[PATCH 2/3] mf: Partially implement sample copier transform.

[Francois Gouget]

On Mon, 9 Mar 2020, Nikolay Sivov wrote:

> That's strange. I don't immediately see what could trigger loader critsection
> deadlock at this point, half way through tests. Other run with different
> locale works fine too. I don't know how to use these results productively, as
> it never crashed like that for me. Crashes we have on tests page are also
> different and related to some queue shutdown issue most likely.

It may not be much comfort but I've seen this type of critical section 
timeout in WineTest runs before:

https://bugs.winehq.org/show_bug.cgi?id=48112


For instance more recently in this debiant 32 bit French run:

https://testbot.winehq.org/JobDetails.pl?Key=66562#k102

023c:err:ntdll:RtlpWaitForCriticalSection section 0x7e77ad00 "../../../wine/dlls/user32/user_main.c: user_section" wait timed out in thread 023c, blocked by 0279, retrying (60 sec)
0279:err:ntdll:RtlpWaitForCriticalSection section 0x7bd2d1e0 "../../../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0279, blocked by 023c, retrying (60 sec)
023c:err:ntdll:RtlpWaitForCriticalSection section 0x7e77ad00 "../../../wine/dlls/user32/user_main.c: user_section" wait timed out in thread 023c, blocked by 0279, retrying (60 sec)
0279:err:ntdll:RtlpWaitForCriticalSection section 0x7bd2d1e0 "../../../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0279, blocked by 023c, retrying (60 sec)

These look very much like the timeouts that happened for your patch.
I don't think this is related to the locale. It seems rather random:

Timed out?                 (deb10)(deb10)
        66562  66468  66386  66270  66145
win32     n      n      n      n      n
 fr_FR    Y      n      Y      n      Y
 ja_JP    Y      Y      Y      n      n
 zh_CN    n      n      n      Y      n
wow32     Y      n      n      n      n
wow64     n      n      n      n      n

The same errors also happen on cw-gtx560 so they're not the result of 
some QEmu bug.

These critical section timeouts are also hard to notice in the WineTest 
runs (e.g. on test.winehq.org) since they don't prevent WineTest from 
completing; and the additional timeouts, if any, are lost in the middle 
of the other failures. The only way to really notice them is to check 
the Wine traces.

If anyone figures out what's up with those that would be great.


-- 
Francois Gouget <fgouget at codeweavers.com>