Alsa 1.0.0rc2 OSS Midi is still broken for recording from external MIDI devices. The time values of note on and note off events are severely corrupted in MIDI recordings that contain a sequence of chords of two or more different notes played at the same time. However, interestingly, MIDI recordings which contain only a sequence of _single_ notes work ok because the bug was partially fixed in alsa-driver-0.9.1 released around 15th March 2003.
Only OSS-mode MIDI recording from /dev/sequencer is broken, e.g. using an OSS client like the old Rosegarden 2.1pl4 (http://www.cannam.demon.co.uk/rosegarden/distribution/rosegarden-2.1pl4.tar.gz or ftp://ftp.cs.bath.ac.uk/pub/dream/ROSE/rosegarden-2.1-sources.tar.gz) whereas ALSA-mode MIDI recording works fine, e.g. using the new Rosegarden 4 (http://heanet.dl.sourceforge.net/sourceforge/rosegarden/rosegarden-4-0.9.5.tar.bz2) To reproduce this bug, you'd need a MIDI keyboard for sending external MIDI events to your soundcard and also an OSS client for MIDI recording which must be the _old_ version 2.1pl4 of Rosegarden (newer versions of Rosegarden do not do OSS-mode MIDI recording), preferably my statically linked Rosegarden 2.1pl4: http://www.liam2.demon.co.uk/rosegarden.static.bz2 (sources http://www.liam2.demon.co.uk/rosegarden.tar.bz2) I made three OSS MIDI recordings to show how to trigger the bug: Attachment #1: I repeatedly played the ascending notes of the scale of C-major (or whatever name your locale gives to a 440Hz pitch) one note at a time at a speed of roughly 1 note/second for about 12 seconds; this recording plays correctly from beginning to end. Attachment #2: I played the notes C and E together repeatedly at a regular speed of roughly 2 chords/second for about 20 seconds; observe the very irregular, incorrect timing of the recorded notes. Attachment #3: I repeatedly played the notes C, E and G together at a regular speed of roughly 2 chords/second for about 20 seconds; notice again the very irregular, incorrect timing of the notes and also how the MIDI file has been stretched from 20 seconds of true playing time to over 3 minutes in playback length. It is actually the same old bug I reported with full details in my email as bug #1 ("Every MIDI file recording ... has invalid time ...") with the subject "[Alsa-devel] Invalid MIDI recordings and silent external playback" on 26th August 2002. Please ignore the other bugs mentioned in that email such as silent external MIDI playback because these have been fixed since alsa-driver-0.9.1 (external playback is ok using "playmidi -e -D 0 file.mid"). Subject: [Alsa-devel] Invalid MIDI recordings and silent external playback URL: hhttp://www.geocrawler.com/mail/thread.php3?subject=%5BAlsa-devel%5D+Invalid+MIDI+recordings+and+silent+external+playback&list=12349 Additional explanations in two followups on 15th November 2002 with Subject: [Alsa-devel] OSS-mode invalid MIDI recordings and silent external playback http://www.geocrawler.com/mail/thread.php3?subject=%5BAlsa-devel%5D+OSS-mode+invalid+MIDI+recordings+and+silent+external+playback&list=12349 Subject: [Alsa-devel] Re: OSS-mode invalid MIDI recordings and silent external playback Maybe this bug is easy-to-fix so it won't be a showstopper for a 1.0.0 release but it's been around a very long time, at least since ALSA-0.3.0pre4 in 1999. I don't know the ALSA driver source well enough to be able to find it. Thanks, Will
oss1-nobug.mid
Description: MIDI audio
oss2-buggy.mid
Description: MIDI audio
oss3-verybuggy.mid
Description: MIDI audio
