior to 16.6.10 when this bug was introduced. Release 17.3.03a, August 13th 2002 ====================================================================== Changes Since 17.3.02a ---------------------- [ajc] 23:21:37 12 Aug 2002 - Studio Monitor Mix Matrix Added additional mix matrix for studio monitoring. Enabled by an entry in aman_cfg which specifies number of outputs. A corresponding DSP mixer resource with this number of outputs must be available. All Dream/Station cfgs now have an 8x2 or 10x2 matrix reserved for stmon. [ajc] 23:23:44 12 Aug 2002 - Mixer Meters Increased number of mixer meters to 32. Latest MixerProtocol (8.6) redefines mixer data packet to have one more meter address bit and one less data bit. The change is backwards compatible. Mixer meters now handled by MIXINT as code in PLAY/INT now works in VSN domain, but mixer meters are still MSN domain (see bugfix below)! Optimised PX processing mixer meters considerably, as the old code in PLAY/INT was doing a lot of processing which was not required by the mixer. [mcc] 11:12:32 12 Aug 2002 - extended monitor modes Changed order and names of monitor modes to be more intuitive. 0 INPUT old SOURCE mode 1 AUTO old TRACK mode 2 RECORD input in Record only (muted in stop) 3 MUTE muted in Record 4 REPRO monitor track playback in all transport modes For all consoles, the default monitor mode at startup is INPUT. The meter window now displays the current monitor mode. The Merlin InputMon key is now a hyper mode, with a simple UI to conrol extended monitor modes. It should be self-explanatory. Pressing the key toggles between INPUT monitor mode and one of the other four monitor modes (default AUTO). While the key is depressed, the mode can be explicitly selected with the softkeys. [ajc] 23:51:59 11 Aug 2002 - extended monitor modes Implemented Extended Monitoring Modes in MDR DSP program. [mcc] 17:04:55 10 Aug 2002 - extended monitor modes Arm/Monitor provides 3 new modes in addition to SOURCE and TRACK, and is now a choice (choose mode via jog or number pad, press Enter). As a convenience, double-clicking Monitor reverts to previous selection. The new modes are AUTO IN, MUTE IN, and REPRO. Merlin does not yet have this feature. In 17.3 additional DSP mods are still required before the new modes can be used. Bugs Fixed Since 17.3.02a ------------------------- [ajc] 23:28:42 12 Aug 2002 - Mixer Meters Fixed operation of mixer meters which were broken by varispeed submission. Symptom was that meter display was out of sync with audio. [ajc] 10:42:41 12 Aug 2002 - Monitoring bug Fixed short burst of playback when going into PLAY after switching monitor mode from TRACK to SOURCE whilst in STOP. [mcc] 08:31:19 11 Aug 2002 - THRU tracks after Bounce Tracks in THRU mode are now correctly restored after Bounce and Fade-range. Release 17.3.02a, August 8th 2002 ====================================================================== Changes Since 17.3.01a ---------------------- [mcc] 10:07:11 8 Aug 2002 - vari The "vari" command is installed by direct build (make install) but is not included in the release. This command can be used to test varispeed on any product, not just Merlin. It replaces the old "setvari" command which will no longer work properly, and which is deleted when this version is installed. Varispeed is now enabled only by Varispeed dongle, R&D dongle is not required. [ajc] 01:02:17 8 Aug 2002 - Varispeed Ratio updates in SHUT/SPOOL modes Propagate varispeed updates immediately in all transport modes. [ajc] 01:01:37 8 Aug 2002 - Mixer Notification Implement MX_NOTIFY packet to inform mixer of transport mode changes [mcc] 14:31:37 7 Aug 2002 - import disables track groups Track group selection is disabled while Import file is open. [mcc] 07:09:44 7 Aug 2002 - station startup The stn startup script now uses 48tk project format for all configurations. The configurations available are... stn 24: 24/8/12 stn 32: 32/8/16 (2+ QDCs required) stn 48: 48/8/16 (2+ QDCs required) The default configuration (no parameter) is stn 48. [mcc] 13:44:19 2 Aug 2002 - Varispeed The VariSpeed menu for Merlin is nearly complete. Only the choice of fixed ratio is still missing. Both the R&D and Varispeed dongles are required. [mcc] 13:44:19 2 Aug 2002 - ViVid support removed Support for ViVid has been removed. [ajc] 00:54:48 31 Jul 2002 - Varispeed Various fixes; fixed possible buffer wraparound bug (-> clicks & distortion); Reduced max playback speed for programs built with varispeed enabled. This is necessary to avoid catastrophic DSP overloads requiring blue-stop. LoopJog now works as expected in Varispeed mode. Note: changing varispeed ratio now causes actual loop length to vary since it is specified in terms of project timecode (frames)! Perhaps this should be changed, but how? Bugs Fixed Since 17.3.01a ------------------------- [mcc] 11:14:55 29 Jul 2002 - Satellite mouse FMC V1.0 build 123 starts automatically in fullscreen mode, with no title bar buttons. Thus various problems are fixed that were caused by clicking mouse outside FMC window, or by clicking the FMC kill button. Caveat: the mouse scaling may not be correct until after the first power cycle following installation of this version. Release 17.3.01a, July 26th 2002 ====================================================================== Changes Since 17.3.00a ---------------------- [rls] 18:10:28 25 Jul 2002 - DMAN_SONY Added ALT_JOG env variable to Send Shuttle cmds instead of Jog in Jog mode This is needed to make the Avid News Cutter work in JOG mode. [rls] 15:27:28 25 Jul 2002 - MADI 24 channel Added AMAN environment variable MADI96K_24PER If SET and SR=96k then the MADI frame will be 28 channels long carrying 24 channels of 96K audio. A second MADI card will carry I/Os 25-48. MADI96K_LEGACY can also be set to get a 56ch frame with 24ch of 96k. This allows two separate 24 track MADI machines to be connected to an MFX with two MADI cards at 96K. [ajc] 01:44:28 25 Jul 2002 - Varispeed Initial release of varispeed code. UI not yet implemented but you can set the varispeed ratio using "setvari " where is relative to 1.0. To switch varispeed off use @setvari -f". [ajc] 23:21:47 24 Jul 2002 - tclib Moved tclib sources to global libraries directory (no longer MDR-specific). Also moved tclib.h to global MFX/tclib.h. All modules now use the runup/rundown functions in tclib except the MFX console code (68000) which is in MFX/CONSOLES/CONLIB/. Bugs Fixed Since 17.3.00a ------------------------- [ajc] 00:48:49 23 Jul 2002 - Rundown Error Fixed arithmetic error in the rundown() function used for converting timecode at 29.97 frame rate. This error resulted in a result 1 TUNIT too high in some cases; in most cases this would have no further ill effects. Also added inline versions of runup/rundown to tclib.h. [mcc] 18:30:14 22 Jul 2002 - ML4 import 30ND A typo in the ML4 parser prevented import of files with 30ND framerate. This has now been fixed. An emergency workaround is to change the framerate in the ML4 file to say "RF30ND" instead of "FR30ND". Release 17.2.09a/17.3.00a, July 22nd 2002 -- 17.2 BRANCHED HERE ====================================================================== Changes Since 17.2.08a ---------------------- [rls/cea] 13:23:56 19 Jul 2002 - SONY Slave ID Chris added code to set frame rate bits when sending the Fairlight device ID Also fixed bug when parsing sony_id file. Comment lines were not ignored! Added *@SONY-OLDID in tcs_cfg to disable the new frame rate bits in the ID [mcc] 17:23:56 18 Jul 2002 - SysStartup postponed until MDR ready Audioman will now wait until MDR is ready before sending SysStartup(63) to mixer. Bugs Fixed Since 17.2.08a ------------------------- [mcc] 12:41:22 17 Jul 2002 - system hang changing frame rate When changing frame rate from a larger to smaller timecode 'day' (eg, from 24 to 30DF) the system would sometimes hang if the transport position was in the dropped region before midnight. The track display would freeze, showing blank timecode position. For example, while in 24fps, goto 23:59:59:0 and then either select 30DF explicitly, or open a 30DF project. [mcc] 15:12:49 15 Jul 2002 - WAV import bugs Fixed bugs introduced into WAV import when MDR software was converted to C++ in 16.6.02a. In particular, BWF files were not being correctly recognised. Release 17.2.08a, July 12th 2002 ====================================================================== Changes Since 17.2.07a ---------------------- [rls/cea] 13:02:17 11 July 2002 - Midi Time Code generator Implemented a basic Midi Time Code (MTC) generator. The "Gen" button on the console controls both timecode generators (LTC and MTC). Also the "Gen" Offset is applied to both LTC and MTC. MTC quarter frame messages are generated while in Play or Record mode. When not in play/record a full MTC SysEx message is sent every 400ms. The Midi UART port is used directly by the MTC generator, so don't try to use /midi (on PX) while in MDR. [steve] 10:22:17 09 July 2002 - /nvr/setup defaults Changes to the default /nvr/setup file have been done in an attempt to make systems more "turn key" for end users. Bugs Fixed Since 17.2.07a ------------------------- [mcc] 11:25:19 12 Jul 2002 - Windows dfnserver Some redundant overhead has been eliminated from Windows versions of dfnserver and file_gopher. The processing of the dfn_cfg file has been enhanced. NETWORK: The default NETWORK is now 10.1.1.0. Each NETWORK defined in dfn_cfg is checked to ensure the server has an IP address on the specified subnet. If not, an error message is printed and the subnet will not be scanned. If at least one valid NETWORK is not defined, dfnserver will print an error message and exit. MEDIA_VOLUMES: The drives to be included in MediaLink file system can be explicitly specified with this new keyword. The syntax allows a range of drives to be specified, for example: MEDIA_VOLUMES E J M-R HIDDEN_VOLUMES: The drives specifed will never be included in MediaLink file system, even if they are also specified as MEDIA_VOLUMES. If no MEDIA_VOLUMES are specifed, then HIDDEN_VOLUMES behaves as it always has. In particular, an old dfn_cfg file will still result in the same drive selection. The syntax now allows a range of drives to be specified. [mcc] 12:21:52 11 Jul 2002 - ML4 import Fix two recently introduced ML4 import bugs: Version 4.1 ML4 files were not accepted. Introduced in 17.2.04. Uninitialised file_status could cause audio files to fail to attach correctly. Introduced in 17.2.02. [mcc] 13:52:36 07 July 2002 - More copying directories on dfnserver Corrected conversion of NULL path to "" in previous fix. This bug was more of a loose cannon than indicated below. It was not directly related to copy at all. The two functions affected were those to add a new file to dfn cache, and to find a file in the cache. These are used by many different operations. Whenever the pathlist length was a multiple of the allocation quantum (probably 4, 8, or 16 bytes) then a byte was being cleared beyond the allocated buffer. The effect this would have was entirely in the hands of the gods. [awk] 04:27:11 05 July 2002 - Copying directories on dfnserver The following copy would cause dfnserver to crash on a medialink machine runing WINDOWS 2K. From the m4 browser a copy from scxx/dirA/ to scxx/dirb/dirc/. Copies involving different path names would work ok. The problem was caused by the heap becoming corrupted. The corruption only seemed to affect particular W2k machines but never the less other W2k machines, NT4 and OS9 machines would also have been affected. The results being totally unpredictable. Release 17.2.07a, July 3rd 2002 ====================================================================== Changes Since 17.2.06a ---------------------- [mcc] 16:31:19 1 Jul 2002 - explicit input type selection A method to explicitly select input type has been implemented in MFK and Satellite input patching menus. While Patch or Group operation is active, pressing a number-pad digit will select the corresponding input type: 0=analogue,1=digital,2=madi. If there are no inputs of that type then there is no change. [mcc] 11:11:51 25 Jun 2002 - Windows file gopher Some of the recent changes to os9 file gopher have been ported to Windows file gopher: The file type/suffix is only evaluated once. The attempt to read raw HFS resource fork has been removed. This did not work properly anyway and is no longer necessary thanks to 3rd-party apps like Macopener, MacDrive5 etc. [mcc] 08:09:43 25 Jun 2002 - New24 for mixers The Project/New24 softkey is not disabled for Dream mixers. [ajc] 22:42:33 24 Jun 2002 - ajc: MADI I/O mapping Implement mapping of user I/O numbers to hardware channels for MADI. From user perspective MADI I/Os are now always consecutive and change according to the number of channels per MADI port. [rls] 14:46:59 24 Jun 2002 - MADI 96K Legacy framing mode. MADI 96K Legacy framing mode is controlled by environment variable: setenv MADI96K_LEGACY If SET and SR=96k then the MADI frame will be 56 channels long carrying 28 channels of 96K audio. Each 96K channel uses two consecutive MADI channels. MADI96K_HIGHDENSITY can also be set to get a 64ch frame with 32ch of 96k. Also changed default setting to 28 channels at 96K The variable MADI96K_HIGHDENSITY must be set to get 32 channels at 96K. The variable MADI_HIGHDENSITY must be set to get 64 channels at 48K. It is preferable to add these to /nvr/aman_cfg [ajc] 16:11:59 14 Jun 2002 - MADI outputs in MFX multitrack mode Tracks now assigned correctly to MADI outputs at 96k (28 or 32ch frame). MADI inputs are still numbered 64 per card. Bugs Fixed Since 17.2.06a ------------------------- [ajc] 11:05:11 26 Jun 2002 - dongle buss count Number of main busses available now limited to dongled value [mcc] 09:27:11 26 Jun 2002 - file browser reset A very old bug has been fixed that caused the file browser to reset. The browser view would "collapse" and pop to the root of the network. The bug was in the Windows version of dfnserver and manifested only on servers with more than one IP address. The bug could strike whenever the application (as opposed to the user) set the browser position. It was first noticed when creating a new folder on the server. Once the folder is created, the application is supposed to move the browser position into the new folder, but instead the browser would reset and pop the view back to the root. [mcc] 17:01:23 21 Jun 2002 - ML3 import bugs A null name for CHANNEL, WAVE or MARK caused a fatal error, and the entire file to be rejected. This no longer occurs. Substitution of underscore for wildchars was being performed for names other than clip (WSEG) names, and should not have been. [mcc] 11:17:19 17 Jun 2002 - stereo WAV on track 48 Fix old bug, fatal exception caused by interleaved stereo clip on track N (the last available track, where there is no track N+1). Stereo grouping of track N would also cause the exception. [ajc] 16:11:31 14 Jun 2002 - MADI 96k mode Corrected operation of MADI at 96k Release 17.2.06a, June 14th 2002 ====================================================================== Changes Since 17.2.05a ---------------------- [ajc] 14th June 2002 - MADI MADI should now be working in 64 channel mode. As the DSP programs have been changed substantially testing should not just concentrate on MADI operation, but also include complete system (esp. audio) integrity! It would also be a good idea to test Dream & Station with MADI cards fitted... even we aren't selling this cfg at the moment it is sure to come soon. Testing should include fitting the MADI card(s) on various QDCs, as not all DSPs are running the same programs. Bugs Fixed Since 17.2.05a ------------------------- None Release 17.2.05a, June 13th 2002 ====================================================================== Changes Since 17.2.04a ---------------------- [ajc] 01:13:47 13 Jun 2002 - qdcdump Qdcdump now has "Symbol.table" cmd which loads the symbols for the DSP program running on the currently selected DSP. The name of the program is stored in IRAM by qdcrun and audioman at the time it is loaded. [ajc] 01:13:05 13 Jun 2002 - Feed FX Mode Feeds in FX Groups now use FX mode of first feed in group. [ajc] 01:11:47 13 Jun 2002 - 64 Channel MADI fixes Fixed level ramping bug; fixed DSP input patching in mixer programs. Updated usercode of clock CPLDs to ensure an update is forced by sysinfo. [mcc] 16:55:22 12 Jun 2002 - diskinit v2.2 Diskinit is built using defs from LIBS/FLFS/DEFS. The old sources & defs in KERNEL/IO/FLFS are now unused. [ajc] 01:54:51 12 Jun 2002 - 64 Channel MADI MADI card now supports up to 64 channels at sample rates of 48k or less. At Fs=96k the number of channels is automatically reduced to 32 max. At the moment the MADI channel numbering is fixed at 64 channels per card, so if two cards are installed the channels on the second card will start at 65. This issue is yet to be addressed. NB: only rev 4+ QDCs will support MADI 64 channel mode. Bugs Fixed Since 17.2.04a ------------------------- [spr] 12:47:18 12 Jun 2002 - Mediad/DFN error/warning Mediad now only informs DFN about device and target removals if mediad told DFN about the device in the first place. This is part of the phantom FAT32 device fix submitted by mcc below. [mcc] 12:27:18 12 Jun 2002 - phantom FAT32 device FAT32 partitions were not being mounted correctly by dfnserver. The result was a unit with no target device and target device with no unit. The target was displayed as a "phantom", and the unit would cause problems if it went offline or was reinitialised. Release 17.2.04a, June 11th 2002 ====================================================================== Changes Since 17.2.03a ---------------------- **************************************************************************** * Please insert new entries at the top of this section, not at the bottom. * **************************************************************************** [initials] describe the change here, please. Bugs Fixed Since 17.2.03a ------------------------- [mcc] 11:05:21 11 Jun 2002 - ML bugs Revision -------- ML revision numbers have been incremented due to bug fixes. ML3 revision now 3.1. ML4 revision now 4.1. ML Import bugs: -------------- Waveframe & AKAI Changes to omdl.h in 16.6.13a introduced faults in Waveframe and AKAI import. These have been fixed. At the same time, several other bugs were detected and fixed. Files >= 2Gb Incorrect use of signed storage type introduced potential bugs for 2Gb+ files, except for MFX project files. Object & Mark numbers Incorrect use of 8bit storage type caused import of mark and object numbers > 255 to fail. ML4 parser missing prototype No prototype for ceil() caused error for WAVE size. ML export bugs -------------- Inappropriately signed values. Use of %d instead of %u caused the following values to be incorrectly shown as negative if >= 2^31: ML3 OBJECT size & offset MARK time. ML4 fileseg start & size file cookie & size Release 17.2.03a, June 6th 2002 ====================================================================== Changes Since 17.2.02a ---------------------- [ajc] 01:44:26 6 Jun 2002 - mods to tools tree - Mods to esplib for win32 compilation; esplib.dsw/esplib.dsp added. - Use gcc3 & Qt3 for compilation. This involves setting up the environment in fairlight_env. To switch back to an older tree it is now necessary to reset QTDIR to /usr/local/qt. Still investigating ways of automating this! [rls] 15:22:14 31 May 2002 - AO1 Xilinx Rev 15 again Aout_ctrl Xilinx rev 15 - Changed watchdog timeout flag. Now the watchdog is fed before reading the timeout flag. Timeout flag clears when read. [mcc] 13:15:00 31 May 2002 - dfnserver -u option The dfnserver -u option causes the dfn_cfg file to be read from the user's home directory. This is only for Windows version. The USERPROFILE environment variable is assumed to be the home directory, eg, "C:\Documents and Settings\user". [mcc] 19:56:43 30 May 2002 - undo/redo counters displayed The undo/redo counters are shown on the track display. [rls] 18:22:14 30 May 2002 - AO1 Xilinx Rev 15 Aout_ctrl Xilinx rev 15 - Divider control bit for 96k so that watchdog timeout can be adjusted to compensate for double clock. Timeout is now 680ms at 48k and 96k. Bugs Fixed Since 17.2.02a ------------------------- [mcc] 16.35:17 5 Jun 2002 - move,rename open files FLFS now also prevents moving or renaming of open file. [mcc] 13:07:14 4 Jun 2002 - delete open files FLFS file mgr allowed open files to be deleted. This serious bug is thought to have been responsible for corruption of some disks. It has now been fixed. An attempt to delete an open file will return error #253, Non-Sharable File/Device Is Busy. [mcc] 10:58:22 1 Jun 2002 - read-only objects Open objects in read-only project files are now correctly closed when the project is closed. If a project is read-only, QDC now sets bit6 (0x40) of the object presence flag in the ObjectPresent packet. The object can still be opened but cannot be updated. If the object is not present, then a new object cannot be created. Release 17.2.02a, May 28th 2002 ====================================================================== Changes Since 17.1.12 --------------------- [mcc] 12:04:55 May 27 2002 - os9 file gopher Some redundant and inefficient methods have been corrected in os9 file gopher. [mcc] 12:31:19 May 25 2002 - MDR set trackname Added MIXLIB mdr_trackname() to set track name or seed name. [steve] Friday May 24 13:02:31 - Analogue Output Relays The AO1M cards installed in a system will now cut out when the system crashes. This is a combination of both Xilinx firmware and software. [mcc] 11:49:22 24 May 2002 - Aiff file support Aiff files are now supported sufficiently to be used for ML import. They are recognised and installed in the MediaLink cache, and are accepted as attachments to a project file. They do not yet have cookies, so beware, filenames must be unique and files must not be renamed after attaching. Because cookies get "locked in" once defined, I don't want to define aiff cookies until certain of the method to be used. The file suffix must be .AIF. ML3 files referencing aiff files should specify AIFF file type, eg: FILE 1 { name = "filename.AIF" type = AIFF } [rls] 11:42:14 24 May 2002 - QDCINIT I/O printout The number of AES inputs is now listed correctly by qdcinit when there are more than 32 inputs installed on a QDC. [ajc] 01:45:29 24 May 2002 - TSB Decouple option Added option to decouple TSB I/O for primary and secondary DSPs. This is required in some DSP configurations to avoid TSB-related clicks. Option is enabled by specifying a '=' separator instead of ':' on the program load line in aman_cfg. Currently used in Dream48 only. [ajc] 01:45:01 24 May 2002 - high density DSP I/O Preliminary changes for high density DSP I/O to support 64 channel MADI. [ajc] 01:44:42 24 May 2002 - XPC server Added xpcserve to /dd/startup. [rls] 18:42:14 23 May 2002 - AO1M relay hardware watchdog Revision 14 of the AO1 xilinx now has a watchdog which opens the relays after 341ms at 96k (682ms at 48k). [mcc] 15:02:14 18 May 2002 - 17.2.01b: new FLFS FM reinstated The revised build of FLFS file mgr is reinstated after big fright in 17.1.12a. Please flfscheck your FLFS disks before installing this version to ensure the file system is intact to start with. The checking of file RW attributes is once again enabled by default. [steve] 17th May - 17.1 Branched Here ...... [mcc] 18:26:11 16 May 2002 - embedded objects Ensure all paths are eventually closed. Removed unnecessary truncation of transfer in eos_write_blocks(). The "space" arg for ObjectOpen now only applies to creation of new object. New objects are not copied to temp file. Bugs Fixed Since 17.1.12 ------------------------ [ajc] 21:45:10 23 May 2002 - buss mixer failure Buss mixer input was not being disabled when deactivating buss. This caused the buss mixer to malfunction after a few buss group reconfigurations. [mcc] 12:28:19 23 May 2002 - stuck transport mode Added some test code to diagnose and repair stuck transport mode. When a transport mode gets stuck, press Blue-PlayMenu and see if that fixes it. Please report the tvt info printed by consoled to me. [mcc] 11:47:25 22 May 2002 - ML3 parser: wrong interleaved Fixed bug in parsing WAVE format string: The interleaved channel count (3rd char of format string) was only converted to integer if old version 'N' or 'I' not new '1' or '2'. Channel count was thus 49 or 50 instead of 1 or 2. This caused unnecessary copy of borrowed audio during import. Also fixed fatal debug bug: If MDR U option set then some ML debug could crash the system. [awk] 18:25:04 21 May 2002 - anetd -z option (WINNT only) This option did nothing on the WINNT version of anetd.exe. It will now display debug messages as advised in the help screen. [mcc] 18:43:19 21 May 2002 - ML3 import null clipname Fixed ML3 parsing bug that caused fatal error when WSEG name = "". There was already code in place to substitute "Wave #" for null clipnames, but this didn't get a chance because the parser bombed out first. [mcc] 18:43:19 21 May 2002 - ML import multiple errors for same file Fixed ML import bug that would repeatedly try to attach a file that cannot be found, once for every waveform that referenced the file, perhaps thousands. [mcc] 09:10:45 20 May 2002 - diskinit RBF bug Fixed old bug in diskinit that affected RBF (only) when disk capacity less than 4Gb and -z option not used. The bug caused the partition size to be wrong sometimes. Release 17.1.12a, May 15th 2002 ====================================================================== Changes Since 17.1.11 --------------------- [mcc] 07:43:19 15 May 2002 - revert FLFS to old sources FLFS file mgr is reverted to old sources because a bug is suspected. Also RW attr checking is disabled by default (osk G option inverted). Bugs Fixed Since 17.1.11 ------------------------ [mcc] 18:01:32 15 May 2002 - hangup in project close A bug in embedded object mgt left bit(s) set in coarse dirty bitmap. This caused MDR to hangup forever when closing the project file. A possible fix has been implemented, plus a safety net workaround if the fix doesn't work. Release 17.1.11a, May 14th 2002 ====================================================================== Changes Since 17.1.10 --------------------- [ajc] 14:25:04 12 May 2002 - Audioman XPC Server XPC Interface to audioman now implemented (experimental) and available for testing. XPC Server process also implements mixer coefficient update via XPC. Interface specification in $FAIRLIGHT/DEFS/XPC/qdcfuncs.h & -/amanfuncs.h. [ajc] 14:23:30 12 May 2002 - XPC Generator Initial release of the XPC (eXternal Procedure Call) generator program (xpcgen). This version generates Client and Server code for function calls with simple, array, structured and pointer types as parameters and return value. More complex types such as variable length arrays and C strings are not yet implemented. Return of data via pointer parameters (out & inout parameters) is working. Limited C++ parsing implemented. Output is C code which should also be C++ compatible. More detailed documentation to follow! [ajc] 14:21:06 12 May 2002 - Buss Direct Out Added direct out property to busses. This can be attached pre/post fader, pre/post dynamics, pre/post insert. [mcc] 11:52:13 10 May 2002 - FLFS The FLFS file mgr and dynamic boot rom now use the new consolidated FLFS libraries. Version advanced to 17.1.10b. [steve] Thursday May 9 16:14:05 EST 2002 - Deleting Directories Problems with deleting directories using espdeldir,anetd and ffs have now been fixed. All file managers have been tested and fixed. [steve] Thursday May 9 15:54:20 EST 2002 - ROM Boot SCSI Port A serious bug in the ROM has been found and corrected. The problem only happened when the boot device was set to any plugin SCSI controller. The effect was that the boot disk would have been corrupted when being accesed by two device names. i.e. /dd and /sd21. This problem does not effect boot disks that use the onboard SCSI controller. [rls] 17:40:12 8 May 2002 - MADI card support Revision 2 of MADI xilinx - Added Reset control in AMAN and Mute status register bit. [mcc] 19:15:10 7 May 2002 - diskinit V2.01 Diskinit options for FLFS: -m=size - partition size in sectors (only useful for testing) -bb - Clusters are filled with 0xdd while scanning for bad blocks Updated version to V2.01 only to mark these changes, as well 24 April mods. No change is intended or expected in the layout of the disk. [mcc] 18:29:55 7 May 2002 - dfn_cfg Removed uncommented modhist block. Removed obsolete DOMAIN (unused since Rev14). Removed MAX_FILES from Windows version as this now only applies to os9 system. [rls] 17:50:12 7 May 2002 - MADI card support Added control of MADI card registers in AMAN and qdcinit: Output muting is set on or off Frame length set to 56 channels (or 32 for 96k) CSB is loaded and working. [mcc] 10:32:12 6 May 2002 - new FLFS libraries A new consolidated version of FLFS lives at MFX4/LIBS/FLFS. Two libraries are compiled from the same files in SRC: flfs_fm.l - library for OSK FLFS file manager flfskit.l - library not for file manager So far only flfscheck (V3.04) has been modified to use flfskit.l. The file manager does not yet use flfs_fm.l, but soon will. Apart from flfscheck, there is no significant change to release software. [mcc] 21:24:42 5 May 2002 - prohibited filename chars The following characters are additionally prohibited when entering new file and folder names from all consoles: <> angle brackets "' double and single quote ` single backquote : colon [mcc] 13:21:14 4 May 2002 - FLFS file permissions Added read/write permission testing to FLFS file mgr. Prior to this, the read & write permissions of files and directories meant nothing. Now FLFS will only allow files with write permission to be opened for write, deleted, or renamed. A directory must have write permission to be deleted, renamed, or have files created or deleted within it. [rls] Friday May 3 11:47:59 - MADI Xilinx MADI transmitter and receiver now working in 32 channel mode. No CSB support yet. Bugs Fixed Since 17.1.10 ------------------------ [ajc] 14:17:57 12 May 2002 - Feed Dynamics Disappearing [?] Changed allocate_fxgroups() to not abort if a group can't be placed. This is only correct when a single group is being allocated; possibly explains the mysterious "disappearing feed dynamics" problem which has been reported but not precisely characterised yet. Needs more testing. Release 17.1.10a, May 1st 2002 ====================================================================== Changes Since 17.1.09 --------------------- [djs] The OSK R option has been inverted again. Set R option to enable early responses from ANETD for a small performance gain at the expense of error reporting (not recommended). Bugs Fixed Since 17.1.09 ------------------------ [mcc] 11:29:17 May 1 2002 - anetd delete folder bug A very old bug in anetd (os9 version only) caused deleting of a folder to fail. This resulted in the user being unable to delete a remote folder on another QDC (or mfx3). It is uncertain whether the bug was specific to FLFS disks, however the fix has been tested with both FLFS and RBF, but not FAT32. To take effect, the fix to anetd must be installed on the remote node. In other words, it is the machine where the folder lives that needs the fix. Release 17.1.09a, April 30th 2002 ====================================================================== Changes Since 17.1.08 --------------------- [rls] 11:30:57 30 Apr 2002 - qdcinit MADI outputs now patched in 32 channel mode. [rls] 13:51:57 29 Apr 2002 - qdcinit MADI inputs now patched and working correctly in 32 channel mode. [steve] April 29 11:54:52 EST 2002 - Noded Scan Range The noded scan range is now adjusted from the /nvr/setup file. To change this, enter "setup" at a shell prompt and add a line like the following in the [Web] section ... SCANHOSTS = 1-32,254 This will scan IP addresses 1-32 and then 254 in the current class C subnet for other QDC's, MFX3+'s and Medialink servers. If the SCANHOSTS directive is not present, the entire class C subnet will be scanned as before. [mcc] 12:13:57 26 Apr 2002 - fslib Enhanced efficiency of free space update. Added new function to fslib to get free space only, not all the other stuff, and changed dfnserver (os9 only) to use this to service request to update the free space for a unit. Fixed some error recovery bugs in fslib that left paths open, and memory un-freed. [mcc] 21:07:19 18 Apr 2002 - MADI input patching MDR input patching UI supports MADI inputs in Merlin, Satellite and MFX/MFK consoles. Bugs Fixed Since 17.1.08 ------------------------ [mcc] 16:11:17 26 Apr 2002 - fix diskinit version stamp Fixed bug just introduced by 24 April mod, version stamp was 0.0. [mcc] 09:57:44 24 Apr 2002 - obscure diskinit/flfs bug fixes Obscure bugs fixed - they were unlikely to be causing a problem. An endian swap was neglected but the case seems to only occur for real NTFS disks. [ajc] 22:47:24 23 Apr 2002 - Monitor/Playback Ramp Clamping; Bug Fix Clamp Ramps to correct small gain error introduced by monitor 'Pops' mod. Also corrected bug in monitor SOURCE mode (stereo-mixed-out mode only) introduced by the same mod. [mcc] 14:51:22 19 Apr 2002 - Wrong environment for 'be' startup The environment was not correct when using the 'be' startup method. As a result the mshell was not being used by the gfx window mgr. This was a minor fault and probably of no consequence. [mcc] 14:51:22 19 Apr 2002 - mshell writes .history during power-down The mshell option to save/load the .history file was set. This could cause the mshell to try to write to the file during power-down sequence, which is a very bad thing. The mshell -s,-r & -h options are no longer enabled by USR/.login. Release 17.1.08a, April 18th 2002 ====================================================================== [steve] April 18 11:38:24 2002 - WIN32 Executables This is the first build that the WIN32 executables have been build from a checked out tree using CVS/VC combination. Please check medialink software works as expected. [mcc] 12:41:19 17 Apr 2002 - pfix V2.0 The pfix command is reinstated into the release. This is Version 2.0. It has been ommitted for some time waiting for an upgrade to correctly handle the new 24/48-track file format. This has now been done. [ajc] 00:13:53 15 Apr 2002 - Check TSB capacity; fix dream48/48/48 The Dream 48 configuration was using too many TSB slots, so it didn't work at 48k unless "fast" QDCs (ie with -15nS DPRAMs) were fitted. This is the probable cause of the "buss-mixer" problem observed two weeks ago. Some changes have been made which should make Dream 48 work on "normal" QDCs (-20): excess resources have been removed from aman_cfg entry to reduce TSB usage; unused DSP input resources (ie DSPs with no I/O card fitted) no longer use up TSB slots; maybe the number of MS-decoders (96) can be reduced?! Also, audioman now checks TSB usage whenever sample rate changes and writes a big error message to the TVT if the TSB capacity has been exceeded. NB: should only ever happen during alpha testing!!! [djs] If the osk R option is set, ANETD will use a slower, more robust method of handling and responding to write requests. This change has been made in an attempt to correct errors with backup and embedded objects. [ajc] 15:04:13 9 Apr 2002 - Feed FX Groups Audioman modified to correctly handle Feed FX Groups of any width up to 7.1. [ajc] 15:03:50 9 Apr 2002 - MADI Preliminary support for MADI card added to qdcinit, audioman [rls] 17:51:22 9 Apr 2002 - AES2 card support for production version. Updated AES2 Xilinx to Revision 2 for production AES2 cards with XCS20XL parts Modified setup of AES2 card in QDCINIT to support production version. The prototype AES2 cards are no longer supported (only 4 built). [mcc] 14:21:22 5 Apr 2002 - revise stereo group method TEST: A small change to stereo groups has been made, please check that stereo clips are still recorded on 2-track edit groups. [mcc] 14:21:22 5 Apr 2002 - refresh file size If osk E option is set (opt k e) then getstat SS_Size on FLFS disks refreshes path descriptor size from file. If this results in any change then this tvt mesg is printed: "REFRESH SIZE TO " Bugs Fixed Since 17.1.07 ------------------------ [mcc] 09:03:55 16 Apr 2002 - Blue-stop disarms Thru mode tracks Blue-stop did not correctly disarm Thru mode tracks if no other tracks were armed in Ready mode. The audio stopped but the tracks were left semi-armed. Now all tracks are fully disarmed by Blue-stop. [ajc] - 15:03:11 9 Apr 2002 - Monitoring 'Pops' (ported from 16.6) Implemented constant amplitude crossfade for monitor/playback switching in DSP code. This resolves audible artefacts observed when monitoring switched from source to track on punch-in/-out. [steve] Monday April 8 13:45:07 EST 2002 - SYSINFO Sysinfo used to silently ignore CPLD's in the system that were blank. This has now been corrected so they are displayed just like any other CPLD that is out of date. Release 17.1.07a, April 5th 2002 - NAB 2002 Release ====================================================================== Changes Since 17.1.06 --------------------- [rls] MADI card Added JTAG support for the MADI card [djs] Embedded Object Server The EOS implementation is now complete. The mixer can extract and embed objects in the project via the FAME protocol. This enables title load and save in Station/Dream. Bugs Fixed Since 17.1.06 ------------------------ None Release 17.1.06a, April 3rd 2002 ====================================================================== Changes Since 17.1.05 --------------------- [ajc] 22:58:19 29 Mar 2002 - Stereo Linking for Dream Renamed DefineDynamicsGroup->DefineFXGroup. Made mixint issue LinkFeed requests for stereo linkage as DefineFXGroup not implemented in audioman yet. [spr] Thursday March 28 11:17:45 EST 2002 - Dynamic ROM The Dynmaic ROM is now compiled in ANSI mode. Before it was in "backwards compatible mode". No bugs were found during the conversion. Bugs Fixed Since 17.1.05 ------------------------ [mcc] 17:32:05 25 Mar 2002 - unlink stereo input patch Fixed bug that swapped left instead of right input when unlinking stereo group. Release 17.1.05a, March 21st 2002 ====================================================================== Changes Since 17.1.04 --------------------- [mcc] 10:18:51 20 Mar 2002 - Serato V1.1008: dialog mode Serato algorithm for dialog has been introduced. The original algorithm is for music. The Wave menu of all consoles has been changed as follows: Algorithm softkey selects Serato M (music) or Serato D (dialog). Scale softkey selects quality 1 (highest) to 3 (lowest). WARNING: although this seems to work, there is a small but worrisome bug that smells like memory corruption and is still being investigated. The bug does not seem to affect the audio, only the debug output. PLEASE TEST: compare the results of SRC (VARI) operations using Serato and original Discrete transform. We need to determine if there is any point using Serato to perform SRC, because it is many times slower than the Discrete module. Theoretically there should be no difference in the output, because SRC is not something that can be done with greater quality, it is either done correctly or not. Note that the VARI submenu will use Serato if the Algorithm is set to Serato, even though the Algorithm sofkey is not present. To use Discrete SRC, just set Algorithm to not Serato. [cea] 19:13:20 19 Mar 2002 - Allow for V363 EPC controller The diagnostics, PX and PY roms, and pcilib.l have been changed to also accept the V363 pci controller as a replacement for the V360. [mcc] 16:22:11 18 Mar 2002 - stereo input patching In Dream, MDR should now record stereo clips on 2-track groups. [mcc] 10:01:55 16 Mar 2002 - Disabled ProTools ProTools is disabled. No source files have been removed yet. [mcc] 14:01:44 13 Mar 2002 - play menu is hyper mode The Play menu in Dream-family consoles is now a hyper mode. [mcc] 14:01:44 13 Mar 2002 - Clip/Range key LEDS The Clip/Range key LEDS now reflect the range state. [cea] 19:55:04 12 Mar 2002 - $MFX4/LIBS/MFXPROJECT Created the initial version of mfxproject.lib. Currently only tested for compilation under WIN32. [mcc] 17:04:55 12 Mar 2002 - range style In Dream-family consoles, the Setup/Range softkey is gone. The range style is now permanently MANUAL with the same-mode-key-toggles-range behaviour disabled. [mcc] 09:46:31 12 Mar 2002 - track group selection The track group selection rule has been changed from principle of positive assertion to toggle-member. [cea] 15:51:25 11 Mar 2002 - $MFX4/MFX/DEFS/mdrfiles.h Added _WIN32 conditional compilation. The _OSK sections will eventually be removed leaving a platform independent header. [rls] 15:47:19 8 Mar 2002 - TBASE Fixed bugs in resync attempt code added below [rls] 13:27:19 7 Mar 2002 - Startup Moved JTAG scan so it runs before SCSI caching. This may help with the "JTAG hangs during bootup" bug. [djs] LAME-SLAVE-DELAY This new variable in tcs_cfg allows the specification of a delay between bytes transmitted on the 9-pin interface. Some slow machines cannot handle the continuous data generated by the masculine 68k PY (eg. Windows NT based slaves). A delay of 3-5ms seems to make the v-motion happy, and we speculate it will do the same to an Avid. [rls] 15:47:19 1 Mar 2002 - TBASE - resync attempt If the PLL is still not locked after 4 seconds, internal sync is selected for up to 4 seconds then the original source is reinstated. [mcc] 12:07:19 1 Mar 2002 - jog 9pin slave Jogging is allowed in remote 9pin slave mode. [rls] 2002/02/25 Assume QDC has 15ns DPRAMS if no JTAG ID part fitted Bugs Fixed Since 17.1.04 ------------------------ [ajc] 19:30:44 17 Mar 2002 - 'snatches of audio' bug Modified audioman to mute inputs during all patching operations - removed some apparently invalid optimisations! [mcc] 12:00:00 15 Mar 2002 - recdata level parameter A bug is fixed that caused the level to be wrong if the level parameter for recdata is omitted. This parameter is optional and now correctly defaults to full scale level. [mcc] 11:54:22 27 Feb 2002 - recdata bug fix Recdata was inadvertently broken by the activation of record buffers independently of arming. This is fixed. The independent activation is now exploited by the recording software. This means that dropping-in by arming will not be delayed in certain cases. For example: Patch input 1 to all tracks and put transport into record. Then drop-in track 1 by arming it. Now arm track 2, then 3, then 4, etc. The resulting clips will be seamless, with no gaps, and they will have handles. Release 17.1.04a, February 21st 2002 ====================================================================== Changes Since 17.1.03 --------------------- [ajc] 12:07:40 21 Feb 2002 - QDC clock frequency (*EXPERIMENTAL*) Increased default DSP clock frequency by 5%. If this appears to be causing any problem please advise me! [ajc] 12:05:57 21 Feb 2002 - jtag modifications jtag command modified to work with SVF files generated by Xilinx 4.1 tools. Added cplderase procedure to erase all QDC cplds; added erase procedures to individual cplds in qdc and pxy. [rls] 11:36:52 21 Feb 2002 - Modified qdcinit to detect AO1M cards directly. AO1M card is detected by reading back a relay control pin. The JTAG id chip is no longer needed. [ajc] 18:02:52 19 Feb 2002 - build release 17.1.03d [ajc] 17:52:21 19 Feb 2002 - MWOS Tree INET Directory In an attempt to rationalise the MWOS include directories for the OS9 ISP (networking) environment I have created a "shadow" INET directory under $MFX4/DEFS. This will be searched ahead of the MWOS directories and contains copies of all files in the $MWOS/OS9/SRC/DEFS/INET directory which is actually obsolete. However the newer defs files in $MWOS/SRC/DEFS/INET are incompatible with some of our existing sources so we still need them. This roundabout method has been used to avoid the need to change the MWOS tree which would cause older rev 17 trees to not compile. The changes may be integrated into the MWOS tree at a later stage. Since all networking code is affected by the change in defs files this release (17.1.03d) must be thoroughly tested in all aspects of network operation!!! [awk] 11:30:00 19 Feb 2002 - Medialink Build Environment Moved the ONCRPC package and the executable qawk.exe to be located within the m4 tree rather than being located externally which is the current situation. This allows the tree to be more self-contained and hence it should now not need these external packages to be installed first for a build to complete. [ajc] 15:56:42 5 Feb 2002 - built 17.1.03b [ajc] 15:53:24 5 Feb 2002 - audioman Support buss group formats 6.1, 7.1 & LCR. Fixed gain offset when compressor enabled. [rls] February 4 15:57:00 2002 - Improved mfxping utility Added options -l=length -d=delay -w=wait -e=data (erase data) -v -vv (verbose) Fixed bug causing "Bad Response" to be printed instead of "No answer" (the sigmask is being set somewhere in the OS - added sigmask(0) to clear it) [djs] The ML4 reader now truncates long string names, instead of stopping with an error. [steve] January 29 11:53:53 2002 - POWERD The system will *not* force a power off 5 seconds after the power switch has been detected as pressed. If the system does not shut down now it is due to a software/hardware problem. With the 5 seconds timeout in place, any problems were being silently hidden. There is also some concern if write operations to FLFS disks while the power is removed may have on disk integrity. Bugs Fixed Since 17.1.03 ------------------------ [ajc] 18:02:52 19 Feb 2002 - build release 17.1.03d [ajc] 17:50:54 19 Feb 2002 - Mixer & Console comms over TCP/IP The problem with TCP/IP comms in Dream systems has been fixed by setting the socket flag TCP_NODELAY which disables coalescing of small packets. [mcc] 12:16:35 12 Feb 2002 - SuperTrim bugs The cancelling of expanded clips now works correctly and does not undo real edits by mistake. The EDL compiler was not always invoked correctly and so the display & play systems could get out of sync with the actual EDL. This has also been fixed. It was discovered that the user is not prevented from dropping-in to record while clips are expanded, resulting in havoc. The system now cancels the expansion at drop-in, if possible, otherwise the expansion is retained (this will always occur in Overwrite or Tape recording modes). A bug was fixed that made the selection of tracks very slow while a trim or slip is active. [mcc] 12:16:35 12 Feb 2002 - No undo while editing EQ Undo is disabled while editing clip EQ - it always should have been but was not. Also, EQ window did not use correct lock-out protocol and was thus susceptible to a fatal race with the main editing task. [mcc] 12:16:35 12 Feb 2002 - Mishandling of Track Sel key Solo, Safe, Mute, Disable and EQ menus mishandled Track Sel key in Dream family consoles. The Track Sel key now correctly has no effect in those menus (it is a minor bug that the LED does not indicate this, and remains dim). [mcc] 12:16:35 12 Feb 2002 - Mandatory range can't be turned off In Dream family consoles the Range ON key would incorrectly allow the range to be turned off in menus which do not allow this, such as the Track menu. [ajc] 12:35:18 12 Feb 2002 - 32k sample rate Fixed incorrect clock timeout in DSP s/w causing failure at 32000 sample rate [mcc] 14:02:44 7 Feb 2002 - jog panel params at 4x normal rate A bug was introduced in Dream family consoles with the Blue-jog fix: the jogging of editor panel parameters was permanently in "high gear", ie, 4x normal rate. This has been fixed. [mcc] 20:17:31 6 Feb 2002 - AB2 bugs 1. re "tinyint" - publish would fail if any clip had PAN set left of centre because a negative value is not allowed by SQL for pan. The same problem also applied to exportab1 and abtext commands. 2. Several bugs in fetching of clip data for audition & paste caused the resulting clip to be corrupted: EQ, fade alpha/sigma, level and panning were all garbage. [ajc] 14:05:53 1 Feb 2002 - qdcinit IO Xilinx load (XDONE timeout) fix; improved error reporting [djs] Fixed a bug that caused ANETD to crash when a read request had an invalid file handle. The bug was caused by an attempt to dereference a NULL pointer (buf). The bug was found by opening a project on another (server) QDC, and rebooting that QDC with the project still open. The server's ANETD process would crash on restart after displaying a message like: Wrong INODE $00000380 FD=$00000000 [mcc] 13:06:39 29 Jan 2002 - power-down vulnerability The quit command did not wait for ack from mdrmain. Instead it merely waited for two seconds. Thus, when quit is forked by power-down (smart switch or command), there was nothing to prevent the power from being switched off during the final flushing and closing of the project file, or during the deleting of the mdr_undo file. This put the project file and the file system of the project disk and boot disk at risk of corruption. The quit command now waits for an acknowlege and then waits a further two seconds. The mdr_undo file is no longer deleted when mdrmain exits. It may now take longer to power-down, but previously it was not taking long enough. [mcc] 10:12:44 28 Jan 2002 - minor LED anomalies Fixed LED behaviour for some Dream-family softkeys (LED didn't go dark when inactive): Digi menu - Inp Sync, Stereo Setup menu - NTSC [djs] ML4 reader fixes: The ML4 parser had a typo causing FR29DF to be parsed incorrectly. ML files containing 29 drop-frame would not read. This has been corrected. In addition the parser now outputs correct line number with error messages when the ML4 file has DOS-style line end characters. In addition the parser can now properly recover from errors and parse other files. Release 17.1.03a, January 18th 2002 ====================================================================== Changes Since 17.1.02 --------------------- [steve] SJanuary 18 13:37:18 - Sysinfo Modified to display out of date PX/PY static and Kernel ROM images even if not a "golden" machine. [djs] A new modifier "delay" has been added to the sony_id file. This is used to declare the number of frames ahead of a drop point the drop command should be sent down the 9-pin wire. For most devices the default of 3 is correct. It can be altered for machines that are slower or faster at dropping into record. It can also be used to offset drop points on a remote machine. [mcc] 18:56:11 11 Jan 2002 - 17.1.02b: espmsg changes Espmsg rev 11: os9 events are no longer used. Max message length is now defined by each receiver. [djs] Wednesday December 19 03:40:08 GMT 2001 ML4 files are now written to files with an .ML4 extention. The import process is the same for .ML and .ML4, in that is auto-detects the file type and uses the appropriate import method. [rls] December 19 14:00:00 EST 2001 - How to program YGLUE with Rev 19 I have checked in files for programming YGLUE with rev 19, if you need to revert for testing purposes. The procedure is to manually edit - /dd/usr/jtag/cdl/pxy.cdl at line 122, (12 lines after - "device pxy-yglue") change from: else run builtin xc9500_program("cpuglue20.svf"); to: else run builtin xc9500_program("cpuglue19.svf"); Unless you have a rev 4 pxy, in which case you want the previous line. Now run "jtag -v yglueprog" [ajc] 22:59:46 18 Dec 2001 - jtag program update Allow jtag to handle assemblies with no ID component (qs3j245/309) fitted. ID component may now be specified 'optional' in cdl file. [rls] December 18 17:00:00 EST 2001 - New PXY YGLUE CPLD Rev 20 A CPLD bug has been discovered which causes PX to hang if PY is reading PX DRAM, and PX does a MOVE16 from PY DRAM at the same time. Fortunately PX doesn't yet do any MOVE16 instructions. This bug is fixed in CPUGLUE rev 20 which only needs to be put into YGLUE. XGLUE will stay at rev 19. Sysinfo will detect the YGLUE revision and instruct the user to run "jtag -v yglueprog" followed by a "restart -y". It is important to be patient while the programming runs. The erase... and program... stages may take several minutes to complete. It is also important to do the restart immediately afterward, as the system may hang if other commands are issued. [djs] December 18 ??:??:?? EST 2001 - ML4 ML4 import and export is now ready for testing. ML4 files are exported using the "filex -n" command. ML4 files will be autodetected upon import. ML3 export and import is still present and default. (This change was actually added to release 17.1.02) Bugs Fixed Since 17.1.02 ------------------------ [djs] Sony 1-in-30 drop in bug. This bug caused drop-in to not occur roughly 1 in 30-40 times. The 1 in 30 remains, but now the effect is to drop in/out 1 frame late instead of not at all. [steve] December 18 15:17:35 EST 2001 - UPGRADE A problem has been found with upgrade that caused a system upgraded from 16.X to 17.X to fail on the next reboot. This has been fixed but will require 16.5.24 and 16.6.16 to be installed first. At the time of writing, neither of these releases have been bult yet. A work around is to "load copy" before starting to upgrade or .... if in the situation where a message saying "**** csl traphandler mismatch ****" when attempting to boot is displayed, you must boot to the ROM and type "bootgen /sdXX /sdXX/dd.bf" where XX is the port and SCSI id of the disk you are booting from. After this has been done, all should be OK. This problem will also happen when downgrading from 17.1.02a (only) to 16.X. All later versions of 17.X will be OK. [djs] December 18 ??:??:?? EST 2001 - ML4 Fixed minor bug in ML4 import, which would have resulted in an inaccurate wave size calculation for waves that did not start on a byte boundary within the file. Release 17.1.02a, December 18th 2001 ====================================================================== Changes Since 16.6.14/17.1.01 ----------------------------- [mcc] 14:15:22 16 Dec 2001 - abtext V1.02 Clip and file records now include record numbers, starting at 1. The -c and -f options can be used to change the starting numbers for clips and files respectively. The -n option causes record numbers to be omitted. The -e option allows different end-of-line styles to be selected: PC, os9, Mac or Unix. PC is the default. See "abtext -?". [mcc] 16:31:05 14 Dec 2001 - abtext timeunit fields Abtext V1.01 outputs all timeunit fields as signed decimal because that is what SQL needs. [ajc] 00:08:48 11 Dec 2001 - xcc upgrade Now that the MWOS-3.2 tree is installed xcc has been reconfigured to remove the "hacks" previously required for C++ programs using exceptions. The new compiler is now always used; the no-exception library is linked unless -qx is specified to the linker. This change only affects C++ programs. All C++ OS9 modules should be remade from scratch (make clobber all) to verify that there are no compile- or run-time problems. Bugs Fixed Since 16.6.14/17.1.01 -------------------------------- [chris/steve] - December 17 13:39:11 EST 2001 - IOGLUE/SCSI Hang The "SCSI hang" problem has been fixed by version 46 of Intglue. The fix is in the arbitration handling of IOGLUE. This fixes a deadlock/retry problem when PY was reading from PX's DRAM across the PCI bus. It resulted in a double read from a "read once" register on the SCSI card. [mcc] 08:18:41 14 Dec 2001 - FLFS FM: spurious fixup words in MFT Fixed bug that left spurious "fixup" words in the MFT of every file created by the file mgr (as opposed to files created by diskinit). There seems to have been no serious side effect, it was just wrong. [mcc] 08:18:41 14 Dec 2001 - FLFS FM: system file attribute handling The system files created by diskinit (including the root directory) do not have the pseudo osk file descriptor attribute. Attempts to write to this attribute for such files (eg, by attr, touch or tcopy) were not supported properly. For example, touching the root wrote garbage to every field but the date. It was suspected but not proven that tcopy could damage the root directory by this route. Therefore it was decided that writing to the osk file descriptor should be disallowed if it does not already exist for the file. Consequently commands like attr & touch will now return an error for these files. Tcopy no longer does this, and so it will not get an error. [spr] December 12 11:03:34 EST 2001 - TCOPY The command tcopy now copies files to and from FLFS disks as expected. Release 16.6.14, December 11th 2001 ====================================================================== Changes Since 16.6.13 --------------------- [mcc] 07:49:11 10 Dec 2001 - AudioBase new abtext command The new "abtext" command reads an ABF file (ie, abase.master, abase.srch or abase2.srch) and writes two output text files, one containing the clip data and the other the file data. These files are tab-delimited text files with PC eol format. All three files must be specified, see "abtext -?". [mcc] 07:49:11 10 Dec 2001 - AudioBase form layout Filetype, User3 and User4 fields have been added to the form layout. All four user fields have been expanded to the max length (24). Note that the form layout is defined by the editable text files /dd/USR/ABASE/abase.form (AB1) & abase2.form (AB2). The new fields and widths can be defined in any hires release. The main reason for the editable layout was because, in lores, not all fields could be displayed at once. In hires gfx there is plenty of room for all fields. [mcc] 07:49:11 10 Dec 2001 - AudioBase 96K 96K sample rate is now supported by AudioBase 1 & 2. [mcc] 07:49:11 10 Dec 2001 - AudioBase2 check URL length It is suspected that URL buffer overflow might be responsible for some mysterious fault reports. URL overflow is now detected but will cause a fatal assert (the stack is corrupted so there is no point continuing). To test, edit all text fields in some records to be maximum length and confirm that those records can be returned by a search. Also, issue searches where all fields are active and text fields are maximum length. Watch for the "A Task has Exited Fatally" popup, accompanied by tvt message "*** URL OVERFLOW". Please inform me immediately if this occurs. The code can easily be changed so that URL overflow is virtually impossible but first we want to know if it is happening, as this would certainly explain some of the odd faults that have been reported in AB2. Bugs Fixed Since 16.6.13 ------------------------ [mcc] 12:41:44 9 Dec 2001 - Publish WAV: Illegal Filetype error Fixed an old bug which caused error "Illegal Filetype for Operation" when publishing a single WAV file or folder of WAVs if the file or folder was not in the root directory of the disk. [ajc] 16:46:28 7 Dec 2001 - FAME Stereo Linkage Problem (from 16.5) Fixed a bug in audioman which meant that feeds were never unlinked correctly after having been linked. [djs] ML v3 importing was inadvertantly broken in the last release. There was an error in the handling of clip start/stop and hook times. This has been corrected. Release 16.6.13a, December 5th 2001 ====================================================================== Changes Since 16.6.12 --------------------- [mcc] 11:31:00 3 Dec 2001 - new two-handed movement rule In Dream family consoles, the rule for invocation of a "two-handed edit" has been changed. If the track selection has not changed then the transport must have been moved explicitly by a locate operation, or by turning the jogger in jog or shuttle. [steve] November 26 13:08:05 EST 2001 - REMOTED Modified remotedc so it issues a "restart -y" command when a restart command is received instead of calling the debugger. [steve] November 24 17:32:53 EST 2001 - BREAK The break command no longer resets the machine if the debugger is off. It will now simply return the to command shell that issued the break. If you need to restart or reboot the system, you must now use the "restart" command. Bugs Fixed Since 16.6.12 ------------------------ [steve] November 22 14:55:21 EST 2001 - SCSICACHE There was a problem with "scct" which is part of the SCSICACHE system. On startup, the function _os_loadp() was causing some data corruption. This function call has been replaced with _os_load() and seems not to cause the corruption anymore. [mcc] 13:25:48 1 Dec 2001 - long filenames not cleared Long filenames were not fully cleared by the enter_name() function that is shared by copy, extend, rename, etc. This was an ancient bug that affected all consoles. [mcc] 13:25:48 1 Dec 2001 - bungled panel mgt by Pro Tools The Export/ProTools softkey procedure had several bugs in the mgt of the panel resources (LCD and LEDS). Faulty browser intialisation caused tvt message "A3 BLOWN". The timecode display was not inhibited and was thus corrupted. The procedure did not cancel correctly when another mode key was pressed or if ProTools key was pressed again. The LED did not flash. [mcc] 12:36:22 30 Nov 2001 - RangeON key LED maintenance RangeON key LED was not being maintained properly. [mcc] 12:36:22 30 Nov 2001 - No project new24 in Station & Dream Project menu new24 softkey is inhibited in Station & Dream because the mixer cannot yet be reconfigured to 24 tracks. [mcc] 14:51:11 28 Nov 2001 - possible fix for SCSI hangups The PY static TVT was identified as the probable cause of SCSI hangups. It has been globally disabled as a result. [steve] November 26 13:01:03 EST 2001 - SCCT (SCSICACHE) Fixed major bug in scct re configuring STDIO for a system state task. Thsi bug exists in all versions of software to date including revision 15. [mcc] 13:44:19 23 Nov 2001 - Satellite bug fixes Blue-Jog transport gives 4x velocity increase. Play/Jog LED status bug fixed. Zoom-Jog no longer toggles zoom setting. Level menu track softkey is only enabled in stereo output mode. [mcc] 13:44:19 23 Nov 2001 - jump-to-clipname track selection bug Obscure bug fixed that could leave console track selection out of sync with QDC in certain conditions (when no jump occurred). Affected all consoles. [steve] November 22 14:55:21 EST 2001 - FAT16 FAT16 disks can be mounted again. This turned out to be a bug in the IRBF driver which Daniel Simon has been informed about. The work around was to read sector zero twice as the first attempt generated an "Invalid Sector Size" error. [ajc] 00:47:59 20 Nov 2001 - I/O Loopthrough after crash Fixed bug which resulted in inputs looped through to outputs after a system crash or system restart without quitting first. Release 16.6.12a, November 19th 2001 ====================================================================== Changes Since 16.6.11 --------------------- [steve] November 19 14:55:00 EST 2001 - diskinit v2.0 Diskinit has been increased to version 2.0 for this build. No changes to the code since mcc's 1.42 submission. [mcc] 10:37:12 16 Nov 2001 - diskinit V1.42 for FLFS Changes to initialisation of FLFS disks. Bug fixed in bad block scan. System clusters are always scanned for bad blocks and if any are found the disk cannot be initialised. If any bad clusters are found then the user must confirm whether to continue or not. The initial allocation of space is contiguous - there are no worrisome holes in the bitmap. The volume size, as given in the ID sector, is truncated so that there are no non-existent clusters in the final byte of the bitmap. Versions of flfscheck earlier than V3.03 will report this as "ERROR: Sector Count Mismatch" and fail. [tino/-steve] Tuesday November 13 19:26:51 EST 200 - FMC A new version of FMC has been submitted by Tino. This is version 1.0.26.0. This maps the folder +/- keys correctly and the Macro layout has changed as per the latest spec. [mcc] 08:32:21 13 Nov 2001 - macro key layout, FMC version test The Satellite macro key codes have been revised as per spec. The version of FMC is tested by 68K - if out of date then it prints "Console FMC Software Not Correct Version". The version must be V1.0.26 or higher. [spr] Monday November 12 14:31:21 EST 2001 - SYSINFO sysinfo has been changed to detect the version of the FLFS file system on the boot device. If this version is less than 1.41 the user is informed that the boot drive needs re-initialising. The QDC software will NOT run in this state. The disk must be re-initialised before any QDC software will run. [djs] NODED changes. When attempting to detect a remote machine's RPC capability, noded now only waits 0.25s for the tcp connection. It also makes three attempts. This should stop long timeouts that sometimes occur, the result being a more up-to-date list of nodes in the file browser. [djs] Added some debugging output to NODED on tvtline screen #9. This is useful for checking if the application is still running. [spr] Thursday November 8 17:32:16 EST 2001 - Unified FLFS check A unified FLFS check that can check the disks from any version diskinit as been submitted. Before this, diskinit and flfscheck had to be "paired". This is no longer the case. Sector zero from FLFS disks are more stringently tested before lauching a full check. (i.e. RBF disks dont say that the FLFS file structure is damaged!). [rls] 12:26:31 8 Nov 2001 - qdcinit Modified AES2 card setup to eliminate noise from unconnected AES input. [mcc] 19:24:17 7 Nov 2001 - 'enduser' alias Defined 'enduser' alias for 'password -h'. [mcc] 10:03:22 7 Nov 2001 - Import/WAV files search for 96K The Import/WAV files search form now allows 96K to be selected. [ajc] 00:26:31 5 Nov 2001 - qdcinit Modified AES(1) card setup to try to eliminate noise reported by FltJP from unconnected AES input. [mcc] 13:42:15 2 Nov 2001 - Track Sel terminated by binnacle target release Releasing the binnacle target key terminates the Track Sel procedure. In other words, you can release the target key first, before Track Sel, when performing two-handed editing. [djs] Projects can now be exported in ML4 format. Please note that this is still under development. The ML4 export is performed with the "filex -n" (new) command. [ajc] 21:27:32 29 Oct 2001 - DSP Monitoring mod Delay onset of monitoring a little (currently 64 samples) to ensure that monitor & playback signals do not overlap. This is an experimental mod which should reduce the severity of the punch-in "pops" reported by Enya studio. [ajc] 21:29:40 29 Oct 2001 - Mediad Handle non FAT32 partitions on DOS-partitioned disks. Also do not create a user DD for the "whole device" since this is meaningless on a DOS-partitioned disk [ajc] 21:30:59 29 Oct 2001 - Tvtline - display line numbers for reference Bugs Fixed Since 16.6.11 ------------------------ [steve] November 19 14:55:00 EST 2001 - Networking All networking modules have been modified to use new DNS lookup code from the Microware library netdb_resolv.l. This fixes the problem with AB2 not being able to use an IP address in the /nvr/setup file. [steve] November 19 14:55:00 EST 2001 - hosts file The hosts file has been modified not to include internal fairlight machine addresses in the release file. [steve] Thursday November 15 13:25:32 EST 2001 - MEDIAD/DFN/FILEGOPHER Fixed an unwanted side effect of mounting disks with unknown files systems as RBF. File gopher (via DFN) would be informed of such a device and would attempt to do a file search. The first access would generate an "Invalid Access Mode". This has been fixed by not informing DFN (and therefore filegopher) of disk where the format is not known. The device is still built however so it can used as a name for diskinit. This fix is in the mediad module. [djs] Fixed a condition in noded's ping test that would occasionally hang the daemon. The application was setting an alarm with a very short time, as a way of aborting a network operation. Under high load the alarm would trigger too early, and the following network call would wait forever, effectively stopping all node detection. [mcc] 10:55:02 7 Nov 2001 - 'prev gap' on From menu The From menu shows 'prev gap' instead of 'next gap' on all consoles. [ajc] QDC#1025 17:33:52 6 Nov 2001 - QDC crashing at 96k Reduced maximum DSP playback pitch at 96k to prevent QDC hanging. [mcc] 02:41:55 31 Oct 2001 - Satellite Rec Menu The Rec Menu (head,clip,here,again) now works in Satellite. Release 16.6.11a, October 29th 2001 ====================================================================== Changes Since 16.6.10 --------------------- [mcc] 10:58:44 29 Oct 2001 - Serato V1.0834 Version 1.0834 from Serato. Consumes less memory. Needs testing to verify that it still works correctly. [mcc] 10:46:27 29 Oct 2001 - FFS diagnostic changes The -c option for FFS disables the DFN connection cache. On machines with R&D dongle (feature_develop), FFS measures the elapsed time of a few critical operations and reports if more than one second, or if FFS debug enabled (MDR F option). These changes were made to help diagnose the delay that sometimes occurs when browsing over nodes. [mcc] 15:27:10 22 Oct 2001 - embed -w option The new -w option for the "embed" test program causes the file to be written, ie, the object is extracted from the project. Bugs Fixed Since 16.6.10 ------------------------ [mcc] 10:33:20 27 Oct 2001 - recording problems due to garbage handle The value stored in console NVRAM for record "handle" was not range-checked before transmission to the host at startup. Any value up to the inherent limit of 255 frames could be asserted. If the handle was too large then recording would fail in various ways, including spurious reports of "DISK NOT FAST ENOUGH - RECORDING MAY CONTAIN ERRORS". The value in NVRAM is now rigorously tested. It is set to zero if invalid, not the maximum value of 30 frames. [mcc] 11:18:21 24 Oct 2001 - ML EQ import/export faults Two bugs have been fixed in the handling of EQ in ML files. For export, most EQ was not included in the ML file. For import, all shelf types became low-shelf. Release 16.6.10a, October 22nd 2001 ====================================================================== Changes Since 16.6.09 --------------------- [mcc] 16:34:19 18 Oct 2001 - look first in project dir for attachments When trying to locate and reopen attached external files, MDR will look first in the same directory as the project file. The FFS search engine is only deployed if a file is not found in the project directory. This should speed up the opening of a project when the attachemed files do in fact live in the project directory, as will be the case after a chain of autoextends. [mcc] 09:53:25 18 Oct 2001 - FFS search engine Some changes have been made to FFS and the client interface in preparation for complete pathlists to attached files in the project, and to enhance the effienciency. 1. Pathlists that have no node (eg, /sc20/foo/...) will only be searched for on the local disks. Previously the entire network was searched. Actually such a pathlist should cause an error, but I think some rogue applications depend on this to work. However they do not expect the file to be on another node. 2. Another redundant MDR-FFS-DFN transaction eliminated when fetching file cache info for complete pathlist. Previously the ffslib interface used the pathlist to fetch the FILEINFO, retrieved the FID from the FILEINFO, and then used the FID to fetch the FILEINFO again! Incredible. 3. When MDR is attempting to locate a file by complete pathlist, it no longer uses the search engine at all. [steve] Monday October 15 16:17:10 - Machine file Location. The location of the machine file has changed with the public ftp directory structure. It is now located in /pub/DONGLES/features. [steve] Monday October 15 16:17:10 - Networking changes. The /nvr/setup file has some new members. These are ... DNSNAME = a.domain.of.your.choice.com This enables you to set up the DNS domain so DNS lookups via DNSSERVER will work as expected. Note DNSNAME has nothing to do with DOMAIN. DNSERVER = 1.2.3.4 This sets the IP address of a DNS server on your network. Set it up to resolve names correctly. GATEWAY - 1.2.3.4 This is the default route for all IP traffic. Any network that can not be reached will be sent to this host. Now it is possible to access the Internet directley from an QDC/MFX if the above are set up correctly. Here follows a transcript from a QDC/MFX session. DNSNAME = fairlightesp.com.au DNSSERVER = 192.100.105.1 GATEWAY = 192.100.105.1 steve:chd /dd steve:ftp ftp.fairlightesp.com.au Connected to ftp.fairlightesp.com.au. 220-Welcome to Fairlight FTP 220- 220-If you have any problems with this FTP server 220-please email: webmaster@cinfo.com.au 220 Service ready for new user Name (ftp.fairlightesp.com.au:steve): machine Password (ftp.fairlightesp.com.au:machine): 331 Password required 230 User logged in Connected to ftp.fairlightesp.com.au. Mode: stream Type: ascii Form: non-print Structure: file Verbose: on Bell: off Prompting: on Globbing: on Hash mark printing: off Use of PORT commands: on Overwrite: off Directory Recursion: off ftp> dir 200 PORT command successful 150 Opening connection -r--r--r-- 0 29388 29388 Oct 13 09:13 features -r--r--r-- 0 78 78 Oct 13 09:13 features.crc -r--r--r-- 0 116880 116880 Oct 12 12:18 machine -r--r--r-- 0 78 78 Oct 12 12:18 machine.crc 226 Transfer complete 262 bytes received in 0.07 seconds (3.66 Kbytes/s) ftp> bin 200 Type set to I (Image) ftp> get machine 200 PORT command successful 150 Opening connection (116880 bytes) setting machine to 116880 bytes 226 Transfer complete 116880 bytes received in 10.60 seconds (10.77 Kbytes/s) ftp> quit 221-Bye from Fairlight FTP [mcc] 10:27:19 11 Oct 2001 - embedded OBJECT array in omdlex data module The data module created by mdrmain for omdlexport now includes the array of embedded OBJECT descriptors. [mcc] 11:32:04 10 Oct 2001 - redundant MDR-FFS-DFN transactions eliminated A number of totally redundant transactions used to occur between MDR and FFS-DFN when a project file was opened, created or extended. One was to fetch all the information from the file cache for the newly opened file (sample rate, file type, backup date, etc) and then just throw it all away. Another goody was to use the node and unit of the path as parameters to fetch the node and unit of the path. These transactions were introduced at the start of networking in rev15, and have now been eliminated. This gives a noticeably faster project-open time, provided there are no attachments to search for. Alas, these changes do nothing to improve the speed of searching for attached files. Bugs Fixed Since 16.6.09 ------------------------ [mcc] Monday October 22 10:00:06 - LTC Fixed spurious initialisation of tcs_tbase_getmsn vector that caused read_tbmsn() macro to return VSN instead of MSN. [mcc] 07:54:25 19 Oct 2001 - input monitoring after Wave menu op The input mode and monitoring status were not being correctly restored after a Wave menu (or any ODIF) operation. In particular, tracks in THRU mode were not reinstated correctly. [mcc] 08:14:55 18 Oct 2001 - incorrect backup date displayed The file window displayed the wrong backup date, the last edit date was displayed instead. Introduced with mods for hi-res graphics. [mcc] 11:44:42 17 Oct 2001 - THRU tracks become ARMED In Arm menu on Dream-family consoles, THRU mode tracks magically became ARMED when you left the Arm menu. Release 16.6.09a, October 10th 2001 ====================================================================== [rls] 16:02:24 08 Oct 2001 - jtag Added erase procedure for QDC_RESET CPLD needed by production test. Bugs Fixed Since 16.6.08 ------------------------ [ajc] 23:34:47 9 Oct 2001 - Various bug fixes - Handling of mixer 'OscMode' packet fixed - oscillator can now be turned off! - Fixed latent bug in real-time metering which was activated by previous bug fix for ramping - and which resulted in only the first meter on each DSP working. - Corrected code which updates ramp levels after ramp TC has been changed. It is no longer necessary to blue-stop after changing ramp time constant. - Additional checks on integrity of TSB commands received by DSP. Release 16.6.08a, October 5th 2001 ====================================================================== Changes Since 16.6.07 --------------------- [mcc] 11:25:44 5 Oct 2001 - Satellite macro key layout The macro key definitions are adjusted to match the layout of the new key mat (1 2 3 on bottom row). [djs] Large amounts of networking code have been converted to ANSI. There should be no functional difference in this conversion, however it would be worth keeping an eye out for any behavioural differences in the system. [mcc] 22:50:19 25 Sep 2001 - 16.6.07c: new LCD protocol The new LCD protocol is enabled for Dream family consoles. If the FMC software package is not up to date then the LCD will not work properly. See Tino or Greg Adams if this is so. You can revert to the old protocol by setting NEWPROTOCOL to zero in $MFX4/MFX/CONSOLES/EMU/lcds.a. [mcc] 14:58:22 20 Sep 2001 - 96K buffering If there is enough WRAM installed the size of playback and record buffers is doubled at 96K for 48tk configurations only. [spr] Thursday September 20 10:41:43 EST 2001 - DVD-RAM support As CD-Writing is now longer supported, all CD Drives now appear as normal hard drives. This means that DVD-RAM drives can be read and written to as if they were a hard disk. This has been seen to work by formatting a DVD-RAM on Windows 2000 and moving the drive to the QDC. The file system on the DVD-RAM has to one that the QDC supports. These include FAT32, FLFS and RBF. Other file systems may work but have not been tested in R&D. The performance of DVD-RAM drives is very slow (4 tracks @16bit 48kHz) for recording and about double for playback. [mcc] 17:16:21 16 Sep 2001 - v16.6.07b: Editor super-sleep mode The Dream-family console editor software now provides a super-sleep mode (idle super mode) that the mixer should invoke to enter a mixer super mode. The mode is activated by a key depression of pseudo key #149 and deactivated by a key release of that key. [mcc] 17:16:21 16 Sep 2001 - New LCD porthole protocol The new LCD porthole protocol has been implemented but is disabled conditional upon NEWPROTOCOL in EMU/lcds.a. The new protocol is not compatible with current version of FMC (whatever that may be). Bugs Fixed Since 16.6.07 ------------------------ [mcc] 10:58:01 5 Oct 2001 - failure to prime, zero-sample deqel A very old bug has been fixed that could cause the MDR playback system to fail to correctly prime and start a channel after a transport locate operation (jump, goto, even ffw-play). The worst symptom of this was that the transport could begin to play before all channels were primed and started. The underlying cause of the fault was the insertion of a deque element with a duration of zero samples. This occurred when the seek time (transport position) was one timeunit away from a clip boundary AND when the seek and boundary times both convert to the same sample. Of course that is only possible at 44K1 or less. To fix this, the seek time is now adjusted if it is too close to the nearest clip boundary. This bug was introduced in 12.0 in 1993! [mcc] 20:11:51 27 Sep 2001 - softkeys that did not cancel Several softkeys did not cancel when pressed again. These were: Project menu new & new24 softkeys; Name menu project, track, clip & seed softkeys; Mark menu name & time softkeys. Additionally, in Mark menu, when name or time are completed or cancelled, the number softkey is now reselected. [mcc] 22:50:19 25 Sep 2001 - console message problems Fixed two ancient bugs in Project/delete: 1) If filename was too long (more than about 25 chars) then the "PRESS ENTER TO DELETE.." message buffer overflowed with unpredictable results. 2) If delete was cancelled then the browser position was not refreshed in the LCD and remained blank. Fixed ancient bug - the sysmesg printed by the confirmation query ("PRESS ENTER TO DELETE/COMMIT/DISPOSE...") was not null-terminated and could potentially ramble on into junk. [ajc] 00:10:24 22 Sep 2001 - qdcinit Fixed error in programming of bare boards [awk] 17:16:21 21 Sep 2001 Project menu - move problems Fixed a problem in the routine check_for_lock(), from file_winnt.c. This routine had not been modified with the unlimited file enhancements, consequently the check for locks was unreliable at best or could cause dfnserver to crash at worst. This bug only occurs in the NT code, the OS9 code is ok. [mcc] 15:07:29 20 Sep 2001 - Corrupted mark list detection Fixed a bug where a corrupted mark list in the project file was not being detected, the first result of which was the track display exiting with Error #102. The list is now checked more thoroughly and reset if any error is detected. It is not known how the test project that revealed this was corrupted in the first place. [awk] 17:16:21 19 Sep 2001 FFF#1011 Project menu - move problems Fixed the bugs as stated below for the QDC and the Medialink code. Found a problem in the dfn_move_1() routine where moving a file from a sub-directory to the root on the same disk would cause dfnserver to crash. The files are moved correctly before the crash would occur. Also found another bug where moving files down from a directory or the root of a disk to a sub-directory on the same disk would cause a problem. In this case the files would be moved correctly but the display would show otherwise. This was caused by dfnserver's file cache not correctly being synchronised to reflect the changes after the move had taken place, and hence the display was not correctly synchronised either. As the QDC code and the Medialink code share the same source files then these bugs would appear in both QDC machines and Medialink servers. Release 16.6.07a, September 14th 2001 ====================================================================== Changes Since 16.6.06 --------------------- [spr] 14th Sep 2001 - TCS is now compiled in ANSI C instead of C++ In an attempt to find out why LTC sync fails to work sometimes, the TCS system is now being compiled in ANSI C. This is hopefully only a tempory measure. This will make the code (nearly) the same as what is in 16.4 which seems not to have the problem. [RLS] 12th Sep 2001 - Changed TBASE to allow all sync sources to work at 96k. Can now use a 48K word clock or AES reference for a 96K sample rate and vice-versa as well. 96k run-up and run-down fixed. Also a 48k machine can be locked to a 44k1 digital input (only) and vice versa. This could be enabled or disabled from the UI. [mcc] 16:34:55 12 Sep 2001 - From/To dark when not available In all Dream family consoles, the From/To key LEDS are off when the keys are not available for use. Bugs Fixed Since 16.6.06 ------------------------ [mcc] 16:34:55 12 Sep 2001 - sleep key when no project open The Station/Dream 68K will now honour the "sleep" key when no project is open. Release 16.6.06a, September 11th 2001 ====================================================================== Changes Since 16.6.05 --------------------- [spr] 11 Sept 2001 - Beep on power up The new QDC power supply takes a long time to start. It is unpredictable when it was OK to release the power switch. Now the QDC will sound a beep when it is OK to release the power button. [ajc] 12:30:28 10 Sep 2001 - asm21k mods Disallow multiple instructions on one line unless delimiter ;; is used. Modified DSP sources accordingly. This change immediately caught some bugs (luckily all harmless!) where the delimiter ";" was used instead of ","! [mcc] 16:42:02 10 Sep 2001 - New LED rules The new LED rules have been implemented for Satellite and the editor section of Station & Dream. Active mode LEDS no longer flash. Certain LEDS now pulse when you are NOT in the mode and the target of the mode is active: Arm/Solo/Safe/Mute/Disab - pulse if any tracks are armed/soloed/muted/safe/disabled. B/Up - pulses while transfer in progress ADR - pulses while cycle in progress Bugs Fixed Since 16.6.05 ------------------------ [ajc] 12:26:25 10 Sep 2001 - Mixer Ramping Bug Fixed long standing bug in mixer ramping which resulted in zero ramp time. Set ramp time to 7ms in aman_cfg - optimisation still required. Added aman request to set ramp time from "aman" cmd. Ramp time can be set by issuing cmd "aman ramp nn" where nn is ramp time in ms. Must blue-stop afterwards to reload correct levels. [mcc] 16:42:02 10 Sep 2001 - Zoom in mixer mode You can now Zoom whilst in a mixer mode on Station and Dream. Release 16.6.05a, September 10th 2001 ====================================================================== Changes Since 16.6.04 --------------------- [mcc] 7th Sep 2001 - Fully functional 96kHz playback and editing A new sample rate of 96kHz has been added to the QDC engine. This can be selected on the DIGI memnu like any of the existing sample rates. 96kHz is a dongled feature. Please make sure your machine is enabled. [mcc] 12:56:22 5 Sep 2001 - console name in LCD The EMU 68K program prints the correct console name in the startup LCD, Satellite, Station, or Dream. Bugs Fixed Since 16.6.04 ------------------------ [spr] FFF#??? - Setp 6th 2001 - Fixed AO/AI dongle bug. A typo meant that the count of AO's was being ignored and was using the number of AI's instead. This bug in AMAN has been fixed. This had a major impact on machine they were dongled for zero inputs (i.e. a playback system only). In this case the number of outputs was also set to zero. Release 16.6.04a, September 4th 2001 ====================================================================== Changes Since 16.6.03 --------------------- [ajc] 11:31:23 3 Sep 2001 - Mixer surface over TCP/IP Mixint now supports TCT/IP connection to the mixer surface in client mode. This means that the surface is the server and MFX actively connects to it. The server IP addr/hostname and port number for this mode are now stored in /nvr/setup in section [Surface]. If the key "Server" is present then a client mode connection to this host is opened on port "Server_Port" (default 2721); else if key "Client_Port" is present mixint waits for a connection on that port (default 2801); else the HSSL is used as normal. [spr] August 29 14:48:39 EST 2001 - PY Activity LED We now use the 68040 "stop" instruction. The botton LED on PXY will now extingish when the processor is in the stopped state. The intensity of this LED when TCS is running should be proportional to the sample rate. 32kHz being dim and 96kHz being brighter. When no active tasks are running at all the LED should be off. This is the case after starting qserver. If the qsys option Q is set, the LED will flicker as the qprocs display is built for viewing by "tvtline qprocs". This mod makes the PY LED have the same meaning as the PX LED. Bugs Fixed Since 16.6.03 ------------------------ [spr] 15:01:13 3 Sep 2001 - conclt reconnection fixed. The Console over TCP/IP client (conclt) was not reconnecting if FMC was restarted. This has been fixed. [mcc] 13:58:24 3 Sep 2001 - lost "Max Clip Duration" error If recording stops due to max waveform size (E_MAXSPACE), the error message is now correctly reported. [mcc] 07:38:14 30 Aug 2001 - AutoExtend error recovery Fixed failure to recover from AutoExtend error (re FFF#1005). If AutoExtend failed, the system was left in an undefined state and would probably crash with an exception from one of the MDR tasks. This would occur if the new file was created okay, and the extension process had already begun, and then an error happened. Eg, not enough disk space. Instead, what will happen now is: transport stops all tracks disarmed no project file is open MDR (Disk) goes offline the new file that failed to open properly is deleted an error message is printed, eg: AutoExtend Failed - Media Full. AutoExtend can be forced to fail by setting environment variable EXTERR to 1. Release 16.6.03a, August 29th 2001 ====================================================================== Changes Since 16.6.02 --------------------- [mcc] 20:06:17 28 Aug 2001 - autoext command AutoExtend test mode has been removed. The new "autoext" command can be used to force AutoExtend at any time (provided you are recording) for testing purposes or otherwise. [spr] August 28 10:46:20 EST 2001 - Fake 96kHz The smoke and mirrors method of enabling 96kHz has been removed now that work is under way to support 96kHz for properly. -- 16.2.02b -- [spr] August 28 10:46:20 EST 2001 - Remove tvt debug When using TCP connections in mixint too much debug was being sent to the tvt system. This debug is now optioned. Bugs Fixed Since 16.6.02 ------------------------ [ajc] 00:20:54 29 Aug 2001 - monitor faders "freezing" Fixed bug in MDR program in processing of mixer update queue which resulted in all updates after the first 1024 being ignored(!) [spr] August 29 11:03:43 EST 2001 - Shutdown Sony Remote on PY When QDC/MFX shutsdown, the sony CMAN driver was leaving aa task running. This task now shuts down with the rest of the tasks on PY cleanly. Release 16.6.02a, August 27th 2001 ====================================================================== Changes Since 16.6.01 --------------------- [mcc] 10:11:17 25 Aug 2001 - Serato V1.0833 This new version of Serato fixes the bug that was introduced in V1.0832. [mcc] 12:10:15 24 Aug 2001 - MDR doesn't consume disk remnant MDR will not consume the last remnant of disk space. The disk is considered full when the free space is 1Mb or less. [ajc] 12:04:40 22 Aug 2001 - Fast QDC support Qdcinit modified to detect QDCs with -15ns DPRAM & run TSB faster if possible. qdc.cdl modified to display correctly on 'jtag -v' (requires jtag v3.03). Fixed bug in jtag cmd re parsing of ?: expressions (used by new qdc.cdl). [ajc] 12:49:35 22 Aug 2001 - Buss Mixer cfgs All station & dream cfgs (except dream48/48/48) now have buss mixing enabled. [rls] 20 Aug 2001 - Added support for AO1M option to AO1 card. The AO1M is a modified AO1+AOH card with relays to isolate the outputs and is mainly used to drive monitoring outputs. Audioman controls the relays via the aman_mute_all() function. This ensures that no clicks occur when changing sample rates etc. or if AC power is interrupted or switched off. [ajc] 23:38:25 17 Aug 2001 - AOUT-CTRL Xilinx rev 12 Added control bit to disable analog power-down when DEGn asserted. [ajc] 23:39:25 17 Aug 2001 - Buss Mixer update Tidied up audioman; fixed old bug in DSP code which caused incorrect setup of TSB outputs for certain combinations of DSP resources. [spr] Friday August 17 14:01:37 EST 2001 - Version 16.6.01d [ajc] 00:48:41 17 Aug 2001 - First release of buss-mixer system Station 48 cfg has buss (sub-)mixer enabled for initial testing [spr] 16th Aug 2001 - new command "mfxrc" - The Remote Control Client A port of the UNIX command "mfx" has been ported to OS9 in preperation for its use when transfering data to/from the Dream range of products. The name has been changed to "mfxrc" as we are unable to have a command called mfx due to a directory of the same name in /dd/USR/CMDS. Not much can be done now, but for testing it is possible to copy files to and from localhost. Once the remoted service is running on a PC this command will be much more usefull. It will also be possible to transfer "emu.raw" directly to the PC from the UNIX development system. Examples follow .... steve:chd /dd/USR/MFX steve:mfxrc localhost -d=/dd/dd.bf >bf1 MFX/QDC Remote Control Client v1.01 Size of File on path 3 is 224938 steve:cmp /dd/dd.bf bf1 Bytes compared: 00036eaa Bytes different: 00000000 steve:mfxrc localhost -u=/dd/USR/MFX/bf2 6720 WARNING(ntracks==CC_NCHAN && nlives==NLIVE) FAILED MulTk ... BAD TRACK 33 USING 1 TOO MANY LIVE FEEDS! TOO MANY TRACKS! [mcc] 17:18:22 25 Jul 2001 - mdrmain compiled as C++ The MDR main task is now compiled as C++. Bugs Fixed Since 16.6.01 ------------------------ [mcc] 17:50:52 26 Aug 2001 - TAPE mode autoextend leaves tiny clip Autoextend in TAPE mode could leave a tiny clip fragment at the transition time if there was a clip spanning that point before the recording. The duration of the fragment would be a few subframes or less, depending on the default-fade duration. The fragment consisted of a tiny bit of the clip that used to be there, before it was erased by the TAPE mode recording. [mcc] 18:09:51 24 Aug 2001 - input sync assert A bug is fixed that could cause a fatal assert when selecting input sync. The assert occurred when the input number in console NVRAM was garbage. [mcc] 12:10:15 24 Aug 2001 - drop-in with little free space All armed tracks will now drop-in successfully when there is less than a few minutes of recording space available. Previously, one or more tracks would fail to start recording. Now you should be able to drop-in with as little as 2 or 3 seconds of space remaining. This partially fixes FFF#1005. However, no change has been made to autoextend, so there will still be a point beyond which it will not be able to drop-in and then immediately autoextend. But this point is now only about 10 seconds before the file is full. In other words, the width of the "window" within which FFF#1005 may occur has been reduced to about 10 seconds. Re testing: These are substantial changes. It is not autoextend that needs testing but rather basic recording in NEW mode with autoextend OFF. It is the "running out of space" that needs testing, particularly when only a FEW tracks are recording. When you run out of space, this is what should happen: 1. The transport will drop-out and stop. 2. The clip tails will be reasonably aligned, within a second or two of each other. 3. The longest clip will end no more than a second or two before the transport position (cursor). 4. The free time shown in the meter window should be zero, but may still show no more than 1 or 2 seconds remaining. 5. An error message is printed: "Media Full" means the disk (not the file) doesn't have enough useful space left. "Project File is Maximum Size" means the file is full. "Max Clip Duration Exceeded" means the waveform size limit has been reached before either the file or disk were full. This should only occur when recording a single mono track. (The free time display doesn't "know" about this limit and will still show some time remaining). Please test that these things in fact happen as described. Test the filesize limit by starting a new file on a disk with plenty more then 4Gb free. Test the disk space limit by starting a new file on a disk with less than 2 or 3Gb free. Test with small numbers of tracks because ironically this is the worst case for the software. Test 1 mono tk, 1 stereo tk, 3 mono tks, 2 stereo tks. Also test "provocative" numbers of tracks like 47, 31, 23, 17. Things to watch out for: 1. Objectionable jerks & stalls in the scrolling of the track display during recording, perhaps accompanied by the tvt message "load_mdr_time() TIMEOUT". 2. The error "Waveform Segment List Full". [mcc] 13:46:11 23 Aug 2001 - ML export totaltracks The totaltracks= item in ML files has been corrected to be the highest referenced track number instead of a count of used tracks. [mcc] 18:13:33 7 Aug 2001 - ML import Fixed three bugs which manifested once the omdlr app had been invoked to reduce the number of tracks, etc. The symptoms included dotted stereo clips (that should have been mono) and a fatal assertion. [mcc] 08:19:55 3 Aug 2001 - mixer track selection The _seltracks message structure did not have a header. [rod/spr] New SIO1 Sync Xilinx revision 57 Corrected an error in the "Ready" logic. This was causing a contention with the LTC Xilinx Ready output which can result in a bus error when accessing the LTC generator. The symptoms were ... "Q kernel exception - bus error" at fault address 2c000818 (LTC control) Which then resulted in "IPI Timeout". This was occurring on 3 SIO1 cards in production, but it could have occurred to any machine in the field. Release 16.6.01a, July 25th 2001 ====================================================================== Changes Since 16.5.05 --------------------- [spr] 25th July - Path Extentions Disabled Path table extentions for FFS/DFNSERVERS and NODED have been disabled. By setting PATHEXTON in the startup script this can be enabled again. The select function supplied by Microware can only support upto 32 paths. This caused problems when more then 32 paths were open. [mcc] 10:22:11 25 Jul 2001 - Serato 1.0819 This version of Serato is more reliable than the last, and equally fast. [ajc] 23:08:35 23 Jul 2001 - various - Created defs file for pico-time-units (PTU) type. - Added mixer interface over TCP/IP (packet fmt unchanged); this is enabled on cmd line by specifying device "tcp:" or "tcp:" or by setting environment variable MIXER_COMMS_TCPIP[=]. - Enable QDC I/O during ODIF operation - ODIF programs can now print messages to TVT. I/O could be sent elsewhere if anyone wants this. [mcc] 17:38:17 23 Jul 2001 - 16.5.05c This version has the ML export changes and input level bug fix. [mcc] 16:11:19 19 Jul 2001 - totals in ML export The total*** keywords were supported for ML import but were not being generated by ML export. Now we accumulate and include in the output file the total numbers of files, tracks, edits, waves and marks. These are printed at the END of the project block. [mcc] 19:17:21 14 Jul 2001 - 16.5.05b: "New 24" softkey If the configuration is greater than 24-tracks, the Project menu now provides a "New 24" softkey to create a 24-track project file. This feature is available in all products except fame & prodigy. [mcc] 18:16:02 14 Jul 2001 - data moved A few Mb of MDR data have been moved from MC_XPRIVATE to MC_MDR ram. [mcc] 16:17:21 12 Jul 2001 - extensive minor changes Extensive minor changes have been made to prepare mdrmain for compiling (soon) as C++. [ajc] 23:29:59 11 Jul 2001 - Dynamics Mods Two new dynamics processing modules have been added to the real-time DSP program - a limiter and a compressor/limiter. Effectively this means that the expander/gate can be reconfigured to be a limiter. All functions now have an additional parameter hold-time which is applied before the release begins. The RTFX menu (BLUE-EQ) & display page (ESC-G) have been updated. A summary of DSP resource usage is now printed when the real-time programs are assembled; an error is generated if the DSP is overloaded. Bugs Fixed Since 16.5.05 ------------------------ [mcc] FFF#none 19:24:12 23 Jul 2001 - input level A recently introduced (16.5.05a) bug caused input levels to be incorrect if nonzero attenuation selected. [mcc] FFF#???? 09:16:42 19 Jul 2001 - ML import EDIT LIST MISMATCH If an error occurred during the EDL-construction phase of the import, then an EDIT LIST MISMATCH error was stupidly being reported. Release 16.5.05a, July 7th 2001 ====================================================================== Changes Since 16.5.04 --------------------- [spr] 9th July 2001 - FLFS file/directory length fixed at 28 chars. A new version of FLFS that limits the maximum length of a file name or directory name has been submitted. This is set to 28 to be consistant with the other file managers in the system. [mcc] 13:24:30 6 Jul 2001 - Serato v1.0816 This new version of the Serato package is much faster than previous versions. There is a fault, however, which only happens occasionally - sometimes it works perfectly. The fault is that the output is distorted in a strange harmonic way, as if amplitude-modulated at low audio frequencies. I have submitted this version because it needs to be tested on other systems than mine to see how it behaves. The reason for the sporadic nature of the fault is a complete mystery. [ajc] 00:56:13 6 Jul 2001 - Mixer Protocol mods New packets MX_RecordLevel & MX_Resource have been added. Record level control is implemented to 1/10dB resolution with smoothing. [ajc] 00:54:54 6 Jul 2001 - RTFX menu reinstated The RTFX menu and display page has been reinstated in MFXMFK console. The code has also been linked into EMU console but not yet tested (any volunteers?) [ajc] 00:53:26 6 Jul 2001 - various - some include files had prototypes using FILE which were conditional on _STDIO_H being defined - this didn't always work as the file might be included before stdio.h. The protos now use the underlying structure tag struct _iobuf (Microware specific) instead so don't need stdio.h to be included. [spr] 08:38 5th July 2001 - new CG5 Xilinx's (rev 22 and 8) This revisons should stop the shifting of the entire display by 8/16 pixels for good. The hardware now detects thsi "shifted" state and corrects itself. [ajc] 23:12:57 3 Jul 2001 - MRT DSP Programs Real-time DSP programs (QDC/MRT directory) are now generated automatically by a python script which reads aman_cfg and builds all the required DSP programs. The multitude of subdirectories in MRT has been removed from the repository. NB: python 1.5.2 or higher is required! [spr] 08:38 3rd July 2001 - new CG5 Data Xilinx (rev 7) A new CG5 DATA Xilinx which prevents corruption of the top 50 lines has been submitted. This corruption was only visible on the latest CG5's from production. The revision of this Xilinx is now visible on the startup screen and sysinfo detects if this Xilinx is out of date too. [mcc] 14:08:51 2 Jul 2001 - 16.5.04d The console is informed of the max number of tracks, that is, the number defined in the environment. A bug in Adjustable NTRACK is fixed, new 48-tk projects were not being created correctly when down-configured to 24-tks. [mcc] 11:51:19 30 Jun 2001 - Adjustable NTRACK 24-track project files can be opened from 32 or 48-track configurations without converting the file to 48-tracks. If you answer No (Clear) to the "Convert file to 48-tracks?" dialog then the file will be opened but not converted. Instead, the MDR software is down-configured to 24-tracks. A system message appears (briefly, alas) that says: "USING 24-TRACK CONFIGURATION". This is not possible for fame or prodigy products. New projects will be 48-tracks. A "New 24" softkey will be added soon to create 24-track projects from 48-track configurations. [spr] Thursday June 28 1:07:15 EST 2001 - FAT32 enhancments - 16.5.04c Edition 156 of IRBF has fixed the returning of disk space problem. We now have (AFAIK) a fully functional and working version of FAT32! An additional feature which has now been tested is FAT16 disks. Disks created by Windows95 in FAT16 formated can now be read on OS9. There is at this stage no way for OS9 to format a FAT16 disk however. [ajc] 23:53:34 27 Jun 2001 - JTAG mods Disabled auto-instantiation of all members of top-level assemblies. This causes jtag to crash when processing the test version of the qdc cdl files, though not when processing the normal cdl files. The reason for this has not yet been determined. For the time being the auto-instantiation is only per- formed when entering interactive mode (which is where it's needed) so the cdl files for testing can't be used interactively at the moment. [spr] Thursday June 28 04:16:36 EST 2001 - FAT32 enhancments - 16.5.04b Edition 155 of IRBF addresses these issues .... Free space calculation is now correct. If the drive returns that x bytes are free then x bytes can be added to a new or existing file. This was not the case before. The "seek" caching algorithim has been improved so MDR projects can be played backwards as well as they could be played forwards. Before this change, playing a MDR file backwards or even just jogging so the reverse que came into action would cause a noticable delay while the FAT's where chased on the disk. This has now been addressed. A problem still exists that when a file is deleted not all the clusters allocated to the file will be returned to the free pool. This is being looked at now. This should not impede FAT32 testing however. Bugs Fixed Since 16.5.04 ------------------------ [spr] - HFS Disks HFS disk can be mounted again. Release 16.5.04a, June 27th 2001 ====================================================================== [ajc] 00:41:31 27 Jun 2001 - JTAG command update Modified busstest to display more information about signals in error. Added interactive mode with some basic commands. Interactive mode is not complete yet. In particular handling of ctrl-C is incorrect. Fixed bug in component deactivation which caused port values to be incorrect the next time it was activated (mystery busstest bug). Changed slist/dlist headers to match those used in audioman (all inline). [ajc] 09:37:54 22 Jun 2001 - aodump Submitted an update to the AOUT Xilinx (now rev 10). I added a feature a long time ago which allows a number of output samples to be captured and stored in RAM (inside the Xilinx) for later readback by software. This got broken a while ago (rev 7 & 8) and I've just fixed it again (rev 9 & 10). The rev 10 xilinx (now in 16.5 tree) increases the number of samples captured to 16 (8 left, 8 right) - the older versions only captured 8. More interestingly, I've now made an OS9 command which triggers a dump and displays the data - this might be useful for diagnostic and test purposes to verify that audio data is reaching the AO card. You specify the output number you want to dump (default 1) and the data format (default hex). It might seem that 4 or 8 samples (per side) is not much use but I've found it to be enough to check that an output is working. (handy when looking for missing outputs!) If a test tone signal (eg sine wave) is used then the signal pattern can also be recognised. The command is called AODUMP and is currently only in the 16.5 tree - could be retrofitted to others if required. The compiled command could be used on 16.3 as-is. [ajc] 17:45:54 18 Jun 2001 - sysinfo Fixed incorrect QDC/Section number display in messages [ajc] 16:02:12 18 Jun 2001 - New Mixer Configs - 16.5.03h Added new configs for Station & Dream; max. mixer cfg now 48/48/48. A lot of new DSP programs added to support these cfgs. aman_cfg format changed slightly to make it look more "high-level"(!). Also added "stargate" variable to control enabling of stargate interface by main. [ajc] 09:33 16th June 2000 - 16.5.03g AMAN - Various bug fixes; improved mixdump output AMAN - Dump input channels as well QDC - Fixed SSI Mute handling Various other changes around the tree. [djs] FAME automation: The QDC was meant to send current timecode packets every 50ms in non-play speeds, but was actually sending them much more often. This behaviour has been corrected. [mcc] 18:31:23 10 Jun 2001 - 16.5.03e: Station & Dream Stargate is NOT enabled for Station & Dream. Input patching from Arm menu is temporarily NOT disabled, for testing purposes. MDR debug is enabled for incoming mixer requests. [spr] 16:49:32 8 Jun 2001 - /nvr/mdr_cfg Entries in the mdr_cfg file can now be made in a file called /nvr/mdr_cfg. Any directives in this file will override settings in the /dd/USR/SYS/mdr_cfg file. This allows a site to set options that will not be reset when a new software revision is loaded. Here is a list of all files that can be customised by the user. /nvr/profile - Script executed after every login /nvr/tcs_cfg - Runs after /dd/USR/SYS/tcs_cfg /nvr/mdr_cfg - Runs after /dd/USR/SYS/mdr_cfg /nvr/aman_cfg - Runs after /dd/USR/SYS/aman_cfg /nvr/startup - Executed after the /dd/startup script When any of these files are present, a warning will be displayed during startup so users/distributors/service know that the system could be running is a customised fashion. [mcc] 13:32:47 8 Jun 2001 - Default input type The mdr_cfg @DefaultInputType option has been reinstated. If there are inputs of more than one type installed, this determines which type will be used for the default input patching at startup. To avoid confusion, the argument is not a number but text, and can be "auto", "digital", or "analogue". Auto is the default - this will use analogue inputs if there are any otherwise digital. The parsing is not very fussy about the spelling. [mcc] 09:51:17 8 Jun 2001 - Zoom toggle Pressing and releasing the Zoom key with no other action will restore the previous zoom setting. This new feature is in Satellite, Station & Dream consoles. [mcc] 11:03:44 7 Jun 2001 - Station 48 track keys All 48 track keys and leds should now work correctly on Station. [mcc] 20:09:17 6 Jun 2001 - 16.5.03d: Station context switching The 68K console software for Station & Dream supports context switching. When raw key #148 is received the 68K will go into "sleep" mode. Whenever the modelevel changes, the 68K will issue MODELEVEL(8) event to the command porthole. The data for this event is: -1 68K is in sleep mode 0 no mode (software version is displayed) 1 mode 2 super mode 3 hyper mode The extra 16 track keys are incompletely supported in this version. It would be advisable to avoid using them - the 68K will get confused. [spr] 6th June 2001 - FAT32 Recycled Bin The "Recycle and _Restore" folders on FAT32 volumes are now not displayed in directory listings within the application . This like all other "system directories" (eg. CMDS/USR/SYS etc.) is in effect no matter what file system the folders are on. This can not easily be made file system dependent. [djs] Changed Vivid blank title names from BLANKY25, BLANKY3N & BLANKY3D to BLANK_25, BLANK_3N and BLANK_3D, to be consistent with the Vivid software. [mcc] 12:56:29 4 Jun 2001 - Distinguish crossfades The track display can distinguish the display of crossfades from fades. This is enabled with a new option on Blue S. When ON, fades are displayed in solid black, crossfades in the original horizontal-bars pattern. [mcc] 12:56:29 4 Jun 2001 - Sync offset in clip info The clip info display now includes the sync offset of the clip from the birthtime of the waveform. The display now truncates at a word boundary if the clip is narrower than the clip info text. [mcc] 09:52:21 1 Jun 2001 - 16.5.03c: Satellite performance enhancements Diagnostic delay (artificial load) removed from 68K software. FMC.exe features packet-based OUT422 porthole, 16bit jogger, and fix for recently-introduced clock interrupt restart bug. The mouse is once again enabled. [spr] 31 May 2001 - New version numbering. The version numbering scheme has changed a little. The 3rd group of digits in the version string will now be incremented when an official build is made. For example, all builds that leave R&D for testing should be like:- 16.5.03a 16.5.04a 16.5.05a etc. If an intermedate build needs to be be made for any reason, it is up to the submitter to increment the trailing letter. Anyone can now do this and is suggested when something major has been submitted. A final release build will follow the same procedure but will not have a trailing letter. So if 16.5.67d was the last ever beta release the final release version will be 16.5.68 [spr/rod] 31 May 2001 - New internal build 16.5.03b Changes to qdcinit to support production testing of AES cards [ajc] 30 May 2001 - qdcinit Added -e option to enable audio loopback - requires QDC Flash v1.29. Added sample rate parameter to -k option (default 48000 as before). Changed flash programming procedure to load flash.bin directly instead of loading bare.bin first (no longer required now that hardware is fixed!) [spr] 30 May 2001 - Splash Screens Preliminary splash screens are now installed for the entire product range. These are not final and will be replaced when the master artwork is availible. Bugs Fixed Since 16.5.03 ------------------------ [mcc] FFF#???? 08:32:17 20 Jun 2001 - Satellite macro bus trap When recording a macro, certain keys caused a bus trap to occur at the end of the recording. This has been fixed. [mcc] FFF#none 08:41:25 14 Jun 2001 - mdrplay exit #105 (divide by zero) The mdrplay task could exit with a divide by zero exception if other than the expected 32Mb RAM installed in a QDC section. This was due to a minor typo and has been corrected. [mcc] FFF#???? 4 Jun 2001 - Clipname truncation Clipnames were sometimes not correctly truncated when the clip is narrower than the name. Clipnames that end with a number are supposed to be shortened from the middle, retaining the number if possible. [mcc] FFF#none 4 Jun 2001 - "striped" bad waveform display The waveform display for waveforms that cannot be loaded or found was striped instead of halftone pattern. Release 16.5.03a, May 29th 2001 ====================================================================== [spr] 11:05:42 29 May 2001 - Revision 3/3a QDC's At the request of Tino and Ross, revision 3 and 3a QDC's will now not be accepted unless the machine thay are running on is a development machine. [mcc] 11:03:40 29 May 2001 - Import menu Borrow and keep can be cancelled by pressing borrow or keep again. [mcc] 12:53:02 28 May 2001 - Import 24-track files In a 48-track configuration, it is now possible to open a 24-track project from the Import menu. The file version of the 24-track project must be at least 0.25. This is the standard file version, created with software 12.0.37 and later. To import from an older file version, open the file first from the Project menu, using a 24-track configuration. This will automatically upgrade the file to 0.25. The borrow and keep operations allow (require) the selection of the destination tracks when multiple tracks are selected. Previously you could only select the destination track if a single track was selected. Now you always select the destination regardless of how many tracks are selected. This is done by pressing a single track key to indicate the destination for the first (lowest numbered) selected track. This assigns destinations to other tracks according to the "selection pattern", like the MFX/Merlin Edit menu. For no change, to import clips to the same tracks in the project, you can press Enter instead of a track key. These changes apply to all consoles, all products. [ajc] 23:00:05 27 May 2001 - Check for QDC Rev 3/3a Sysinfo now checks "Hardware Mod Level" of QDC which indicates the effective hardware (PCB) revision level. An error is now reported if any rev 3/3a QDCs are present since these do not work properly with 16.4+ software. Rev 3/3a cards which have been modified to rev 4 standard (TSBMAP mod) are accepted. [ajc] 22:45:16 27 May 2001 - Audioman, "stn 48" product cfg Major rework of audioman to support internal monitor matrix, buss inserts, 6-band RT EQ and other features required by Dream product family. All high-level configuration is now modelled by connecting logical signal Pads which are independent of DSP resources. This allows most patches and parameter updates to be performed with a minimum of disruption to audio processing. The current FAME protocol level is now 8.1. The first configuration to use the new features is Station48_8_16. --- The MRT (real-time FX) DSP program now also performs mixing functions. This allows mixing & RTFX to be distributed more evenly among DSPs. The current mixer program MIX will soon be defunct. The MDR (disk recorder) program is still an independent dedicated program but may be integrated at some time to allow completely general distribution of DSP resources. --- Under $FAIRLIGHT/DOCS/MRTCalc.xls is an Excel spreadsheet which calculates DSP usage for various resource combinations in the MRT program. [spr] 09:10:42 - 16th May 2001 - Startup Scripts New startup scripts have been added. These include "sat" with options of 24,32 and 48 - defaults to 48 "stn" with options of 24,32,48 which relate to 24/8/16, 32/8/16 and 48/8/16. Again the default is 48. dream with one option at the moment of 48. which is set to 48/48/48. Other configutations will be added as merketing decide what they need. [spr] 09:12:12 - 16th May 2001 - New dongles New product members have been added to the dongle database. These include dream, station and satellite. All Golden dongles now have these products by default. [mcc] 16:51:20 23 May 2001 - Satellite Version incremented to 16.5.02b-3x. Binnacle: For split-clip with range, the cursor must be in the range. The reverse softkey is available when there is no range. The bounce softkey is available when there is a range. Arm menu: The sample depth (width) softkey has moved to here and is now a tri-state switch (enter key is not required). Use the input softkey to access input patching. The Mode softkey should now work (it was inhibited before). Track key LED colour scheme is safe = dim green thru = medium amber armed not recording = pulsing red armed recording = bright red other = dim red Level menu: Individual softkeys are provided to select the domain, clip, track or display. The Info softkey is always provided. AudioBase menu: Search and edit options are on the top softkeys. The file menu can be exited by pressing file again. ADR menu: Options that were previously on more softkey are on upper softkeys. [mcc] 12:58:17 17 May 2001 - v16.5.02b-2x The version is incremented to ensure console compatibility due to a change to the format of one of the request packets. [mcc] 12:58:17 17 May 2001 - Satellite The Trim/Slip menu has separate trim & slip softkeys. The Fade menu has capture head/tail softkeys. A bug is fixed in Fade menu that allowed All Layers and Razor to be turned on. [mcc] 17:17:29 16 May 2001 - Satellite The additional five softkeys are now supported. So far only the Cut, Copy and Erase menus feature "upper" softkeys. [mcc] 17:14:55 14 May 2001 - Satellite The Track Sel key is implemented. Bugs Fixed Since 16.5.02b ------------------------- [ajc} 22:58:34 27 May 2001 - remotedc Fixed buffer overflow, reinstated verbose flag [mcc] FFF#???? 18:57:04 24 May 2001 - ProTools Bad Path Number A buffer overflow bug in writept corrupted the stack and destroyed the path number to the session file, among other things. This resulted in Error #201 - Bad Path Number. [mcc] FFF#none 13:26:52 19 May 2001 - Stargate connection status Fixed a bug that could make Stargate DMS appear to be online when in fact it isn't. There was some confusion in the interface about whether non-zero meant error or true. [mcc] FFF#none 16:41:19 18 May 2001 - send_rpc error message A bug is fixed that caused the tvt error message printed by DFNLIB/dfnlib.c::send_rpc() to report the IP address as 0.0.0.0. See the note in that function. [mcc] FFF#???? 12:28:11 16 May 2001 - local node not responding The popup notification of reconnection to local node is disabled by default as the fix has been proven to work. If desired, the popup can be enabled, by setting the ffs -n option in /dd/startup. The ffs help sheet has been fixed to show this and other options. [mcc] FFF#???? 08:01:17 14 May 2001 - double-click track Fix bug introduced with 48-track support: double-clicking a track key didn't work properly if the first click triggered the assertion of the track selection pattern. Affected MFX & Merlin Edit menu and Satellite Cut/Copy/Erase menus. Release 16.5.02b, May 8th 2001 ====================================================================== Changes Since 16.5.02a ---------------------- [mcc] 16:53:25 10 May 2001 - Satellite & FMC.exe The program for the Satellite console CPU is bundled with the release at /dd/USR/CONSOLE/FMC.exe. To install, use ftp from Satellite to QDC, do binary transfer to root of C: drive. E.g (from Satellite): ftp cd USR/CONSOLE bin get FMC.exe bye Note that FMC.exe cannot already be running. Exit it first if it is. Some bugs and oddbods have been fixed in the Fade menu. The jogger uses 16bit register for smoother control. Bugs Fixed Since 16.5.02a ------------------------- [mcc] FFF#???? 17:32:19 9 May 2001 - local node not responding The fault where the network vanishes from file browser may be due to a stale RPC connection handle in FFS. A change has been made to detect this and fix the problem on the fly, if possible. If the fix succeeds then window will pop-up with the following message: Re-connected to local node. This is good news. If the system continues to work normally after you dismiss this window then a serious bug has been fixed. Please inform Fairlight that you saw this message. The popup will not appear more than once in a session. [djs] The bug with locating a vivid using the DMAN_VIVID driver has been fixed. However, the fix involved disabling the remaining-time-sense operation. This means that the vivid project screen will always show that there is no free space on the vivid drive. This will be fixed in a future release, but for now its less important than not being able to locate! Release 16.5.02a, May 8th 2001 ====================================================================== Changes Since 16.5.01g ---------------------- [mcc] 19:34:11 19 Apr 2001 - binnacle two-handed editing The so-called "two-handed" editing method is now supported on the binnacle Cut, Copy and Trim menus. [spr] 11:35:12 20 Apr 2001 - PY's Vector Base Register PY used to inherit the Vector Base Register (VBR) from the PY static ROM. This has now been changed so the VBR is now in PY's private RAM. [mcc] 16:22:48 20 Apr 2001 - Satellite fixes and features The split operation will work with a zero-width range, so that you can make a single split with All Layers on. The erase operation with Razor ON will allow you to erase an empty range (containing only silence, no clips). A bug is fixed which did not correctly recompile all affected tracks following a razor cut or erase. [awk] 07:12:09 20 Apr 2001 - Medialink Improved the time required for "dfnserver" to build the cache when it starts up. [mcc] 10:34:55 25 Apr 2001 - Satellite The Play/Jog key is implemented. [mcc] 10:35:12 25 Apr 2001 - mdrmain 16bit data refs Changed mdrmain to use 16bit data offsets because the data section is less than 64K (32bit offsets are only required for >64K data). This in turn reduced the size of the code by nearly 30K! [mcc] 08:02:29 26 Apr 2001 - Satellite The MFX Play Menu is installed. The Play/Jog key LED is active. [spr] April 26 16:03:33 EST 2001 - Prodigy Startup Prodigy now has a 48 track startup option. Prodify 32 is still in place and will be removed in a future release. [mcc] 17:44:16 26 Apr 2001 - Satellite The Razor ON option is implemented for Paste. A region of silence equal to the width of the clipboard range is inserted at the cursor position. [mcc] 14:44:51 28 Apr 2001 - Satellite The clipboard offset depends on the current state of Razor ON. That is, it is not fixed at the time of cut/copy but determined when required for display & paste. [mcc] 14:45:19 28 Apr 2001 - [tcio]mask_btst() parameter The first parameter for [tcio]mask_btst() is now a pointer to the mask, not the mask itself. [mcc] 14:21:11 30 Apr 2001 - flfscheck v2.08 The flfscheck fix and copy options are only authorised if the R&D dongle is detected (ie, for R&D machines). These options are experimental and untested, they were only ever intended for the author's private use. Disks with damaged file systems should be delivered to R&D for the software department to inspect and repair, if possible. [mcc] 16:14:53 30 Apr 2001 - Satellite Macros are implemented in Satellite, however the saving and loading of macro files is not well tested yet. The "To Panel" menu is not yet implemented. [mcc] 13:07:29 1 May 2001 - Satellite There are now two types of macro files, Satellite and MFX/MFK. The MFX macro file is currently version 1.2, Satellite is 2.0. The allowable range of macro file versions is sent to the host for verification when loading a macro file. An error results if you attempt to load the wrong file type. The current console macros will remain unchanged in this case. [mcc] 13:17:05 1 May 2001 - Version 16.1.05gx Set version to 16.1.05gx to ensure sync of Satellite & host software for evaluation demo. [mcc] 13:41:15 1 May 2001 - mfxload It is now a fatal error if the Satellite console is out of date. [mcc] 07:27:22 3 May 2001 - Satellite The new two-handed editing rules are implemented (action on release, no timeout). Paste turns ghost off. Browser key actions revised: PageUp/PageDown = Contract and expand (key caps to be changed). UpArrow/DnArrow = Move up/down 1 line Blue UpArrow/DnArrow = Move up/down 1 page (still needs refinement) The Files/Fldrs LED reflects the current browser focus. All browser key LEDS are dark when browser is not active. Regarding "Super Trim": this feature is full of bugs. Please do not confuse binnacle testing by using super trim at this time. The bugs have nothing to do with the binnacle. [mcc] 08:42:11 3 May 2001 - 16.5.01g-1x Incremented version to 16.5.01g-1x to force console sync. Subsequent versions would be *-2x, *-3x, etc. Please follow this pattern by incrementing the number before the 'x' to produce interim versions between official builds. [spr] 11:53:12 3 May 2001 - Fake 96k Enhancment Digital Input sync will now work when started with the 96k option. [mcc] 11:42:01 7 May 2001 - Satellite The Print and Takes menus are installed. The pop softkey has been removed from the Track menu. The active binnacle mode key is red when Razor is on. A Range behaviour inherited from MFX & Merlin has been changed: entry to an optional-range mode would clear a pegged zero-width range (pegged-at-now). This is no longer the case in Satellite, such a range will remain and not be cleared (in Manual range style). The number pad Blue Subf LED and Blue N-Track LEDS tally the current state. The console sound generator is active and making noises. [mcc] 16:46:21 8 May 2001 - Satellite The binnacle Fade menu is properly implemented, rewritten in 'C'. Blue head/tail perform the "capture" function until the additional 5 softkeys are implemented. NOTE - this version still expects the clock interrupt at 1KHz. ============================================================= Bugs Fixed Since 16.5.01g ------------------------- [mcc] FFF#none 10:32:11 24 Apr 2001 - flfscheck v2.7 The flfscheck program had a bug that caused it to think the file system was corrupted when it wasn't. The bug could cause flfscheck to hang forever, bus trap, reboot the system, or report non-existent errors. [mcc] FFF#??? 09:51:02 20 Apr 2001 - fatal TRAPV exception The exception vector to handle trapcc instructions was not initialised for PY, hence the trap was fatal and resulted in the dreaded "Flight IPI timeout" error. This has been fixed, however the trap is still expected to occur because the underlying fault remains unknown. Now, instead of self-destructing, PY will print a message to the tvt, something like this (the details will vary): TBASE-task: trapeq.w 610a at 5021f4a8 Please watch out for and report any instances of this message. [spr] FFF#none April 24 14:53:11 EST 2001 - TRAPV popups In an effort to catch the TRAPV exception a popup will now appear which will have to be cleared by the user if the TRAPV situation should occour. [spr] FFF#none April 26 16:14:38 EST 2001- CG5 Xilinx Revision 21 of the CG5 Xilinx fixes some bugs with the latest batch of CG5's from production. The only visible effects of the bug were observed in the "Patch" page. The Small rectangles displaying the patch matrix were being drawn incorrectly. This has now been fixed. This version of the Xilinx has been backported to 16.3 and 16.4. [mcc] FFF#???? 13:49:17 4 May 2001 - file browser cockup An ancient bug has been fixed in all consoles: the file browser status was not correctly being pushed/reset/popped with changes to panel assertion level (mode,supermode,hypermode). This could leave the browser active when it shouldn't be, for example, if you go to Edit Loop from the MFK Project menu. [mcc] FFF#???? 20:53:57 4 May 2001 - EditMacro/file browsing The file browser was completely broken in the EditMacro/file menu. Only [Blue]+/- and jogger worked, From/To or Jump would cause the EditMacro menu to exit. This has been broken since the beginning of Rev15. Affected files: SRC/MFX4/MFX/CONSOLES/MFXMFK/keytask.a SRC/MFX4/MFX/CONSOLES/M4R/keytask.a SRC/MFX4/MFX/CONSOLES/MFXMFK/macrofile.a SRC/MFX4/MFX/CONSOLES/MFXMFK/MDR/mdrtask.a SRC/MFX4/MFX/CONSOLES/MFXMFK/MDR/project.a SRC/MFX4/MFX/CONSOLES/M4R/MDR/project.a [ajc] May 8 11:43:35 EST 2001 - DSP Crashing on Overload Tailor max playback speed to number of chans; fixes to TSB overload handling. Release 16.5.01g, April 19th 2001 ====================================================================== Changes Since 16.5.01f ---------------------- [mcc] 08:19:49 19 Apr 2001 - Satellite startup script. To use the Satellite console you must use the new "sat" startup script, not merlin. Bugs Fixed Since 16.5.01f ------------------------- [ajc] FFF#QDC 1023: Arming failed on sections with only one MDR program. This has been fixed. Release 16.5.01f, April 9th 2001 ====================================================================== Changes Since 16.5.01e ---------------------- [spr] April 5 12:04:27 EST 2001 - 96kHz The method of editing tcs_cfg to enable 96kHz has been removed. To enable a simulated 96kHz system now start melin with the command line "merlin 24 96k" or "merlin 32 96k" or "merlin 48 96k". Without the "96k" the system will behave as before. (replace merlin with mfx3 if you need to try mfx3 at 96k) If you need "fame" or "prodigy" to run at 96kHz then set the environment variable "MDRFAKE96K" to "1" before starting the product. Only merlin and mfx3 have the extra argument as only merlin and mfx3 will be demonstrated at the show running 96kHz. All this code will be removed when we remove the smoke and mirrors. [ajc] April 5 - IOPACK/IOPLIB Added the following escape codes to control what happens at the end of lines: u - enable autolf v - enable auto backspace (last char on line is overwritten) w - disable autolf&autobs (chars after eol are thrown away) [spr/rod] - April 9 16:40:33 EST 2001 - New AI Xilinx AI1 Xilinx is now revision 5. This fixes a bug with the RDY signal (but is not currently causing problems). [spr/rod] - April 9 16:40:33 EST 2001 - New PXY CDL file Needed to support a hardware ECN about to happen on the PXY card. This is backwards compatble so all existing PXY's should still be OK. [spr] - April 9 16:40:33 EST 2001 - FAT32 FAT32 is now working much better for removale drives. In the mean time please try and keep all file names to the DOS 8.3 format. It is possible for a file name to be accidentally converted to 8.3 format if it uses a long file name. Bugs Fixed Since 16.5.01e ------------------------- [ajc] FFF#none April 5 - MIXINT Allow filter types 4..7 in FeedEQParam & InputEQParam msgs This fixes a fame version 8 protocol bug. [djs] A bug existed where arming and disarming tracks on the MFX would not send the correct arming messages to FAME. This was caused by a structure being too big for the char-array it was being stored in, and corruption of the next thing in memory resulted. The structure size has been corrected. Release 16.5.01e, April 3 2001 ====================================================================== Changes Since 16.5.01d ---------------------- [mcc] 09:48:23 3 Apr 2001 - Serato V1.5 Version 1.5 of the Serato Time & Pitch algorithm is in this release. Some improvement in execution speed is expected. [spr] March 26 12:30:19 EST 2001 - Dongle Checking Much stricter dongle checking is now done. Before this change there was no checking for FAME or PRODIGY at all. Now only what is licenced will be eligibale for startup. [mcc] 16:23:29 28 Mar 2001 - Edit menu CUT/COPY default Pressing Blue-softkey1 will change the CUT/COPY default in Merlin and MFK Edit menus. Shift-softkey1 already did this in MFK but Merlin has no shift. Now blue can alternatively be used for this purpose in both consoles. [spr] April 2 13:14:35 EST 2001 - FAT32 Implementation. The first official release of FAT32 is now implemented. Testing should begin as soon as possible. However, there are some known problems ... 1) Simultaneous access of files on the same disk by reading or writing can cause corruption. So dont share files on the network just yet. Opening a single project is fine. No problems doing this have yet been seen. 2) When a disk is mounted and then removed and modified by another system and then returned to the original machine, the system must be rebooted in order for the changes to appear. 3) Disks can not be checked for integrity yet. Further more Microsofts "chkdsk" will fail to read disks that have been formatted by diskinit. However disks that have been formatted by Windows can be mounted and checked as one would expect. 4) Only the 1st partition will be mounted. All other partitions will be ignored. 5) The boot drive cannot be FAT32 yet. 6) Speed is fast but not quite as fast as FLFS/RBF yet. Work will be done on this when the above problems have been fixed. None of the above problems should prevent testing to commence. To format a disk for FAT32 use diskinit as you normally would except you should specify a -f option. For example "diskinit /sd02 -f" [ajc] 00:10:24 3 Apr 2001 - JTAG command updates Added new facilities for a) blank checking a CPLD before erasing and b) assuming a CPLD is blank when programming. These changes are primarily to speed up programming new QDC & PXY cards. Bugs Fixed Since 16.5.01d ------------------------- [mcc] FFF#none 18:14:11 28 Mar 2001 - divide by zero Fixed bug that caused mdrplay to suffer divide-by-zero exception. This occurred when one or more MDR DCCs have no MDR channels, and was particularly likely for tracks dongle = 16. [mcc] FFF#none 07:17:51 28 Mar 2001 - Blue split The "Blue split" feature (split all clip layers) migrated from Rev15 in June 2000 did not work when there was a range. [mcc] FFF#none 07:19:02 28 Mar 2001 - 48-track Level menu The Level menu in all consoles did not correctly support 48 tracks. The interactive level and pan features would not have reacted reliably to changes of track selection. Specifically, the test for all-track-keys-up was only checking the first 32 keys. [mcc] FFF#none 07:22:31 28 Mar 2001 - Range in Fade menu A minor bug in the Fade menu did not correctly handle the return status from the Range sub-menu in all cases. [spr/rod] FFF#none - 26 12:30:19 EST 2001 - Updated CG5 Xilinx The vertical sync position in Highres CG5 Xilinx revision 19 had the side effect of reducing the back porch by one line. This caused the scan rate to go from 75.03Hz to 75.13 Hz. This has been corrected in rev 20. [spr] FFF#nome - April 2 13:14:35 EST 2001 - Reserved Name Database The was a problem with the reserved name database when the boot drive was an RBF drive. Permissions were being enforced by RBF that FLFS does not do. Release 16.5.01d, March 22 2001 ====================================================================== Changes Since 16.5.01c ---------------------- [ajc] 00:11:43 26 Feb 2001 - "Distortion" Bug fixes Updated QDC_SUPP Xilinx to provide level-sensitive TSB interrupt. This means that the interrupt is not lost if previous sample interrupt processing was not complete before the next interrupt occurred. It appears that this can happen very occasionally due to unpredictable delays resulting from arbitration conflicts for the external DSP buss. Also added hardware to detect interrupt overrun condition. DSP code for MDR and MRT changed to make use of these new features. All DSP code restored to latest state before the distortion problem was discovered. FAME features MS Decoding & Test oscillator are now back in. [spr] 13:13:43 26 Feb 2001 - Enhances for AI1 CPLD The AI1 .SVF file is now not hand patched (version 3 was). Also some future control bits have been added to the control latch. [ajc] 23:06:44 28 Feb 2001 - Corrections to distortion bug fixes [mcc] 20:49:17 2 Mar 2001 - Serato Serato time-n-pitch software version 1.03 integrated. [mcc] 13:12:11 3 Mar 2001 - Merlin Wave & Razor menus The Wave menu is available in Merlin on the key to right of Name Loc. The Merlin Razor menu is now the same as the MFK Block menu. The first softkey selects RAZOR or DUBBER operations. Holding Blue or Shift when pressing the first softkey, sets the default for the switch to the selected state. [ajc] 01:08:23 8 Mar 2001 - Varispeed / TCS mods Dub's varispeed mods integrated into TCS (and a few in PX/MDR). In the process, all modules in TCS have been converted to compile in C++ mode. Be on the lookout for any possible (unlikely) side effects! [ajc] 01:10:48 8 Mar 2001 - Play Task compiled as C++ The PLAY directory has been modified to compile in C++ mode. This can be disabled by removing the "-qp" options from the makefiles in PLAY, PLAY/CC, PLAY/INT and PLAY/WFLD. Quite a few defs files in MFX/DEFS have also been modified to make them palatable to the C++ compiler (but still C compatible of course) [spr] Thursday March 8 16:06:59 EST 2001 - feature file The feature file can now be executed with a -b option. This allows any machine which is currently golden to change ID to any other machine in the database. However, if you change ID to a machine that is NOT golden, then you will not be able to change ID again. If you get in this state, reboot the system and then assign the ID you want. Example run "/dd/features -b=03888484" will change the ID to FairlightUSA service. As this machine is Golden it can be changed again. If you run "/dd/features -b=0617aa62" then once the change has happend you will never be able to change it for the life of the current session. [ajc] 22:58:13 11 Mar 2001 - new DSP program mdr3[n] A 3 channel MDR program has been created for use in configurations which currently have excess DSP resources and for testing 96k operation (the final 96k program[s] may have further modifications) Audioman cfg file aman_cfg has been modified to use this program wherever appropriate, generally where previously the 6 channel program was loaded on the primary and no program (mio) on the secondary. This distributes the load and gives better jog/shuttle performance. NB: total channels per section should always be even to avoid stereo allocation problems. [spr] 12 Mar 2001 - Power Supply Readings All power supply readings and errors are now based on +-6%. Used to be +-5%. This is to take into account errors in PXY's inability to read the same value over many batches of PXY's. [spr/rod] 12 Mar 2001 - HiRes CG5 Xilinx Change - Now rev 18 Fixed CG5 sync timing to exactly match VESA standard 1024x768 @75 Hz. This should fix the problem of the picture being too wide on a lot of monitors. Also changed sync polarity to positive, this may improve performance on some monitors. You will have to adjust your monitor to suit the new timing. The cg5hr diagnostic utility has also been changed to program dotclock for the correct rate. [ajc] 23:36:00 13 Mar 2001 - Added 2 track MDR programs The two track programs are used where previously mdr4[n] was running on one DSP only of a pair. This will improve jog performance. Note that the DSPs can now play at up to 9 times play speed (when lightly loaded); however the play task currently halts all streams at 5x play speed. [spr] 14th Mar 2001 - noded changes Make sure when creating/unlinking modules that the module is the correct type before just blowing them away. This should prevent things like a machine called "kernel" or "mfx232" from comming online and unlinking perfectly good modules. The way that hosts can be given any name is a design oversight. Perhaps all the noded modules should be prefixed with "_". But this would require lots of other code around the tree to be adjusted. [spr] 14th Mar 2001 - 96kHz Changes The SIO1 frequency doubling configutation is now held in tcs_cfg. This doubles all of the standard SIO1 sample rates. It is to be used by testing to see if we have any problems running at 96kHz. To enable it add "@SETENV DOUBLE_SPEED=1" to your tcs_cfg file. This can be in /nvr or in the normal place /dd/USR/SYS. This also enables us just to have one build and try 96kHz instead of the way it was which required two seperate builds. [mcc] 08:28:19 16 Mar 2001 - MFK2 console reinstated. The MFK2 console is again supported. To begin using this alternative key-layout you must manually load the correct software by typing: mfxload /dd/usr/console/mfk2.raw [spr] March 19 16:44:48 EST 2001 - Reserved Name Database Die to a recurring problem of machines being named the same as some existing OS9 modules, a new database is generated after a new version of software is installed. Before a machine can go "online", the name of the machine is checked against all data in this database. If a match is found the software will disallow noded from creating an OS9 device descriptor with the same name. For example, under all previous versions of software it was possible to call a machine "mfx232". This resulted in the real "rs232" scf/serial device descriptor being replaced with a netman device descriptor. Once this had happend, the mfxload command would fail to work. The problem is even more severe if a machine was called "merlin" or "fame". OS9 would find a device descriptor in memory called "merlin" or "fame" and attempt to execute it instead of one of the startup scripts. The reserved name database now fixes this potential problem. NB: in all prior version of software, it is still possible to create a machine called "merlin" and should be avoided at all costs. [mcc] 17:27:44 21 Mar 2001 - fame IO limits There can now be a maximum of 48 inputs or outputs of each type in a Fame or Prodigy configuration. This number was previously limited to 32. Bugs Fixed Since 16.5.01c ------------------------- [spr] - Wednesday February 28 15:34:08 EST 2001 SIO1 Sync Xilinx Rev 55 corrects the internal sync frequency and is needed if the PXY card is Rev J or later. (i.e. has a 14.32Mhz crystal at X1) [mcc] QDC#0989 14:52:09 28 Feb 2001 - No 'S' after Delete A very old bug in Merlin prevented you from typing 'S' after or 'P' after and a few other combinations. [mcc] FFF#none 20:49:17 2 Mar 2001 - Serato bugs Several bugs are fixed in the Serato ODIF modules. One of these caused the right side of stereo to be retarded. [mcc] FFF#none 10:56:42 3 Mar 2001 - Bounce Track N could not be the bounce destination if track N-1 had stereo input configuration (two inputs patched). Fixed by cancelling ALL input patching before the bounce and restoring afterwards. [spr] FFF#none - Tuesday March 6 18:09:11 EST 2001 - Diagnostics sio1test - ltc is working aes sync and wclk sync tests now detect the input rate correctly. [spr] FFF#none - Tuesday March 6 18:09:11 EST 2001 - SIO1 corrected 3ppm error in internal clock mode. (rev 56 SIO1 Xilinx) [ajc] FFF#none - 22:53:05 11 Mar 2001 - Crashing QDC Bug Updated QDC-SUPP Xilinx - now rev 13; QDC Flashware v1.26 The bug seems to be caused by the secondary DSP missing interrupts, possibly due to a marginal timing problem in the rev 12 Xilinx. The rev 13 Xilinx increases the length of the interrupt pulse from 1 to 3 clock cycles. [ajc] FFF#none - 15:38:46 12 Mar 2001 - Restore Output Mode after Blue-Stop After blue-stop audioman was not restoring the correct output mode (stereo/ multitrack); multitrack mode would be active irrespective of menu setting. Now fixed. [spr/rod] 14 Mar 2001 - SIO1/LTC Xilinx Chnage - Now rev 14 This fixes FFF#930, which reports that a 1KHz tone is being output from the SIO1 LTC port when the transport is stopped. When the transport is stopped, there is no longer any output from the LTC port. [spr/rod] March 19 16:44:48 EST 2001 - Revision 19 CG5 Hi-Res Xilinx This xilinx also corrects some other bugs. 1. Auto increment is now modulo 768 the vertical direction. 2. Random shift of entire screen to the right after a palette write is fixed. 3. Early palette change between top and bottom palettes is corrected. 4. Vertical sync pulse starts before Horizontal sync. Release 16.5.01c, February 23 2001 ====================================================================== Changes Since 16.5.01b ---------------------- [mcc] 18:27:13 12 Jan 2001 - ped reinstated The palette editor (ped) has been modified to work with the current 8-plane hi-res graphics format. The application has also been upgraded to use relative screen addressing instead of the old absolute mode. [spr] Tuesday January 16 16:43:55 EST 2001 The system wide tvtline system is now functional for the TCS static tvtline display. The old method (using xtvt) is no longer supported. To see the new system in action start a gfxterm or an xterm and type "tvtline tcs". You should now see the TCS debug variables changing. To exit press control C/E. [mcc] 15:27:19 17 Jan 2001 - startup scripts The merlin,mfx3,fame & prodigy scripts all profile a new file, /dd/USR/SYS/mdr_set, to setup the various MDR environments. The PRODUCT_CONFIG variable is printed for inspection by these scripts and also by mfxstart when starting audioman. [spr] Wednesday January 17 17:38:04 EST 2001 The splash screen during startup has been reimplemented. A new option in the NVRAM setup can turn this feature on and off. All existing system will have this default to off so it will have to be explicity enabled. It is however on by default if NVRAM is ever reconfigured. The screen images need to be revamped for 1024x768 as the ones that are currently installed are just resized from 512x512. The current raster file system is probably past its use by date and a new system using many more colours will more than likly be written very soon. This should now only require modification to 2 files to display a new type of image format. All the code to support it is now in place. [spr] Wednesday January 17 17:38:04 EST 2001 The "slideshow" command has been ported. You can now see a slide show of all the pictures in the /dd/USR/PIX directory. This command can only be run from within the window manager. Shell out and type "slideshow". Control E/C to quit. [spr] Thrusday 8th Feb 09:47:00 Modified the features program so timefx() can take an argument of (2). When this is set, SERATO is enabled. The GOLD macro has also been changed to enable SERATO on all GOLDEN machines. [mcc] 11:51:27 8 Feb 2001 - Serato The Serato time & pitch processing algorithms have been integrated into the MFK Wave menu. The Algorithm softkey has 3 more choices, Serato 1, Serato 2 & Serato 3. (These are dongled by timefx=2). The Scale softkey is disabled (not needed) when Serato is selected. The VARI submenu operations will use Serato[123] when selected, even though the Algorithm softkey is not present on the menu. PITCH operations are still performed in two passes. It may be possible to do these in one pass with Serato, but not yet. The Serato modules are not used for SRC related to other features. [mcc] 11:55:48 8 Feb 2001 - DFN unlimited files Dub's DFN "unlimited files" feature is in this release. This affects all dfnservers, os9 & PC. For os9, the maximum size is still arbitrarily limited to 4096. The file cache is much more dynamic so we should watch out for memory leaks. Test that repeatedly ejecting/inserting removable media does not consume more and more memory. Changes have also been made to low-level UDP functions - something to do with caching of connections. [mcc] 10:38:19 9 Feb 2001 - Serato pitch-shift Serato pitch-shift is now done in one pass. [spr] Monday February 12 12:21:46 EST 2001 Any system can now display the information about any other system in the encoded features file. To use this type "/dd/features -m=NUMBER" where NUMBER is the MACHINE id you want to see. Previously you could only run this on a machine that was "GOLDEN". This was authorised by Ross Caston and Tino on 12 Feb 2001 [spr] Monday February 19 16:08:48 EST 2001 The SIO1 xilinx has been updated to revision 54. This version has the double clocking option that will be required for 96kHz operation. It appears to be after some testing that it is perfectly compatible with revision 53 as long as the double cloking bit does not get set! Also the "sio1" diagnostics program has been reinstated. [spr] Thursday February 22 14:54:13 EST 2001 It is now possible to control an QDC Engine from a Console located on another QDC engine. This is done over ethernet (TCP/IP). The code is NOT enabled by default, and will work EXACTLY as it did before this change. The two critical files that have changed are $MFX4/LIBS/ESP/CONSOLE/comms.c and $MFX4/MFX/CONSOLES/CONSOLED/main.c When the NVRAM option to enable this feature is set, the data to and from the /mfx422 device is sent and received via two named pipes instead. Data to and from these two named pipes is filled and empted by two new tasks called consvr (console server) and conclt (console client). The machine to connect to must be specified in the /nvr/setup file. By default this is set to localhost. This tests the code very well as it emulates an external console but uses the internal console instead, thereby testing the code to it maximum. Using this method jogging works OK although the load on PX is a little higher. NB: The /term device is not handled yet so anything that requires input from /term will have to be done on the local console. Also the console version checking will test the local console and NOT the remote console. [spr] Thursday February 22 16:01:09 EST 2001 More 96kHz features. To test 96K sample mode do the following. edit MFX/PY/TCS/TBASE/sio1.c and uncomment the following line. #define DOUBLE_SPEED_48000 edit MFX/PX/AMAN/io.cpp and uncomment the following line. #define DOUBLE_SPEED_48000 Now when you select 48k rate the machine will actually run at 96k Bugs Fixed Since 16.5.01b ------------------------- [spr] FFF#none Tuesday January 16 11:59:35 EST 2001 A major bug has been fixed in qlibres which could have caused any function calling tvtline or tvtfield to corrupt memory. This has now been fixed. At the same time, an old defsfile (qsysmod.h) has been removed from the tree. This will require a "make clobber; make install" to be done from $MFX4 to rebuild many make dependances. [mcc] FFF#none 09:43:10 10 Jan 2001 - M1,M2 menu data corruption The M1 and M2 menu state variable (statenum) had the wrong storage size (ds.b instead of ds.l). This caused corruption of adjacent memory - which happened to be the "Arm Trks" bank selection. Using Blue-M1 would corrupt M2's bank selection and vice versa. (This bug was actually fixed in 16.5.01b). NOTE: 15.7 has the same bug. [mcc] FFF#none 18:27:13 12 Jan 2001 - M1,M2 softkey behaviour In M1 & M2 menus, "choice" softkeys (LED blinks, Enter required) did not behave correctly. Specifically, pressing the key again did not cancel it. This was due to blown registers for every call to choice_loop() or simple_choice_loop(). For these functions, d0 and d2 must contain the values passed to the CHOICE, MENU or SWITCH. These are the keycode(d0) and key index(d2). NOTE: 15.7 has the same bug. [mcc] FFF#none 15:38:21 16 Jan 2001 - GPI fire softkey The "fire" softkey code had a very old bug. It's supposed to do different things depending on the GPO type (first softkey). But confusion about which data register meant that the behaviour for types >= RECORD would not be correct. PLEASE TEST GPI menu "fire" softkey for all GPI types that support it. Please compare to behaviour in an earlier software version, before this fix. ALL software versions have the same bug. [mcc] FFF#none 15:38:21 16 Jan 2001 - GPI debug crash MDR main task would crash with a buss trap if GPI type MARKS (or LOCS) was selected with MDR U option set. [mcc] FFF#none 15:38:21 16 Jan 2001 - FAULTY PACKET Merlin reported "FAULTY PACKET" when quitting due to ViVid incompatibility. [mcc] FFF#none 18:06:00 16 Jan 2001 - Crazy meter set Under certain conditions, the meter window would display a crazy or invalid set of tracks. Fixed by removing the lo-res bank selection code that had been inadvertently left in hi-res version. Also modified to keep ptr rather than index to meter display format, for increased speed. [ajc] 11:48:53 23 Jan 2001 - Record Level Incorrect Bug introduced in 16.4.01f by change of floating point level range from [-0.5,0.5) to [-1.0,1.0). The record level was twice as high as shown on the meters. [mcc] FFF#none 11:52:21 25 Jan 2001 - Merlin Undo/Redo not disabled Fix bug in Merlin - Undo/Redo were not disabled when they should have been. Undo & Redo are now disabled in Merlin identically to MFX console. [spr] FFF#none Thursday February 15 13:46:58 EST 2001 The ROM no longer bus traps when no CG5 is installed [spr] FFF#none Thursday February 15 13:46:58 EST 2001 PY's ROM has been fixed to the "g" command to enter the debugger works as expected. [ajc] FFF#0988 10:20:00 22 Feb 2001 - Fix fade-range (bounce) bug. AMAN fix for the bounce bug (WR_BOUNCE not being sent to DCCs). You can now bounce a track to itself. [ajc] FFF#???? 10:20:00 22 Feb 2001 - Fix audio distortion All DSP programs reverted to their state as of 16.4.01d, which was the last version which did not appear to exhibit the "distortion" bug. Note: the test tone oscillator and MS decoding are also gone again until I get the newer s/w fixed. Release 16.5.01b, January 10 2001 ====================================================================== Changes Since 16.4.01f ---------------------- [ajc] 15:47:03 21 Dec 2000 Implemented MS Decoding & Test Oscillator in Audioman & MRT Program [ajc] 15:47:03 21 Dec 2000 Changed range of Floating-Point sample values to [-1.0,1.0) {was[-0.5,0.5)} [ajc] 15:53:51 21 Dec 2000 Bug fixes in 21k assembler & linker [ajc] 16:07:54 22 Dec 2000 Implemented STEREOMODE_FOLLOWS_FX flag in aman_cfg. If defined in environment then stereo mapping operations (reverse, monoL, monoR, MS decode) will follow the position of FX processing. This means that if FEED FX is selected the stereo mapping operations will operate on track playback. NB: currently the feed must have both inputs patched to something for this to work. However it is not necessary to arm the feed. This is because the FAME protocol can only send the stereo mode information referenced to an input. To be fixed later... [ajc] 16:11:50 22 Dec 2000 Preliminary work in aman & mixint to send all configuration operations (input patching etc) directly to aman instead of via mdrmain. [ku] 22 Dec 2000 Vivid integration. Most of the code was porting from Andrew Hodgins changes from 15.6 to 15.7. A lot of information can be found in the Release Notes in 15.7 (its actually in Word format on Fridge in documents/mfx3 or thereabouts). For detailed information, refer to the Vivid design documents also on Fridge. Most important part of this port is DMAN_VIVID under PY/TCS/ which replaces the old DMAN_SONY for 9 pin control. Other changes include some to FFS for displaying Vivid title names, and changes to the console so that the connected Vivid can be accessed from the Blue M1/M2 menus. (Good luck!) [mcc] 18:21:10 8 Jan 2001 - ViVid changes The following changes were made as discussed with Ross & Emilijo: The ViVid 9pin device manager is NOT enabled by default in tcs_cfg. To test ViVid you must edit your tcs_cfg file. Read the comments regarding choice of 9pin device managers. Add or remove leading "*" to disable/enable the desired @DEVICE lines. The ViVid dialog (and all ViVid-related code) has been disabled in the new_framerate() function. See also ViVid bug fixes & remedial changes, below Bugs Fixed Since 16.4.01f ------------------------- [spr] FFF#??? - 19:48:32 Dec 2000 - SCSI Driver Fixed non-fatal timing bug. It used to take 6+ seconds to "SCSI_reset()". One second to assert the reset pulse and then due to arithmatic overflow 512 ticks to wait for the SCSI bus to settle. ONE_SECOND is defined to be ((1<<31)|(256*1)). When used as "ONE_SECOND+ONE_SECOND", this would discard bit 31 leaving a result of 512. The bus settling time has been relaxed to be just 1 second now and the reset pulse has been reduced too. This helps mediad find online devices much faster now and fixes a bug in upgrade where devices used to take a very long time to appear. [mcc] FFF#none 20:41:22 6 Jan 2001 - QKernel Memory Mgt. Fixed old bug in QKernel memory mgt. The memory allocation maps were being mismanaged and corruption of memory was occurring. The bug was in QSYS/KERNEL/memalloc.a, function membit(). [mcc] FFF#none 11:06:21 7 Jan 2001 - Duartlib Memory Mgt. Fixed old bug in duart_free() - it used the wrong function to deallocate the memory. Fixed by calling qsdealloc(), not qdealloc(), in DUARTLIB/alloc.c. Luckily, this particular bug had no harmful side effect, but the new tests in KERNEL/memalloc.a detected the fault. The tvt messages "membit NOT SET" or "membit NOT CLR" indicate a fault like this with QKernel memory management. [mcc] FFF#none 18:21:10 8 Jan 2001 - DMAN stack overflow The stack for DMAN_SONY was overflowing. For QKernel processes, the process memory allocation map is corrupted when the stack overflows, resulting in general havoc. Increased stack to 32K for DMAN_SONY and DMAN_VIVID. Also increased the minimum stack size from 4K to 8K in QSYS/KERNEL/fork.a (so no process gets less than 8K stack). [mcc] FFF#none 18:21:10 8 Jan 2001 - Sony QIPICHAIN not unlinked In certain error cases, the Sony device mgrs exit immediately. But they were not unlinking the interrupt handler, causing a fatal exception. Fixed in sonycon_startup() by always calling qvec_unlink() if an error is returned (in DMAN_SONY & DMAN_VIVID). In general, any QKernel process that uses qvec_link() must not exit without calling qvec_unlink(). This could be done automatically if qvec_link() stored the PID in the QIPICHAIN. Then _kf_exit() could scan the lists and remove any entries for the exiting process. [mcc] 18:21:10 8 Jan 2001 - ViVid bug fixes & remedial changes: 1. General remedial changes The VIVLIB library is gone, the MDR interface functions now live in MDRLIB/vivid.c. The header file MFX4/DEFS/ESP/vivid.h has been moved to MFX4/MFX/DEFS/mdrvivid.h. It contains ANSI prototypes for the interface methods and the definition of the shared interface structure: VIVID_DATA. Macros are defined for convenience, e.g., VIVID is a ptr to the shared interface structure. These changes have been made because this interface is for use by MDR processes only, it is not global. A few ViVid defs required by non-MDR processes are defined in MFX4/DEFS/MFX/mfxvivid.h. FFS uses these to build a pseudo-device for browsing ViVid files. The DMAN_VIVID device manager has been ansified. It now compiles clean, free of any warnings. 2. Timecode format confusion ViVid and MDR use different codes to identify the same timecode formats (framerates). There was some confusion between these codes in the software, resulting in the use of a ViVid code to set the MDR framerate (when opening a project with embedded title link). The VFRC type has been defined for ViVid framerate codes, to distinguish them from the MDR codes, type FRC. Conversion functions have been defined: vtomdr_frc() and mdrtov_frc() convert from ViVid to MDR and vice versa. These functions code the following rules: MDR 25fps = ViVid 25 MDR 29ND = ViVid NTSC ND MDR 29DF = ViVid NTSC DF Other MDR formats (24,30DF,30ND) are not supported by ViVid. 3. Buffer overflow by sprintf In MAIN/vivid.c ask_frate_change() the dimension of question[] was far too small, increased to 256. Adjacent memory was being corrupted when the dialog was invoked. 4. Buffer overflow of current_title The dimension of VIV_SH.current_title[] was too small, increased to MAXVIVNAME+1. Adjacent memory was being corrupted. 5. Only send ViVid requests to dman_vivid In TCSMAIN/sci.c vivid_rqst() check for dvc_dmgr == "dman_vivid". 6. Only allocate Q_vivid once The Q_PLAY->Q_vivid structure was being allocated twice by both instances of dman_vivid. Fixed by only allocating if Q_vivid is null in DMAN_VIVID/dman.c. Also clear the allocated memory so it's not full of garbage. 7. Wrong args In DMAN_SONY and DMAN_VIVID, prototypes were broken in timecode.h Once fixed, a bug was detected in timecode.c: in FR_to_timecode() the args for unpkBCDtcode() & cvtFRtoBCD() were wrong due to mis-placed parenthesis. Release 16.4.01f, December 19 2000 ====================================================================== Changes Since 16.4.01e ---------------------- [spr] 12:30:45 19 Dec 2000 Spin up drives in the ROM instead of waiting for "mediad" to spin them up. This means that the drive geometry should always be displayed on the ROM screen correctly now even if it might take a little longer to display all drives if they all need spinning up. [ajc] 15:28:41 19 Dec 2000 Corrected error in FAME protocol rev 8 causing band number of Feed EQ updates to be incorrectly interpreted. Bugs Fixed Since 16.4.01e ------------------------- (none) Release 16.4.01e, December 14 2000 ====================================================================== Changes Since 16.4.01d ---------------------- [djs] FAME mix automation enhancements are implemented. These changes allow timecode to be sent to the mixer regardless of transport speed. [spr] mshell is now the default system wide shell. [mcc] 19:24:10 8 Dec 2000 The no. of meters displayed is derived from the actual number of channels available. Merlin 32 will now have 32 meters. [mcc] 19:24:10 8 Dec 2000 Adjust behaviour of Patch Menu operations to accommodate 32-track configurations better. The excluded tracks are not highlighted. Patch, group and atten softkeys will incur an error if only excluded tracks are selected, or if no inputs are selected for atten. These changes apply in general to a system with less tracks than the current file format. [spr] pyosk works again. It failed to start after making the mshell changes on PX. This is now fixed. [ajc] 11:04:16 11 Dec 2000 - QDCRUN -u option now causes DSP program to be loaded using top half of WRAM. This is useful if another program, eg ramtest, is currently using bottom half. Default bahaviour of qdcrun is now to abort if any DSP program exits with a non-zero exit status. Specifying the -e option reverts to the old behaviour. [ajc] 11:04:16 11 Dec 2000 - QDC/DIAG/RAMTEST -s option causes individual error messages to be suppressed; a summary is printed at the end of the current test pass only. -n disabled all error reports entirely (useful for LA/CRO probing) -d option disables writing to the LED display. [ajc] 11:04:16 11 Dec 2000 - AMAN/CFG Added 32 track mfx & merlin configurations. Changed all Merlin cfgs to use the "n" versions of mdr program which has no EQ processing. [ajc] 11:04:16 11 Dec 2000 - QDC/MDR Added mdr4n program (no EQ processing) for use in Merlin configurations. [ajc] 17:11:19 11 Dec 2000 - AMAN Fixed bug in DSPPROG class: name of mdr program was not being copied from a static buffer used by the parser, so after the cfg file reached a certain size the dsp program names would be overwritten. [spr] 09:41:00 12 Dec 2000 - DISKINIT Diskinit has been modified to format "fat32" media. Usage is the same as for FLFS but instead of supplying a -w option, use -f instead. The irbf as supplied by SimonTech still has a problem within MDR so dont try to record any projects yet. But If you have wav files on a fat32 partition, you should be able to import them into an MDR project. [spr] 10:25:23 12 Dec 2000 - PYOSK The pyosk command now uses mshell as the default shell. The ESP version of SCF is also now used in the pyosk bootblock. This means history is now availible in pyosk. These mods are in preperation for some PY diagnostics that will coming soon. In fact, most commands that work on PX will now work on PY. Including jtag,drt,sysinfo and all of the current PX diagnostic tools. [spr] 01:12:45 12 Dec 2000 - PYOSK Commands can now be executed on pyosk directly from a PX os9 shell. Examples: "pyosk flfscheck /dd" NB: pyosk only has /dd "pyosk jtag -nv" "pyosk "mdir -e;procs -e;irqs >>>/tvt" >>>/nil" Any options after the first non option argument are passed over to the the program to be executed on PY. NB: This setup is NOT multitasking. Only one command can be initiated to run on PY at a time. Another command comming soon "pyfork" will allow concurrect tasks to run. [spr] 02:42:56 12 Dec 2000 - Startup A facility now exists to have a "Quick" startup, with or without networking. When the system is booting a line is displayed that says:- "Press Q [Quick Start] or N [Quick Start with Network]" When this line is displayed pressing Q will continue the startup script but will leave all the time consuming bits out to make startup faster. Pressing N will also start quicker but will also have basic TCP/IP enabled too. (no anetd or noded). This feature is designed for diagnostic use (i.e. cg5hr,sio1test, sio2 and ioc1) and any other diagnostics that do not require a full hardware/jtag scan. To make startup even faster, use the "be login mfx" command and this will automatically login with a shell. A future enhancement of this will be "be login diag", which will place you in the diagnostic executive. NB: The mmu is NOT enabled for a quick start. If you need the speed or the memory protection features of the MMU, type "mmu -ie" at the shell. Both types of startup can also be set by the DIP switches on PXY. DIP switch #B3 selects Quick Start. DIP switch #B4 selects Quick Start with Networking. If the dip swicthes are set it is still possible to override them when the "count down" sequence is being displayed. If you want a normal boot even though the DIP switches are set, just press the space bar. (infact any key other than Q or N). [spr] 09:56:12 13 Dec 2000 - Startup In an attempt to prevent stray keys from activating the quick start and networking setup features during startup, the input queue is now purged before the countdown starts. This means you can now not type ahead to activate one of these features. You can only activate the feature while the count down is active. [mcc] 09:55:28 13 Dec 2000 The .ML file written by the Export/filex softkey or the "filex" command module can optionally include all clips on all layers. If such a file is subsequently imported (opened on Project menu) it will essentially recreate the entire project, with all waveforms "borrowed". To do this, I have defined a new OMDL keyword/parameter for the WSEG block: "layer = n". This specifies the coverage seguence number so that the layers can be preserved. The layer keyword is only included in .ML files, and only when all clips are exported. See also the related "extwave" bug fix, below. Once 15.7 has been modified to understand the "layer" keyword, it too will be able to import these .ML files. This feature is not yet directly supported by the console. To make the Export/filex softkey generate files with all clips, you must "setenv OMDLALL 1" before starting the product. There are ways to make this automatic, contact me if you want that. Alternatively, you can just type the command "filex -a" rather than using the Export menu. [spr] 05:56:51 13 Dec 2000 There is a new target called "release" at the top level of the tree. This target will reconstruct from scratch a release file without having to have a $DEVTARGET online. Just go to the top level and type "make release". All the steps to build a release will be done automatically. This does not clobber the tree, but instead causes the RELEASE image tree to be totally rebuilt. No objects are uploaded to your current $DEVTARGET. Bugs Fixed Since 16.4.01d ------------------------- [mcc] 09:55:28 13 Dec 2000 FFF#none The OMDL "extwave = n" parameter was not being generated for borrowed waveforms. This may not have worried any 3rd-party readers, but it made importing our own .ML files impossible if any waveforms were borrowed from another MFX project file. The extwave parameter is now generated for all external (borrowed) waveforms, in both .ML and .DL files. This bug may have been introduced by recent ProTools-related changes. Release 16.4.01d, December 5 2000 ====================================================================== Changes Since 16.4.01c ---------------------- [dub] FLFSCHECK - fixed problem with shared clusters printout. The FLFSCHECK revision set to 2.04. [dub] FLFSCHECK - Added search for shared clusters between the files. The FLFSCHECK revision set to 2.03. [spr] FAT32/irbf changes. Working from the command line seems fine. However MDR projects still have a problem. This is still under investigation. If possible tests shopuld be done by testing FAT32 media between all windows platforms to see if media can be exchanged. Report any problems to Steve Rance. For detailed explanation of why projects can not be created/opened pls call me. [spr] Before ROMS are blown, the CRC and/or Checksum of the images to be blown are checked. If they fail the programming operation will not proceed. [ajc] Fixed metering problem where not all samples were being shown. [ajc] 11:20:32 1 Dec 2000 Added input level ramping capability to MDR program. Reinstated processing of stereo mappings (monoL/monoR/reverse) by audioman. Implemented clamping of input ramp levels to fix gain error bug. [mcc] 22:29:18 1 Dec 2000 The "sacrificial RAM" parameter introduced into the playtask motion logic in 16.4.01c can be set to N Kb by setenv SACRIFICE N. This may need to be reduced from the default of 198K for 8-tracks per DCC configurations. [mcc] 08:51:19 4 Dec 2000 Added "-r" option to flfscheck to include removed (deleted) files in the "-l" output report. Fixed a buss trap when the device name was not found. [mcc] 19:01:22 4 Dec 2000 Added fckr module to process and report on output from flfscheck. To use, pipe output from flfscheck -l into fckr, e.g.: flfscheck -l /sd00!fckr This will run additional tests for filesystem integrity. A report is printed to stderr if there are any problems. No output means no problem. Bugs Fixed Since 16.4.01c ------------------------- [mcc] 09:45:17 3 Dec 2000 Fixed low-level bug in the tvtput() function - of which there are NINE copies in the system, for some reason! Long messages overran the buffer, corrupting the flags of the next message and thus stalling the tvt output forever. The corruption could extend outside the allocated memory. Release 16.4.01c, November 28 2000 ====================================================================== (This build does not include updated Windows NT executables) Changes Since 16.4.01b ---------------------- [dub] FLFSCHECK - Added additional printout of the segment list per file for -l option [spr] The method by which mediad builds Random Block file managers has been changed. This should have a zero impact. However, it would be good to test HPFS,WFM and AKAI disks to confirm they still work. All other have been tested. [spr] There are no more mfx48, merlin48, fame48 or prodigy32 commands. To start one of these products you should now use one of mfx3, merlin, fame or prodigy with and extra argument. For mfx3 the arguments can be 24 or 48 For merlin the arguments are 24 or 48 For prodigy the arguments are 24 or 32 For fame the arguments are TRKS/LIVES/BUSS where:- TRKS=tracks, LIVES=live feeds and BUSS = the number of buses. Only certain configurations are allowed. To see the configurations available type a "-?" with any of the above commands. NB: THE FAME SOFTWARE RUNNING ON THE PC WILL NOT AUTOMATICALLY CHANGE. YOU MUST CHANGE THE CONFIG FILE ON THE PC TOO OR UNEXPECTED RESULTS WILL OCCUR. NOT ALL CONFIGURATIONS ARE AS YET TESTED OR FULLY FUNCTIONAL. [spr] realigned all text output from /dd/startup after ajc removed 3 dots from the progress messages. These include termmode, sysinfo and setup. [spr] aman now closes paths to configuration files that it opens. [spr] The launch scripts (mfx3,merlin,fame,prodigy) lo longer print out debug information when started (as per request from mcc) [spr] The FAT32 file manager is now in place and working at the shell level. The only known problem at the moment is projects can not be created or loaded from them. This appears to be a dupsys problem in mdrfs. Why it only effects FAT32 I dont know. ....HELP!!!!.... [spr] sysinfo has been enhanced to display non-existant files and file that should display a warning if they exist. (files in /nvr for example) This takes some of the load off the /dd/startup file and centralises all warnings and errors into single executable. [spr] The tvtprintf() "missed messages" message has been re-implemented after it was ommited after porting from the WX/SC combination. [spr] /nvr/.profile has been replaced with /nvr/profile. There was no reason to have this file hidden. [mcc] The efficiency of waveform drawing has been greatly enhanced by exploiting the ability to explicitly set the position with CG5. Instead of a pixel-wise move from the end of one line to the start of the next, it now just "jumps" to the new position. [mcc] Static tvt screen #8 belongs to MDR. Type "tvtline mdr" from a telnet session to see it. The MDR P option activates the output. There isn't very much yet. [mcc] The redundant "NTSC" (29.97) flag has been eliminated from a chain of confusion which led from the console to mdrmain, on through mdrscitask, and finally to tcsmain - where the flag was completely ignored, and has been for years! This is lucky for us, because there was a bug. If you changed the framerate from an NTSC rate to 24 or 25fps, then the NTSC flag remained TRUE. Because TCS ignores the flag, this had no major effect. There was a minor incidence in AudioBase, but only in the choice of temporary audition paste-point. Now this confusing flag is gone. [spr] The product launch scripts (mfx3,fame,merlin,prodigy) have been modified to chack that the correct type of console is being used for the product being launched. [spr] tvtsave has been changed to take an optional argument. This being the name of a file to dump the tvt log into. Without an argument output still goes to stdout. This was done so tvtsave could be used from within the gfxglobal command line. Redirection is not possible from within its command line processing. [mcc] Several changes to improve smoothness of scrolling tracks: For 32 & 48tks, less waveform samples are scanned for display. The "stalled" channel state has been removed from the play task, it served no useful purpose and only resulted in jerky graphics. When transport is locked, mdrwfld does not signal the play task, the periodic signals from mdrint are enough. (This change can be reverted for testing by setenv WFLDNOTRICKS). [mcc] The product launch scripts no longer interrogate the console. Now MDR will not tolerate the wrong console type, and will not run unless a golden dongle is detected. [mcc] There is no "mfx3 32" configuration. [spr] exabyte drviers are now included within the kernel ROM image to facilitate upgrading from tape. Bugs Fixed Since 16.4.01b ------------------------- [djs] FFF#926 "Dropping into Record on a slave machine". This was 2 bugs. The blip at the start of the recording cycle and the slave recording on both passes of autorec instead of just the first. Both these problems are now fixed. [djs] The CG5 oddline bug has been partially fixed by Erik. This bug was observed as the sense of the oddline bit being inverted for the rightmost few pixels on the screen. [mcc] FFF#none. The scaling of track display at zoom 1 was too wide (>12 hrs) in hi-res mode, causing various problems. Fixed by rescaling to 40 secs per pixel. [mcc] FFF#none. The recording waveform display took too long to draw sometimes. Now it starts a maximum of 4 seconds before the cursor, at all zoom scales. [ajc] FFF#none - 23:32:19 16 Nov 2000 Fixed Bug in ODIF operation in audioman. DCC crashed after completion of operation. [ajc] FFF#none - 23:32:19 16 Nov 2000 Fixed Bug in operation of TRIM pot in FAME - only left side of stereo feed was being updated. [mcc] QDC#0949. Y size of clip in 1-track hi-res display caused arithmetic overflow in the display of fades. The fractional precision of a fixed-point data format has been reduced by 1 bit to compensate. PLEASE TEST: create a clip slightly longer than a half-hour. Go to middle of clip, and Fade head and tail to/from that point. You should get 15 minute fades, as that is the maximum fade duration. Now select 16-track display and zoom #16 - inspect start and end points of fades. Do they look okay? Go to head of clip and play all the way through in zoom 16. Is everything okay with the display of fades? [ajc]'s Changes Since 16.3 branch - 21:51:37 13 Nov 2000 ---------------------------------------------------------- [These are all lumped together here as no entries were made in previous release_notes.txt] Audioman has been completely rewritten in C++ and the interface to the DSP programs has undergone some significant changes. Interfaces from other tasks to audioman have for the most part remained unchanged. The major rework was necessary to create a path for expansion of the DSP engine to much larger mixer-based systems and also to allow a much larger degree of configurability than was previously available. Significant changes: Configuration File ------------------ Loading of DSP programs is now controlled by a configuration file which by default is found under the name /dd/USR/SYS/aman_cfg. When audioman starts up it determines the name of the current product configuration. By default this is found in the environment variable PRODUCT_CONFIG; for test purposes the product configuration name can be specified as an argument to the audioman command. Lines in the configuration file are processed initially for all configurations. When a "config" directive is processed following lines in the file are parsed but only processed if the product configuration name matched the configuration directive. Lines starting with "*" are comments. A ";" is a virtual end-of-line which allows several directives to be placed on on e line. The configuration directive specifies one or more configurations with an optional constraint based on the number of QDCs in the system. The specification can contain the usual wildcard characters. The syntax is: { / [+] } A match occurs if matches (with wildcards) the product configuration name, and any of the specs matches the actual number installed. A number on its own means that exactly that number of QDCs must be installed. If a plus sign follows then there can also be more QDCs. A config spec may have several such specifications separated by commas. Examples: config mfx24/1, merlin24/1 ;* mfx24 or merlin24 with exactly one QDC config fame4*/2+ ;* fame4* (eg fame48_48_48, fame42_37_23 etc), ; 2 or more QDCs config prodigy32/2/3 ;* prodigy32 with 2 or 3 QDCs The "program" directive specifies a DSP program or pair of DSP programs and the number of sections (DCCs) which are to be loaded with these programs. The programs for primary and secondary DSPs are separated by a colon. If only one name is specified then both DSPs run the same program. The default directory for DSP programs is /dd/USR/QDC and the default suffix is ".bin". A full path and explicit suffix can be specified. The "set" directive sets an audioman internal variable. Currently the following variables can be set: maxtracks, maxlives, maxbusses, mixenable, debug, amandebug. Syntax: set [=] The "setenv" directive modifies the audioman environment, which may affect the operation of audioman or any program it forks (eg mixint). Syntax: setenv The "stop" directive causes all remaining lines in the file[s] not to be processed. However all lines are still parsed and syntax checked. This directive ensures that another config directive does not unintentionally get "appended" to the desired configuration. The "include" directive inserts lines read from the specified file until end of file is hit, then continues with the current file. includes may be nested to a depth of 8. There is no default path other than the current directory with which audioman was started. If the file cannot be opened for reading then audioman will exit with a fatal error. The file name must be specified without quotation marks or other delimiters. The synonymous directives "sinclude" and "-include" are the same as "include" except that if the specified file is not found the directive is simply ignored. Mixer System (FAME) Signal Processing Architecture -------------------------------------------------- The signal path in mixer-based systems has been changed. Previously a special program (MDX) was run in mixer systems which provided exactly 4 MDR tracks and 4 live feeds and the associated real-time FX processing. This program is now obsolete. The same MDR program is now used in MFX and FAME systems. Real-time processing is provided by a separate program MRT which can be configured to have various numbers of DSP building blocks. Building blocks currently implemented are: - EQ (IIR) with up to 8 x 2nd order bands - Dynamics with compressor and expander/gate - Meters (driven from any TSB signal) - MS Decoder (simple 2x1 mixer) - VCO (oscillator with frequency & amplitude controlled by TSB signals) - Mix Matrix Other planned/proposed/imagined blocks: - Wavetable playback (for clicks etc) - Variable delay-line - Ring Modulator (multiplies two TSB signals) - Non-linear transfer function (table based) - complex FX like reverb etc. A track feed in FAME now consists of an MDR block and an EQ/DYN block which may be placed in either order according to UI control. DSP Resource Descriptors ------------------------ To accomodate the multitude of possible DSP program configurations a system is now in place which allows the DSP program to pass a set of resource descriptors to audioman on startup. Audioman uses this information to build internal objects for each available DSP building block and to decide how to plug them together. The resources for the primary and secondary DSPs of a section are merged together and treated by audioman as a single combined resource pool. This allows for example dynamics processing to be linked amongst all signals paths on both DSPs (which is not possible between separate DCC sections) The MDR, MRT and MIO (I/O-only "idle") programs now use this system; the mixer programs have not been converted yet. Other Changes to Audioman ------------------------- - mixint is now forked by audioman if variable "mixenable" is set to a non-zero value, and is shut down again when audioman exits. Additional parameters to mixint can be specified in the environment variable MIXINT_OPTIONS. - defining DISABLE_DITHER in audioman environment causes dither to be turned off; otherwise triangular pdf dither is applied to all outputs at +/-1 LSB of current word width. Currently known Bugs etc ------------------------ - Stereo invert/mono/swap has not yet been reimplemented - MS decoding has not yet been reimplemented - FAME oscillator has not yet been reimplemented - FAME meters do not return to zero in some cases - patching of stereo inserts causes audio glitch ("double-patching" effect) Bugs fixed recently ------------------- - I/O on mixer DCCs should now work. - incorrect handling of TSBZERO command by MDR program was causing incorrect delay calculation & possible distortion during bounce; now fixed - incorrect handling of ambiguous "txlev" parameter to aman_mixdown was causing the output level pad to be wrong - now fixed. - blue-stop in FAME mode no longer crashes audioman - live feeds were not being correctly allocated to a DSP FX channel - dongle limits for tracks, feeds & busses are now honoured. Other [somewhat-]unrelated changes ---------------------------------- - xcc modified to recognise -qx and -qp options. - new features in asm21k - pxtvtlog "-r" option added to print all messages from the beginning of the buffer - startup modified to start TVT & run jtag earlier; also tee "echo" headings to /tvt (took off a few dots so the tvt messages fit on a 110-wide window!!) [ajc] end Release 16.4.01b, November 10 2000 ====================================================================== (No official build was made of 16.4.01a) Changes Since 16.3.03a ---------------------- [mcc] Version 16.4.01a - 16.3 is now on branch_16_3_03. No official build was made of 16.3.03a. The release notes have been reset and the version advanced in order to make a clean point at which to create branch_16_3_03. [spr] Another library called pxylib.l has been created. Some code was duplicated between the application "power" and the Dynamic ROM. This code has now been moved into the new library and is linked by both. The PSU readings are also now in the system globals area. [spr] The "MFX4:" prompt has been removed and replaced by "SYS:". You would have only seen this if the startup script aborted due to an error or was aborted by the user with cntrl/C/E. The prompt is now "SYS:" so it can be differentiated from the login shell which is "MFX:". [spr] Display the PSU monitoring mode on the Dynamic ROM display screen. [spr] The sysinfo command has been enhanced to inform is the PSU is out of spec and fixes to individual items are now in place. Only Golden machines will be prompted to reblow the Static X and Static Y ROMS therefore preventing users from doing potentially destructive operatrions to there PXY's. The bootblock is also compared to see if it is correct for version currently running. If an operation only needs re-booting (as opposed to a power cycle) this will be displayed. [spr] The noded debug out via the /tvt port has been streamlined. To see nodes going on and off line run "opt k n" from the OS9 shell. All other debug is now on the -z option of noded. [spr] The startup script will now inform the user if the tapcfg file has failed to build correctly. Before it would be silently unnoticed until qdcinit later in the script would abort with "file not found" [spr] The scsi caching system relies upon a "read capacity command" being issued to a device before any disk blocks can be cached. scsicache has no way of inserting scsi commands and snoops commands that are going to the disk. For scsicache to work it must know the blocksize of the device being cached. So an "inq" command has been added after starting scsicache which will force a read capacity command to be issued. The way it was, scsicache would not have been doing anything until mediad was started. - jtag scanning is a little faster now. [spr] sysinfo now displays if there is a missing tapcfg file. This can occur when the jtag scan has had en error and the tapcfg file is missing. [spr] FAME2 Protocol Version 8 is now the default protocol. Preliminary testing looks positive. However, due to my lack of understanding all the features of FAME it has not been extensivly tested. The tree can be compiled in version 7 protocol if needed by adjusting the #define in $MFX4/DEFS/MIXERS/mixer.h. Fame48 does not yet work due to other problems not related to the v8 protocol. [spr] When in Hi-Res, the Dynamic ROM status display is displyed on the right-hand side of the screen. This is only a cosmetic change. [spr] Golden dongle values - The value of "feeds" is the total of tracks + live inputs. For example, a Prodigy32 machine that is licensed to have 32 tracks and 24 live inputs must have tracks = 32 and feeds = 56. Golden dongles now have 96 live feeds and 48 busses as the default. [spr] Dynamic ROM - Mods to display boot order on the setup screen. Also Boot devices that dont need an argument are not displayed with one. The only boot driver that needs an argument is the SCSI driver (the ID to boot to). All other driver dont need one. [spr] Various obsolete UNIX utilities have been removed from the tree. These include modhist,template,codeReview and cdcr. Now that CVS is working nicely these tools are no longer required. (and some people did not use them anyway! :)) [spr] The ESC/Z debug page has been enhanced to display the time of the message. [mcc] The mapping of tracks and live inputs to DSP slots has been extensively modifed. Linked stereo feeds are now always mapped to the same QDC quadrant. Tracks with interleaved stereo clips (stereo WAV files) are handled like implicitly linked feeds. If there is a conflict, the explicit fame linkage takes priority. [spr] It is now posssible to execute tvtsave at anytime to see the current state of the tvt buffers. Also the output of tvtsave now comes out via stdout. If you want it to go to a file redirect its output. This is what the startup file now does. [mcc] The gfxterm module no longer prints spurious "Invalid process id" message. [spr] The Dynmaic ROM startup screen is now really split. The System status is now always in the right hand side of the screen. IOPACK has been reduced back to 85 chars which makes it compatble with the lo-res screen. Anything that is deemed (by me) to be important on the status display is now also shown in flashing red. Bear in mind that for an end user system, nothing should be in red at all! The termcap entries have also been adjusted to suit the new geometry. [spr] rbvccs is now not explicity loaded as it is (and always has been) in the bootblock. There was never any need for it do be in /dd/CMDS/BOOTOBJS as well. [spr] The PX static ROM has been changed to look for a valid Dynamic ROM image if the primary image is corrupt or missing. This means that as long as the emergency Dynamic ROM is blown the system will still boot without having to flip any DIP swicthes. The initail ROM to locate is still determined by the dip switch setting, but if the first choice failes it will advance to the next. If all fail it will automatically fall back to the Static ROM which has to be valid or else the code would not be running. [spr] Non standard system files will now be flagged as a warning when the system starts. This is so we know that these files are in use. [mcc] Bounce and Fade-range are now supported in Fame. [mcc] In old MFX console, Blue-M14 and Blue-M15 can be used to switch the track-key bank, 1-24 or 25-48. [spr] New product launch scripts. The only files that have changed are mfx3, merlin, fame, and prodigy. These are backwards compatible with the script of the same name before this change. However, now you can type "merlin 48" to start merlin in 48 track mode. With all of these commands, entering with a "-?" will give a help sheet. Not all of the fame/prodigy configurations are complete. Also the default shell for the system is now mshell. If you notice anything different between the old and the new please call me ASAP. The old sctipts "merlin48", "mfx48" still exists until such time the new scripts are accepted by all. Bugs Fixed Since 16.3.03a ------------------------- [spr] FFF#none The GFX window manager could start writing to the video RAM before IOPACK has finished scrolling. This has been fixed. Now the window manager will wait for all IOPACK activity to stop before continuing. [mcc] FFF#??? Fixed old bug in Reverse waveform feature that caused misalignment of output waveform under certain conditions. [spr] FFF#none Fixed pystart not to re-enable MMU if it was not enabled in the first place. Before if the MMU was manually disabled, starting pyosk or qserver resulted in the MMU being re-enabled. [spr] FFF#none Fixed a startup problem in pyosk when trying to use the GPLATCH facility of intglue. A bug in the system still exists that is clearing intglue/inten6/gplatch. I will have to get a logic analyser on this to find the bit of code that is clearing this bit. [spr] FFF#none Iniz /tvt and /tvtcntl in the startup script so the tvt driver (tvtd) does not call its terminate routine. Before this mod, if the terminate routine was called it would clear the INTGLUE/INTEN6/GPLATCH bit which would cause certain messages comming via the GPLATCH not to be displayed. In fact this is what what caused pyosk not to start. This fixes all known problems with pyosk and means comments in the previous paragraph are now not true. [spr] FFF#none Added misssing rbuccs from kernel rom image. Without this it was not possible to mount or upgrade a system if any RBF disks were online. Whoops. [spr] FFF#none The "tvtsave" program now works as expected. Before it was possible that messages at the end of the tvtlog.txt file were actually from a much earlier session and nothing to do with the last session. This occasionally also caused a vast ammount of messages to be saved into the tvtlog.txt file that should not have been writtem at all. [spr] QDC#0929 GPO's fixed when tcs_cfg invert flag was enabled. [spr] FFF#none Minor, display only Y2K bug fixed in pxtvtmsg. [mcc] FFF#none Various problems with Bounce have been fixed. These were new bugs, introduced in 16.4. Bounce now seems to work correctly in both mfx and fame. [mcc] FFF#none A fame configuration problem has been fixed - live feeds on non-MDR QDC sections are now correctly detected. No Release 16.3.03a, October 26 2000 ====================================================================== No official build was made of 16.3.03a. Changes Since 16.3.02f ---------------------- [mcc] Version 16.3.03a Separate 24 & 48-track builds are no longer necessary or meaningful. All the product startup scripts are supported by this one release. Note that fame48 and prodigy32 will still not work correctly, so don't bother trying them yet. [mcc] Added -u option to "update" script to cause unconditional update of all differing files. [mcc] New file - gfxconfirm.h. You no longer need to include the entire GFX planet in order to see a prototype for gfx_confirm(). Just #include . [ajh] Port of HFSFM and Protools import/export from rev 15. [ajh] Searches for nvr network setup keywords now case insensitive. [ajh] FAT16 2GB support [spr] The "less" and "setime" commands now work within a gfxterm. [spr] Termcap entries corrected for gfxterms. [spr] gfxterms now have a template device descriptor called dgt000x000_000 [spr] The Dynamic ROM is no longer built under Unix. This is because it is nearly 4MB and 99% of the files in it are duplicated elsewhere in the release. Now the Dynamic ROM image is built when the system first boots after an upgrade. This takes about 30 seconds to build. The size of the release file before this change wss 8.6MB. It is now down to 7MB after compression. [mcc] The mixer-controlled track parameters are no longer part of the track descriptor. Tracks have a "runtime extension" (TREX) to hold these and any future data that are not to be saved/loaded as part of the project file. [spr] The nominal power supply for the +-12 rail is now 12.6 volts. (verbal from Jon Keeble). The error range has been increased to +-5% to take into account various reading errors. The current rane of monitoring values are: Nominal 05.00v Upper:05.20v Lower:04.80v Nominal 12.60v Upper:13.10v Lower:12.09v Nominal 12.60v Upper:13.10v Lower:12.09v Nominal 03.30v Upper:03.43v Lower:03.16v Anything outside of these ranges will be considered an error. [spr] The "power" command has been modified to display on the /tvt display if the power rail deviate from there specifcation. Once this has been proved to be correct a sysmesg() should be invoked instead. This message should never really appear unless there is a problem. To see the power rails in real time issue the command "opt k p" and look at the TVT output. To turn off use the command "opt k -p". This entire facility is under the control of the NVRAM which has been modified to turn this feature on by default. If you are running a version 3 PXY then you will have to turn this option OFF. (rc from the ROM) [spr] The upgrade command now only starts mediad once to find availible devices to install onto. Before with some drives that took a long time to appear, you could be waiting for a long time. This should impove matters. [spr] The Dynamic ROM now prints Narrow SCSI ports in green. It is not an error to have them narrow so now they are displayed like everything else that is OK. Bugs Fixed Since 16.3.02f ------------------------- [spr] The command line utility "bdd" had a long standing bug of not setting up the address of the device correctly. This has now been fixed. [ajh] Partitioned volumes (any filesystem) are now recognised as /pXXX rather than /cXXX. [ajh] Netman and Anet(OSK+NT) now support Setfork setstat operation. [ajh] NT Noded thread now started well after config file has been processed. Fixes occasional error "No networks to scan" and failing medialink transfers. [ajh] If local node is involved in a backup and it is PRIVATE, it MUST perform the backup. Fixes the error "Module Not Found" seen during copy. [ajh] Fix diskgopher fatal if removeable ejected during caching [dub] FFF#914 - Fixed problem with import of files created on little endian machines. Also ANSI-fied readomf.c, and corrected modification to OMF 2.02 toolkit source headers done by JY in order to avoid redefinition of enum MEDIADATA. [mcc] FFF#none - Fixed bug in parsing of sony_id file introduced 16.2.02e. The symptom was the message "fairlight id not found in id file". Release 16.2.02f & 16.3.02-48f, October 20 2000 ====================================================================== Changes Since 16.3.02e ---------------------- [djs] Up-to-date dongle files have been taken from the ~ian directory and the ~ian/edit script has been updated to re-commit the dongle file into cvs. [mcc] The gfxlaunch window in which the mfxstart script runs will hang around until you press a key if the script exits with an error. [spr] The PX Dynamic ROM has been modified to detect and setup a new machine ID if a removable dongle has been inserted into the parallel printer port socket on IOC1. If this plug in dongle is in use, it will override the backplane ID. [spr] The "adc" command can now be compiled and is included in distribution. This involved changing the intglue.[dh] files so the "adc" sources could be compiled. [spr] The PX Dynamic ROM now displays the low voltage powers rails. If your power supply is out of range then these numbers will be displayed in Red. Will only work correctly on rev4+ PXY's. Michael Carlos still has revision 3 and therefore will display garbage. [spr] The PX Dynamic ROM now does not have any type ahead. Before we had a 1 char type ahead which caused users (and me) much grief when trying to work quickly. Now, we have no type ahead. This makes things run a little smoother. [spr] When NVRAM is erased, corrupted or a version mismatch has occured, the entire contents of NVRAM must be re-entered. Before it was possible to only enter either the System or booting preferances. This could leave one still erased or in a corrupt state. [spr] The /dd/ISP/ETC/Makefile has been removed. To rebuild the networking modules now, you should enter "setup" at the OS9 command prompt, make any required changes (or may be no changes) and then exit the editor. When the editor exits the networking modules will be rebuilt. This file is also removed if it exists at boot up. [spr] an unused directory was permanatly removed from CVS. This was $MFX4/LIBS/YGL/examples. All code is this directory was also in EXAMPLES2 which still exists. [spr] The ROMS display of online disks now correctly displays disks that are greater than 100GB. Before the number was correct but out of allignment. [spr] IOPACK now uses the version number of the currently loaded Xilinx to determine if it should run in HiRes or LoRes. This should prevent future backwards compatiblity issues. [mcc] The Merlin LCD/Volume setup menu is accessible when no project is open. [mcc] The EXTFILE node & unit fields are defunct. The version of MDR files is now maintained in the EXTFILE. [spr] Modifed /dd/startup so /tvt device is not used before drivers are installed. [spr] The "features" file can now regenerate the data.h, rand.h and plugin.h files. Template records have also been added to the top of the data.h file so a simple cut and past/modify cycle can be used. This regeneration facility can only be used on "GOLDEN" machines for security reasons. [mcc] Wrapped struct _embed around embedded object fields in project HEADER. All references to these fields are via macros EMBED_ENDOBJ and EMBED_ODFSD. [spr] The "be" command can now take an argument of "normal" which will delete the /nvr/session file. [spr] Any arguments given to the "be" command can now be passed over to the shell script that overseas an application launch. The only example of this so far is the "be login" command. It is now possible to type "be login mfx". This will launch the login program with the arguments "mfx". The end result of this is the same as "be shell" but may be used in the future as run time options for the fame and mfx launch scripts. [djs] Tracks 25-48 can now be armed when the system is a 9-pin slave. The Sony ID code has changed from FAA0 to FAA2, because other devices use this information to determine how many tracks the slave has. There are still problems arming tracks from an MFX3+, but two QDC Engine machines work together nicely. [spr] Detect emergency Dynamic ROM image. This is only checked on Golden machines. This means that when machines are made in production, a warning will be given informing them to blow the emergency backup Dyanmic ROM. Once the machine leaves it will no longer be Golden and subsequent releases will NOT inform the user that the Emergency backup ROM is out of date. The end result is that if an accident happens programming the Primary Dynamic ROM, support could tell them to put DIP Switch #B1 in the ON position and reboot. This should get the user booting to the disk from where they can complete the programming job. It does not matter the the emergency backup ROM is not always upto date. It is there purly as a method to boot the system. This method will also come into place soon with the Static ROMS. As even though the ROMS are NOT exeactly the same build, the have enough code to boot the system and jump over to the Dynamic ROM. The only forseable reason why the PX Static ROM will have to be reblown is if the CG5 Xulinx changes. If this is the case then all machines, golden or not will be informed that the CG5 Xilinx is out of date and the PX ROM has to be reprogrammed. The end result of this is that the only ROMS the user should ever have to ultimatly blow is the Dynamic ROM (for which their will hopefully be a backup ready incase of a catastrphie [sic]) and hence the reason for this modification. [spr] blowall now has a -e option to blow the Emergency Dynanic ROM. [spr] added all of the plugin dongle numbers into the database files. Production from the next release onwards should now not have to the the dor file (Dongle OverRide) and get in a panic when the network is down for whatever reason. [spr] Removed these rather large files CPUGLUE.svf FPGALD.svf xglue.svf yglue.svf. These are not required in the release file yet as there is no way of reprogramming these Xilinxes from PXY yet!!. The size of each of these files is just under 1MB and therefore will reduce these size of the release file considerably. Bugs Fixed Since 16.3.02e ------------------------- [mcc] FFF#none Fix bug introduced in last release. Reading embedded objects from attached file with different format (no. of tracks) than the current project would read garbage. Now it incurs an error with the message "File Format Not Compatible - filename.MT". [mcc] FFF#none Fix potential buss trap when Pack file with embedded objects. [dub] FFF#none Fix the bug in FLFS section of DISKINIT which would allow allocation of the file past the physical disk space. Revision number for diskinit set to 1.29. [spr] FFF#none Fix BLUE ZOOM cursor position calculation. This was a cosmetic bug left over from the hires port. [dub] FFF#none Command line utility "badcopy" added to SRC/MFX4/APP/OS9. It is to be used for recovery of the files that reside over the damaged disk space which otherwise couldn't be copied or backed up by any other means. It is recomended to copy such files to another physical disk. [dub] FFF#none Corrected help sheet for diskinit that indicated -b option as not available for initialization of FLFS disks. Scan for bad blocks is available on FLFS partitions. [mcc] FFF#none Fix memory mgt bug introduced in 16.3.02d. It is now safe to use AudioBase File operations again. [cea] FFF#none Fixed Command table in diagnostic application sio1. [spr] FFF#none re-enable MMU when PY is only restarted. [spr] FFF#nome fixed bug in missing fpsp040 in PYOSK's merged bootblock. Now jtag will work under PYOSK. This will ultimatly be used for programming XGLUE from PY and YGLUE from PX. It has no effect on the MFX application. [spr] FFF#none When IOPACK is restarted after the GFX window manager shuts down the IOAPCK cursor colour was not reset to white. This is now fixed. For some reason it did does not appear in 16.2 probably due to the pallette changes between lores and hires. [dub] FFF#none DISKINIT - Writing revision number of the DISKINIT into spare bytest of the FLFS boot sector. Revision of diskinit set to 1.30. All partitions formatted prior to diskinit 1.30 will have diskinit ID as 0.00. FLFSCHECK - Verifying the revision of DISKINIT that formatted the partition if option -v is specified. If discrepancy is detected, warning is printed. Example 1: Disk formatted with diskinit revision lower than 1.30 MFX:flfscheck /sd00 -v Partition initialized with DISKINIT Rev 0.00 ************************************************************************************ * Warning: This disk should be re-initialized with DISKINIT version 1.30 or higher * * Make sure that all existing files on this disk are backed up before * * disk is initialized. * ************************************************************************************ Example 2: Disk formatted with DISKINIT Rev 1.30 or higher MFX:flfscheck /sd04 -v Partition initialized with DISKINIT Rev 1.30 Parition initialization info is up to date Release 16.3.02e & 16.3.02-48e, October 8 2000 ====================================================================== Changes Since 16.3.02d ---------------------- [djs] Sample-based run up/down is now operational. Spurious (and incorrect) tvt output has been removed. [djs] TCS environment variables can now be placed in NVRAM, using the file /nvr/tcs_cfg. This uses the same entries as the normal tcs_cfg file, but the advantage is that it is not replaced when installing trees, so site-specific settings are retained. Note that only SETENV lines are valid in the NVRAM file. The application will read the NVRAM file (if it exists) after the regular one. [mcc] The buggy Microware rename command has been replaced with a copy of esprename. That is, esprename and rename are now exactly the same program. Should it ever be required for some reason, the original Microware command is now called os9rename. [mcc] Changes to both consoles to diagnose comms problems at system startup: Watchdog not patted in 1KHz interrupt so console will reset if multitasking locks up. Improved escaping and checksumming of packets for transmission, does less with ints disabled. Fixed inconsistency in checksum calculation, now (8-bit-sum AND $BF) at all 4 rx/tx points. If fatal serial error occurs the consoles will now report FATAL ERROR instead of just making noise and/or rebooting. These are the fatal errors that may be reported: INVALID TERM PKT # - the id is invalid or unsupported CHECKSUM WRONG - serial data may have been corrupted FAULTY PACKET - protocol error, software bug or data corruption Rx422 STATUS 0x - receiver overrun, parity error, queue full [djs] A new diagnostic program, adc, is now available. It measures the power supply rails to the PXY. [mcc] The number of tracks available for playback and recording is limited to the dongle value. For example, if you are only licensed to have 24 tracks, then you may use mfx48, and open 48-track projects, but only the first 24 tracks will be available for playback and recording. The number of fame live feeds is not yet limited in this way. Bugs Fixed Since 16.3.02d ------------------------- [djs] FFF#820 "Merlin Console freezes if multiple RW or FF key presses made with 9 pin online" [mcc] FFF#none Fix bug that occasionally made mfx erroneously appear to be "busy", thus inhibiting backup, etc. Ensure mfxstatus is cleared when mdrmain exits. [mcc] FFF#none Fix bug that occasionally resulted in a burst of garbarge at the end of an exported stereo WAV file. A bug in WPX/ODIF infrequently left the last buffer unflushed. It seems this only affected interleaved-stereo output. MFX3 has the same bug. [mcc] FFF#none Fix garbled error handling by WAV import. It would sometimes produce no error message at all, or "Operation Terminated" depending on the actual error. Made WAV import/export tvt output less verbose. [mcc] FFF#none Fix garbage in exported WAV file headers. Some files I created contained the string "(c)1996 Microware Corporation". Neat bug, huh? [mcc] FFF#none Fix hires file window bug - it had one too many lines. [mcc] FFF#none Fix effect of CG5 autoincrement-readback bug on track display SYNC field, by changing no-clock colour to white text on red background.