On Fri, 24 Jun 2022, Kacper rączy wrote:
This doesn’t seem right, my changes didn’t touch anything file or threadpool related. This patch introduced completely new API.
Only the errors listed in this email are attributed to your patch. So you can ignore the file failures.
[...]
=== w1064v1507 (32 bit report) ===
ntdll: threadpool.c:273: Test failed: expected info.userdata = 0x10000, got 0 threadpool.c:278: Test failed: RtlDeregisterWait failed with status 103
I have no record of these two failures ever happening in the WineTest runs but it makes sense that they are related to bug 53130. Also I managed to reproduce them with and without your patch so it's as if they are more likely to happen outside of WineTest (maybe exclusively on Windows 10 1507?):
https://testbot.winehq.org/JobDetails.pl?Key=117566#k201 https://testbot.winehq.org/JobDetails.pl?Key=117569#k201
Anyway you're in the clear. And I added them to the list of known failures since they are so rare and thus likely to cause false positives.
=== w10pro64 (testbot log) ===
WineRunTask.pl:error: The previous 1 run(s) terminated abnormally
The full log is actually: | TestBot process got stuck or died unexpectedly | WineRunTask.pl:error: The previous 1 run(s) terminated abnormally
The first line indicates that the server-side process running the task got stuck. So that's not caused by your patch.
Maybe this should be handled differently (downgraded to a warning?) but there is stuff that I need to figure out first: * Did the server-side process get stuck or did it die? Or neither? A better trace would be nice. * It may instead be an inconsistency between the Task status and the VM status. That may suggest there is a race condition somewhere. * This task has no start timestamp which is wrong (it has happened before but is rare). Again this may indicate a race condition of some sort.