Release Checklist Watching Brief
From Audacity Wiki
Any items marked $ are believed fixed, awaiting testing.
- Clicks during recording reported by James and Gale (the latter might have been USB trasmission problems and had major dropouts as well) and Gerald Capodieci (all on Windows). Might be sample overwrites. See .
- Force quit Audacity to test crash recovery ability, and if an unforced crash occurs try to note down at once the steps that preceded it, and note if recovery does not work. Problems are already known where crashing with two tracks open leads to multiple projects being recovered (essential); and where if a track is imported then another added after the autosave interval, recovery fails (priority aim to). Complete recovery failures (empty projects) have also been seen on single tracks which are only generated tones.
- Test Chains extensively because we are unsure what problems there may be and report them in this section. Check especially for effects where requested parameters are not implemented in the exported file, or not correctly remembered after restarting Audacity. Known issues:
- Noise Removal parameters in Edit Chains can't be used to capture a noise profile, even in CleanSpeech mode
- (reported as causing a crash on) Mac: LL and BP find that if you zoom in either before starting record or during it, the first assertion in Envelope::GetValues() may fail if the display scrolls. See .
- MJS fixed a reason for some assertions occurring: when two envelopes were merged, and have no points, the overall length of the envelope was not getting updated.
- RA: I'm reasonably sure that putting an envelope point well beyond the start / end of the envelope can cause writes to memory outside the object, i.e. stack corruption. I fear all this means that the causes of the assertions need to be looked at and resolved, to ensure that when the envelope length should change, it does. Maybe the clip class needs some some more code adding so that it's impossible to change the length of the clip without the envelope being updated (I'm thinking hooks in the set functions that change the rate and / or number of samples, driving similar changes through to the envelope.) This might make it much harder to not update the envelope, and so the code rather simpler.
- Assertions corrected on 24 April 08 to test that mTrackLen is the length of the envelope and the underlying track.
- Mac: Mac OS X crash on recovering files. Users have reported two issues 1) a freeze on Intel Macs (2) a non-crash situation on Intel and PPC where you recover, receive a "junk after document" warning and are presented with empty tracks. See and download an example auto save file and _data folder. [BP] tried on CVS (Intel Mac) and 1.3.3-beta on a ppc and could not recreate.
- Mac: open a few 1.2.x projects to see if a crash occurs, but believed no longer an issue
- Mac: Force a crash, recover, then start and cancel any process and items in Audacity menu are unavailable. LL but can still reproduce this on Intel Mac and PPC on OS X 10.4. See: . Hardly anyone else can reproduce it, and there is no obvious fix. If it remains isolated on particular machines then it will remain as it now is, merely release noted.
- Watch out for Undo or other actions e.g. removing a Label Track causing a crash in clean or unsaved projects which have not crashed (reported by Darren on Linux). See Linux-specific issues. Gale has tested these issues on Win XP and completely failed to reproduce. This may be a Linux only problem.
- Keep an eye on Help > Audio Device Info for devices that wrongly report "0" channels, any behaviour in Audacity that is affected by this and any changes that occur in Audio Device Info when switching devices. Release versions of 1.3.4 have an issue where the slider in Advanced Mixing Options jams when exporting WAV/AIFF if the selected output device is reported as having "0" output channels. CVS fixes the jam by treating "0" channels as "2" but this does not fix the misreporting on a device that otherwise seems to work. GA found in 1.3.4 release that switching to an output device reporting "2" output channels freed the slider, and the slider remained free when switching back to the device previously reporting "0" channels, because it then reported "2" channels correctly.
- All toolbar buttons completely unresponsive and have no effect on one Windows XP machine but not the other (reported by Gale - happened twice, does not necessarily resolve after reboot and no abnormal CPU or memory usage evident when it happens). Also test for sticking Control Toolbar buttons on Mac (often reported in 1.2.5/6 but usually better in 1.3.x).
- Some reports on Windows that after running Find Clipping, Audacity is always slow or liable to freeze thereafter, even after restarting the computer or after reinstalling 1.3.4. One comment suggested that deleting audacity.cfg fixed the problem. Should be tested on a 10 minute or longer track.
- Where user has a multiple real playback and recording devices, not possible to scroll down through complete list in Device Toolbar using keyboard. Reported by GA on Windows XP where scroll won't go beyond Modem line#0. Could be related to Audacity problems when encountering devices with no associated mixer.
- Nyquist generation plug-ins don't give the correct time generation. David S on Win 98 finds a 3 second pluck generates 8.5 seconds, for GA on XP it generates 2.756 seconds and GA finds all Nyquist generation is shorter than it's supposed to be by about 8.14%
- Windows: Watch for speaker and mic backgrounds in Device and Mixer Toolbar being black or wrong colour, especially on Windows 2000 and 98. Thought to be a wxBitmap/bit depth issue, not a theming issue. May be an older hardware/timing specific bug when things get cleared and repainted.
- Envelopes have been a major source of problems in the past. In particular, look for when you split a track that contains envelope edits, there may be no end points on a 'split', leading to envelope changes on a 'split'. GA is finding this problem intermittently on Windows XP SP2. MJS can only reliably reproduce it by zooming into sample level, put in an envelope point and then a split within 1.5 samples to the right (about 34 microseconds). It could be looked at after 1.4. Also keep an eye open for error cases that have not yet been covered, and report them. If problems are no worse than in 1.2.5. we may use as is, but we may want to release note them so that people know what to expect.
- Windows: Intermittent failure to detect playback and recording clipping noted on one Windows XP SP2 machine, currently no longer reproducible (GA to report if recurs)
- Windows:Intermittent failure to snap shifted tracks or dragged selections on one Windows XP SP2 machine (this is a consistent problem, it works only about every other time Audacity is launched)
- Keep an eye out for autosave files that open incorrectly for example due to mismatched tags. Autosave files themselves don't have or need a matching </project> tag because Audacity adds this when saving the project, and because the </project> tag would have to be removed and added back at the bottom when the autosave file was added to.
- Keep eye out for abnormal silence added when exporting selection as MP3 - see
- Unduly slow MP3 export compared to 1.3.2. Confirmed by GA on an older 1.6 GHz 1 GB RAM Win XP single machine, MJS did not see the problem on a faster dual core machine, nor GA on a faster XP machine. Thread and user complaints about it .
- Unduly slow processing of AutoDuck - GA finds on one Windows XP machine that processing consistently takes 16 - 20 times the length of the audio.