|
|
Wiki > Main > MetadataOnBufferTesting (compare)
|
Difference: MetadataOnBufferTesting (r16 vs. r15)Metadata-on-Buffer testingIntroductionNicolasPennequin's MetadataOnBuffer SummerOfCode2007 project to revise the buffering side of playback has now produced a largely-working build. His code is hosted in git, and a snapshot of the latest is available from http://repo.or.cz/w/Rockbox.git?a=shortlog;h=mob Before this is committed, it needs lots of testing. SWCODEC-using volunteers are wanted. If you can find a reproducible bug that doesn't happen with SVN Rockbox (or an intermittent bug that doesn't seem to happen with SVN) then please add recipes/details below. Reproducible issuesFor this recipe, I'm using two albums A (19 tracks) and B (10 tracks). All tracks are MP3. My WPS has track X of Y and next track title included.
Expected result: Next track info is displayed as Album B Track 1. Playback continues onto Album B Track 1. Actual result: No next track info is displayed. Album A Track 19 repeats once (displayed as track 1 of 10!). Playback then moves onto Album B Track 1 (displayed as track 2 of 10). Last tested with commit: 23 Oct 2007 - "Makes some variables const and some others volatile" Status: Reproducible on H300 target. White noise, crashes and other badness after enabling crossfade during playback
Expected result: Playback continues. Actual result: Occasionally a burst of white noise is played, this can vary in length but might be as long as 2 seconds. (GodEater reports that this crashes his iPod Video instead.) Last tested with commit: 23 Oct 2007 - "Makes some variables const and some others volatile" Status:Reproducible on H300 target. Should be fixed with commit "Attempt to fix crossfading" (I think the attempt worked), please confirm. Note that playback seems to restart and then stop again after changing the crossfade setting. That might not be related to MoB though. Intermittent issuesFixed issuesWe'll keep these around for regression testing... Reshuffle playlist doesn't work
Expected result: The playlist is reshuffled, any displayed next track name/info is updated (perhaps after a short delay). Actual result: The reshuffle takes place (as can be seen by viewing it), but all buffered tracks are played in the original order, and next track info is not updated. Last tested with commit: 22 Oct 2007 - "Close partially buffered handles when changing tracks " Status: Fixed. Crash when skipping backwards via playlist
Expected result: At each press of SELECT, the previous track should be played (perhaps after a short delay for buffering). Actual result: After several repetitions of step 4, the player locks hard. Last tested with commit: 22 Oct 2007 - "Close partially buffered handles when changing tracks " Status: Fixed. Hard crash with long tracks buffered
Expected result: Browsing is possible while playback continues. Actual result: The device locks up before returning to the browser. Last tested with commit: 22 Oct 2007 - "Close partially buffered handles when changing tracks " Status: Fixed. Progressbar loses sync in the last seconds of a track
Expected result: The progress bar runs smoothly to the end of each track. Actual result: The progress bar resets to 0:00 a couple of seconds before the end of the track. It is synced with the codec, not with the WPS. Last tested with commit: 22 Oct 2007 - "Close partially buffered handles when changing tracks " Status: Fixed. The problem is that while the track is finishing to be played (PCM buffer emptying), the playback code has already moved to decoding the next track. To update the progressbar at this time, Buffer is not refilled if user spins up disk
Expected result: The buffer has refilled. Actual result: No refill is triggered. Last Status: tested with commit: 24 Oct 2007 - "Fix Fixed in commit "Ask for new files when the bug where handles would remain open after a USB connection" disk is spinning and useful data is below the high watermark" For this recipe, I'm using two albums A (19 tracks) and B (10 tracks). All tracks are MP3. My WPS has track X of Y and next track title included.
Expected result: Next track info is displayed as Album B Track 1. Playback continues onto Album B Track 1. Actual result: No next track info is displayed. Album A Track 19 repeats once (displayed as track 1 of 10!). Playback then moves onto Album B Track 1 (displayed as track 2 of 10). Status: Fixed in commit "Revert to calling playlist_next() in check_new_track() as in SVN." Related linksr18 - 26 Oct 2007 - 08:48:29 - SteveBavin
Revision r16 - 24 Oct 2007 - 23:28 - NicolasPennequinRevision r15 - 24 Oct 2007 - 16:05 - NicolasPennequin Copyright © by the contributing authors.
|