'Twas brillig, and Maarten Bosmans at 13/01/11 10:17 did gyre and gimble: > The change I would like to see reviewed, before merging are the read, > write => pa_read, pa_write changes in fdsem.c > https://github.com/mkbosmans/pulseaudio/commit/a02aad4471a07d862cdafc11a06b46c8b54aff55#diff-2 > This is necessary, because on windows the code in pipe.c makes a pipe > out of two sockets instead of file descriptors. That means that a bare > read/write won't work, because it expects an fd. pa_read/pa_write > provides a nice wrapper that uses recv/send in case of Windows > sockets. But as pa_read/pa_write also does a loop to cover up EINTR, > I'm not sure whether the code is still correct on Linux. (Limited > testing of this code on Linux reveals no problems though) I think this commit generally is the one I'm most nervous of (simply because I'll freely admit I don't know the full repercussions). Lennart, if you happen to be stuck somewhere rather than making it to Brisbane, can you comment on this change specifically. The main issue is that this change introduces a cover up of EINTR as Maarten mentions above. This could cause infinite loops somewhere and the change in threaded-mainloop.c may even make debugging client applications a little strange. I'm not really sure, so your (or other suitably enlightened beings) opinions would be welcome. If you want to you can see his full set of changes here: https://github.com/mkbosmans/pulseaudio/compare/mingw32-build although I don't think there is anything particularly controversial in any of the other changes. Col -- Colin Guthrie gmane(at)colin.guthr.ie http://colin.guthr.ie/ Day Job: Tribalogic Limited [http://www.tribalogic.net/] Open Source: Mageia Contributor [http://www.mageia.org/] PulseAudio Hacker [http://www.pulseaudio.org/] Trac Hacker [http://trac.edgewall.org/]
