parallel mode broken in Linux

[loganmc10]

In Windows, selecting the "Parallel" option works (mupen64plus). In Linux, the application freezes and must be force closed.

[ata4]

Hm, that one might be difficult to debug for me... I'll see what I can do about it.

[loganmc10]

I got around to bisecting this, it looks like the bad commit is eef6466 (but actually I tested 5cbd52e, since it's needed to compile, but I assume the problem is in eef6466 )

I had to remove the "build" folder in between testing commits or else my bisecting would get messed up (it would work when it shouldn't etc...)

Anyway so basically d1cb99f is the last commit that worked for me

[Jj0YzL5nvJ]

In my case the application freezes is random in the start, I can't reproduce it with 100% certain.
But I've noticed that it's easier to get all working when firefox is closed.

[ata4]

As far as I can tell, the renderer itself is fine, since the retracer runs stable and fast with parallel rendering enabled on my Ubuntu VM. I'll need to check the M64P plugin interface next.

[ata4]

Oh, never mind my previous reply, I forgot to run the retracer with the -p flag. It was actually a deadlock issue during the initialization in parallel.cpp, because the constructor didn't wait for the threads to reach a mutex before running tasks. It should be fixed now.

[Jj0YzL5nvJ]

I can still reproduce this problem, 1/30 attempts. Is very random.

[ata4]

Haven't been able to reproduce so far. But I'll reopen it in case anyone else knows more.

[loganmc10]

I haven't had any issues since you closed this

[ata4]

I have found two more (silly) mistakes that could have been responsible for the crashes and some abnormalities caused by d52d94f.

[loganmc10]

Can this be closed then? Seems stable to me

[ata4]

Yeah, I hope it is now.

[Jj0YzL5nvJ]

Finally, I can no longer reproduce this issue.