Re: [PATCH 4/4] openal32: Add thunks for EFX functions

[Henri Verbeet]

I suppose 32-bit reads are atomic on x86, provided the variable is
aligned correctly, but shouldn't access to CurrentCtx technically be
synchronized?

[PATCH 4/4] openal32: Add thunks for EFX functions

[Chris Robinson]

On Tuesday 01 September 2009 1:59:54 am Henri Verbeet wrote:
> I suppose 32-bit reads are atomic on x86, provided the variable is
> aligned correctly, but shouldn't access to CurrentCtx technically be
> synchronized?

Probably, but I'm not sure it would be beneficial. If the thread can be 
interrupted between entering one of the extended functions and the function 
pointer being called, it would be a pretty nasty race condition. OpenAL 
doesn't really say what happens when you call a function for a context that's 
not active (current Windows implementations, if they haven't changed too much, 
would likely crash, silently fail, or succeed using the wrong device).

I'm not sure the cost of an extra lock would be worth it when it could still 
randomly crash anyway.