back-port windows / yield fix ?

[Michael Stahl]

On 16.10.2015 20:50, Michael Meeks wrote:
> Hi Michael,
> 
>         I just hit a nasty windows deadlock; and noticed that it's fixed
> on master with:
> 
> commit 482c52e91fe41a52e68827e9bf64a9736427d517
> Author: Michael Stahl <mstahl at redhat.com>
> Date:   Fri Jun 26 13:01:51 2015 +0200
> 
>     vcl: fix Win32 deadlocks from SolarMutexReleaser
>     
>         Turned out to be very reproducible ;-) and your fix above (which
> looks sensible) fixes it.

you can reproduce that with the UI? that's amazing...

>         Any reason not to have that in -5-0 ?

well it's a little scary, let me try to remember the follow-up fixes...

- the assumption that the thread that uses SolarMutexReleaser actually
holds the SolarMutex turned out not to hold

commit c18f11587d37f285a95447dd8996c8b605732e00 "tdf#92460: fpicker: fix
VistaFilePicker using SolarMutexReleaser ... on a thread that doesn't
ever lock SolarMutex."

commit 4231274abaec8e45538788bd67261044d7ef4bea "tdf#76845: vcl: main
thread does not necessarily have SolarMutex... at least not if
VCLXWindowImpl::OnProcessCallbacks drops it and calls arbitrary event
handlers. "

- the LazyDelete related data race still needs fixing, as documented in
commit 815fe9110897d5d8c122a5743e3765b67013dd8b