On Sat, 2014-01-18 at 20:14 +1100, James Bunton wrote: > Hi Tanu, > > Thanks for the quick review! :) > > My mobile phone (Nokia E7 running Symbian Belle Refresh) is connecting > to pulseaudio/bluez running on my Arch Linux on my Raspberry Pi. Without > this patch every few minutes pulseaudio will disconnect the stream and > log "SBC decoding error (-3)". As you can imagine this is quite > annoying! :) Yes, I can believe that :) Please add this information to the commit message too. The commit message should always answer the question "why is this needed". Here's some more wisdom about commit messages: http://who-t.blogspot.de/2009/12/on-commit-messages.html > I see what you mean about the timing code being wrong. What does > pulseaudio use this for? Should I expect to hear some audio glitches > with the current incorrect code, or is it just for display in a GUI? It's used for latency reporting. I don't think the latency information is used much when recording, it's more important for playback (lip-sync), but of course we still should strive for correct values in the latency reports. I don't expect that you would hear glitches even if the latency information is wrong. > I've modified it to use the number of bytes that sbc_decode reports it > wrote to the output buffer. Both versions seem to work equally well for > me. The read_index update is still not quite in the right place. sbc_decode() is called in a loop, and if it fails at any iteration, then read_index should not be updated. So, the update should happen after the loop. > You can see them here: > Repo URL: http://delx.net.au/git/pulseaudio > Patch1: > http://delx.net.au/git/pulseaudio/commitdiff/f66bec680d5a78778ef44910807020724ba4ebae > Patch2: > http://delx.net.au/git/pulseaudio/commitdiff/ba134901487a30633beafe335a5670a6586a65d4 > > I can submit these as patch files with git-send-email if that's > preferable for review. I prefer git send-email. > Thanks again :) Thanks for the patches! -- Tanu
