Ryan C. Gordon comments

Results 1356 comments of


                                            Ryan C. Gordon

there a error compile android arm in mac

Oh, I wonder if he's using SDL headers configured for his development machine, and then trying to build android libraries after that, so it ends up in a bad state...?

Data race in pulse audio initialization

This appears to be a race condition in pulseaudio, not SDL, where it's reusing a socket for multiple contexts but not protecting it. I'm hesitant to change this in SDL;...

Data race in pulse audio initialization

> we can't just slap a mutex around it, because the thread blocks inside pa_mainloop_run for its whole lifetime. So this is the code in PulseAudio itself for `pa_mainloop_run` ......

Data race in pulse audio initialization

Of course that second parameter to pa_mainloop_iterate tells it to block until something happens, and all our code relies on this, so we _can't_ just slap a mutex around it...and...

Data race in pulse audio initialization

PulseAudio has something called a Threaded Main Loop which is probably a more correct solution here, but it's going to take some redesign to incorporate it, so this might not...

Data race in pulse audio initialization

So #7734 is kinda a big change for SDL2 when no one is having any real-world problems, so I have it sitting in a PR for SDL3 right now. I...

Data race in pulse audio initialization

Threaded mainloop is in SDL3 now; do we want to cherry-pick this back to SDL2?

Data race in pulse audio initialization

I'm bumping this out of the 2.28.0 milestone, it's shipping too soon to put this large a change in now. I'll put it in 2.30.0, or it might end up...

Data race in pulse audio initialization

The threaded mainloop code is in for SDL2 (and has been in SDL3 for awhile now). There is still one pending ThreadSanitizer warning, but I'm not confident it's a problem...

Data race in pulse audio initialization

The ThreadSanitizer warning? No, not yet.