--
v3: regedit: Directly use RegQueryValueExW() instead of a helper function.
regedit: Correctly calculate REG_MULTI_SZ size before reading from the buffer.
regedit: Remove unneeded valueDataLen override for DWORD values.
https://gitlab.winehq.org/wine/wine/-/merge_requests/392
I'll have a look at the time failures, I was seeing them locally but not on the testbot run so I assumed they were flukes caused by some decoder differences, looks like it may be something else after all.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/393#note_3477
On Wed Jul 6 17:52:44 2022 +0000, **** wrote:
> Zebediah Figura replied on the mailing list:
> ```
> On 7/6/22 05:08, R��mi Bernon (@rbernon) wrote:
> >> So yes, we could not block in IWMReader::Stop(), but that's not
> >> ultimately going to help anything. We do still need to wait for the
> >> streaming thread to stop at some point.
> >
> > Well we could then call `Stop` and wait for the WMT_STOPPED callback
> notification with a condition variable, releasing the filter_cs so the
> callbacks are free to enter it to make the thread-safety obviously correct.
> Unfortunately we can't release the filter CS while waiting for the
> thread to stop. We need to protect the whole operation.
> ```
I opened https://gitlab.winehq.org/wine/wine/-/merge_requests/393#note_3473 to make WM async reader state transitions asynchronous, so that we can then do the right thing here and wait for the callback notifications while entering the CS and avoiding the deadlock.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/140#note_3475