phosit/0ad
1
0
Fork 0
forked from 0ad/0ad
Code Pull Requests Activity

work around sound-related deadlock and assert on exit (due to OS X's broken OpenAL implementation)

Browse Source 
refs #297

This was SVN commit r7516.
This commit is contained in:
janwas 2010-05-08 19:11:28 +00:00
parent b821f7a289
commit e2fd560983
1 changed files with 19 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
source/lib/res/sound/snd_mgr.cpp
View File

@ -1692,19 +1692,26 @@ static LibError vsrc_reclaim(VSrc* vs)
// clear the source's buffer queue (necessary because buffers cannot
// be deleted at shutdown while still attached to a source).
// note: OpenAL 1.1 says all buffers become "processed" when the
// source is stopped (so vsrc_deque_finished_bufs ought to have the
// desired effect), but that isn't the case on some Linux
// implementations (OpenALsoft and PulseAudio with on-board NVidia).
// wiping out the entire queue by attaching the null buffer is safer,
// but still doesn't cause versions of OpenALsoft older than 2009-08-11
// to correctly reset AL_BUFFERS_PROCESSED. in "Re: [Openal-devel]
// Questionable "invalid value" from alSourceUnqueueBuffers", the
// developer recommended working around this bug by rewinding the
// source instead of merely issuing alSourceStop.
// reference: http://trac.wildfiregames.com/ticket/297
alSourceRewindv(1, &vs->al_src); // stops the source
AL_CHECK;
vs->loop = false;
vsrc_latch(vs); // jw: from Heron's patch
// OpenAL 1.1 says stopping the source causes all buffers to
// become "processed":
alSourceStop(vs->al_src);
// .. so this should de-queue them all.
vsrc_deque_finished_bufs(vs);
// unfortunately that isn't the case on some Linux implementations
// (OpenALsoft and PulseAudio with on-board NVidia).
// wiping out the queue by attaching the null buffer is safer and
// avoids an error in alDeleteBuffers during shutdown on OS X:
alSourcei(vs->al_src, AL_BUFFER, AL_NONE);
// even that isn't enough for versions of OpenALsoft older than
// 2009-08-11, which don't reset AL_BUFFERS_PROCESSED as they should.
// in "Re: [Openal-devel] Questionable "invalid value" from
// alSourceUnqueueBuffers", the developer recommended working around
// this bug by additionally rewinding the source.
// reference: http://trac.wildfiregames.com/ticket/297
alSourceRewind(vs->al_src);
AL_CHECK;
al_src_free(vs->al_src);