#775 describes a bug arising from a change in Microsoft's DirectSound implementation in Vista and later that requires use of a larger input buffer size (hence increased latency). Clients using PA/DirectSound with input buffer sizes that are below some threshold (too small) would fail on Vista and later prior to the fix (i.e. in v19.7 and earlier, and on master prior to March 25, 2025). See #775 for details.
#1008 merged an emergency fix for issue #775 to be included in v19.8. The fix puts a large lower bound on the buffer size and also the output buffer size in the full-duplex case.
#1008 was cherry picked from #782 where you can find further information.
The #1008 fix was "emergency" in the sense that we also increased the output buffer size for full duplex streams even though this is not strictly necessary to fix the problem. The "optimal" fix would allow continuing to use a small output buffer size, thus achieving lower latency. We chose not pursue the "optimal" fix because it would require larger code changes and no one has time for that before the 19.8 release.
This ticket is place to discuss the "optimal" fix. Should it be prioritised? What is involved? How much work is it?
#775 describes a bug arising from a change in Microsoft's DirectSound implementation in Vista and later that requires use of a larger input buffer size (hence increased latency). Clients using PA/DirectSound with input buffer sizes that are below some threshold (too small) would fail on Vista and later prior to the fix (i.e. in v19.7 and earlier, and on master prior to March 25, 2025). See #775 for details.
#1008 merged an emergency fix for issue #775 to be included in v19.8. The fix puts a large lower bound on the buffer size and also the output buffer size in the full-duplex case.
#1008 was cherry picked from #782 where you can find further information.
The #1008 fix was "emergency" in the sense that we also increased the output buffer size for full duplex streams even though this is not strictly necessary to fix the problem. The "optimal" fix would allow continuing to use a small output buffer size, thus achieving lower latency. We chose not pursue the "optimal" fix because it would require larger code changes and no one has time for that before the 19.8 release.
This ticket is place to discuss the "optimal" fix. Should it be prioritised? What is involved? How much work is it?