Why did hid:device start timing out on May 24th?

[Francois Gouget]

The failure pattern shows that hid:device started timing out about 20% 
of the time on the debiant2 VM on or about the 24th of May:

https://test.winehq.org/data/patterns.html#hid:device

debiant2 was last modified on 2021-05-11 so that's not it.

There is no obviously related commit on that date. I tried bisecting for 
a commit by considering a commit good only if the test succeeds 25 times 
in a row. But that was inconclusive.

Similarly I looked for another test interfering with hid:device (with 
the same repeat count), but that was inconclusive too.

So does anyone have any idea what could have changed?
Or better yet, have a fix for it?

Note that the timeout seems to only impact the tests running in a 
non-English locale or a WoW Wine build. In other words anything that's 
not the plain 32-bit English Wine. That's strange too.

[Zebediah Figura (she/her)]

On 6/20/21 3:03 AM, Francois Gouget wrote:
> The failure pattern shows that hid:device started timing out about 20%
> of the time on the debiant2 VM on or about the 24th of May:
>
> https://test.winehq.org/data/patterns.html#hid:device
>
> debiant2 was last modified on 2021-05-11 so that's not it.
>
> There is no obviously related commit on that date. I tried bisecting for
> a commit by considering a commit good only if the test succeeds 25 times
> in a row. But that was inconclusive.
>
> Similarly I looked for another test interfering with hid:device (with
> the same repeat count), but that was inconclusive too.
>
> So does anyone have any idea what could have changed?
> Or better yet, have a fix for it?
>
> Note that the timeout seems to only impact the tests running in a
> non-English locale or a WoW Wine build. In other words anything that's
> not the plain 32-bit English Wine. That's strange too.

My guess is it's bug 51277. I have a fix attached here.

[Francois Gouget]

On Sun, 20 Jun 2021, Zebediah Figura (she/her) wrote:
[...]
> My guess is it's bug 51277. I have a fix attached here.

I had trouble reproducing the failure since it's random and really only 
happens in the TestBot VMs (and there was a bug in the script to 
reproduce it too).

But now it's all sorted out the patch does seem to fix the issue:
* Without the patch I could reproduce the hang after 5, 14 and 10 runs.
* With the patch I got 75 successful runs in a row.
* And after reverting the patch I again had hangs after 3, 1 and 10 
  runs.