[PATCH] winmm_test.exe fix for CALLBACK_THREAD

[Stéphan Kochen]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello again,

"wine winmm_test.exe wave" reliably locks up for me after calling
waveOutOpen() with the CALLBACK_THREAD flag, waiting for a WOM_OPEN
message to arrive. Which never happens because I believe the thread
doesn't have a message queue.

This patch implements the solution mentioned in the "remarks" section of
the MSDN page on PostThreadMessage:
http://msdn2.microsoft.com/en-us/library/ms644946.aspx

Fixed it for me.

- -- Stéphan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFGERkUcFUq0gzqDwQRAlwCAKD1rVdMXjawPzJOddfU5Oh9hK8cagCgzB2g
6b6JbG23fjGqk+Mj9xvHnF8=
=rXu0
-----END PGP SIGNATURE-----

[Stefan Dösinger]

Am Montag 02 April 2007 16:54 schrieb Stéphan Kochen:
> Hello again,
>
> "wine winmm_test.exe wave" reliably locks up for me after calling
> waveOutOpen() with the CALLBACK_THREAD flag, waiting for a WOM_OPEN
> message to arrive. Which never happens because I believe the thread
> doesn't have a message queue.
>
> This patch implements the solution mentioned in the "remarks" section of
> the MSDN page on PostThreadMessage:
> http://msdn2.microsoft.com/en-us/library/ms644946.aspx
>
> Fixed it for me.

Are you sure about modifying a test? Does the test work on windows, or does it 
fail there?

If it works on windows then the test is ok, and wine needs to be fixed 
elsewhere.

[Eric Pouech]

Stefan Dösinger a écrit :
> Am Montag 02 April 2007 16:54 schrieb Stéphan Kochen:
> > Hello again,
> >
> > "wine winmm_test.exe wave" reliably locks up for me after calling
> > waveOutOpen() with the CALLBACK_THREAD flag, waiting for a WOM_OPEN
> > message to arrive. Which never happens because I believe the thread
> > doesn't have a message queue.
> >
> > This patch implements the solution mentioned in the "remarks" section of
> > the MSDN page on PostThreadMessage:
> > http://msdn2.microsoft.com/en-us/library/ms644946.aspx
> >
> > Fixed it for me.
>
> Are you sure about modifying a test? Does the test work on windows, or does it 
> fail there?
>
> If it works on windows then the test is ok, and wine needs to be fixed 
> elsewhere.

the patch sounds fine as the description of the issue is correctly 
covered by the patch
it solves a race, which of course can have different manifestations 
depending on your workload, #cpus...
A+