Hi,
Il 07/05/21 12:14, Marvin ha scritto:
=== w7u_adm (32 bit report) ===
mfplat: 0a28:mfplat: unhandled exception c0000005 at 6D5D7D1C
=== w7u_el (32 bit report) ===
mfplat: 0a50:mfplat: unhandled exception c0000005 at 6D2B7D24
I tried to debug this. Unfortunately I am not able to reproduce it on any Win 7 VM I tried. Querying the testbot many times, I could trim down the crashing test to a minimal example that seems to be a completely correct program, but still crashes on the testbot (with probability around 0.5). So I currently believe that:
* As noticed by François, the problem already existed before my patch set, it was just less probable: https://test.winehq.org/data/patterns.html#mfplat:mfplat. So my patch seems to make it more evident, but the problem was already there.
* The crash remains also if I set the rate to 1.0, which is the default. It goes away if I don't set the rate at all. It also goes away if I request less than 10 frames (which means that I don't reach the end of the file). These facts further convince me that the problem is in the Win 7 implementation of MF, not in our test.
In the end, I think that my patches are correct and should be accepted. If that's considered useful, I can disable the SetRate test on Win 7 to avoid triggering a lot of crashes, but please tell me what is the correct way to do that, because I know that just querying the Windows version is not good.
Thanks, Giovanni.