Re: user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Paul Vriens]

Dmitry Timoshkov wrote:
> dlls/user32/message.c   |    2 +
>  dlls/user32/tests/msg.c |   54 +++++++++++++++++++++++++++++++++++++++++++++++
>  server/queue.c          |   10 +++++++-
>  3 files changed, 65 insertions(+), 1 deletions(-)

Hi Dmitry,

These new tests introduce some failures on Win9x. Could you have a look?
> -
>            ok(ret && msg.hwnd == hwnd && msg.message == WM_USER+1 &&
> -
>               msg.wParam == 0x1234 && msg.lParam == 0x5678,
> -
>               "got ret %d hwnd %p msg %04x wParam %08lx lParam %08lx instead of TRUE/0/WM_USER/0x1234/0x5678\n",
> -
>               ret, msg.hwnd, msg.message, msg.wParam, msg.lParam);

This ok() message is also not totally correct as the msg.hwnd should not 
be 0 but hwnd, I guess.

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Austin English]

On Fri, Jun 12, 2009 at 12:32 PM, Paul Vriens wrote:
> Dmitry Timoshkov wrote:
> > dlls/user32/message.c   |    2 +
> >  dlls/user32/tests/msg.c |   54
> > +++++++++++++++++++++++++++++++++++++++++++++++
> >  server/queue.c          |   10 +++++++-
> >  3 files changed, 65 insertions(+), 1 deletions(-)
>
> Hi Dmitry,
>
> These new tests introduce some failures on Win9x. Could you have a look?

It also broke the gdi32:font, gdiplus:font, and user32:edit tests for me, see:
http://test.winehq.org/data/45629dc050f46988c23ded35ad18bbd8c63e4e65/wine_ae...

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Dmitry Timoshkov]

"Austin English"  wrote:
> > These new tests introduce some failures on Win9x. Could you have a look?
>
> It also broke the gdi32:font, gdiplus:font, and user32:edit tests for me, see:
> http://test.winehq.org/data/45629dc050f46988c23ded35ad18bbd8c63e4e65/wine_ae...

These tests shouldn't be affected.

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Dmitry Timoshkov]

"Paul Vriens"  wrote:
> These new tests introduce some failures on Win9x. Could you have a look?

It's because HWND_MESSAGE support is missing in win9x.

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Paul Vriens]

Dmitry Timoshkov wrote:
> "Paul Vriens"  wrote:
> > These new tests introduce some failures on Win9x. Could you have a look?
>
> It's because HWND_MESSAGE support is missing in win9x.

OK, will have a shot at this.

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Dmitry Timoshkov]

"Paul Vriens"  wrote:
> > > These new tests introduce some failures on Win9x. Could you have a look?
> >
> > It's because HWND_MESSAGE support is missing in win9x.
>
> OK, will have a shot at this.

Perhaps PeekMessage() sets a reasonable last error code in that case
which you can test for, and 'continue;'.

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Paul Vriens]

Dmitry Timoshkov wrote:
> "Paul Vriens"  wrote:
> > > > These new tests introduce some failures on Win9x. Could you have a 
> > > > look?
> > >
> > > It's because HWND_MESSAGE support is missing in win9x.
> >
> > OK, will have a shot at this.
>
> Perhaps PeekMessage() sets a reasonable last error code in that case
> which you can test for, and 'continue;'.

Well actually, I've improved/corrected the ok() messages and it shows:

msg.c:11637: Test failed: 0: got ret 1 hwnd 00000224 msg 0171 wParam 
00001234 lParam 00005678 instead of TRUE/00000224/WM_USER+1/0x1234/0x5678
msg.c:11628: Test failed: 2: hwnd 00000001 expected 1, got 0
msg.c:11632: Test failed: 2: got ret 0 hwnd ABABABAB msg abababab wParam 
abababab lParam abababab instead of TRUE/0/WM_USER+2/0x5678/0x1234

So it's HWND_TOP that has the wrong message 0171 vs. WM_USER+1
and
HWND_BOTTOM is the one that fails.

I will probably sent 3 patches if you agree of course):

1. Improve/correct the ok() message
2. Add a broken() for HWND_TOP and the wrong message
3. Change "if (data[i].ret)" into "if (ret)"

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Dmitry Timoshkov]

"Paul Vriens"  wrote:
> > Perhaps PeekMessage() sets a reasonable last error code in that case
> > which you can test for, and 'continue;'.
>
> Well actually, I've improved/corrected the ok() messages and it shows:
>
> msg.c:11637: Test failed: 0: got ret 1 hwnd 00000224 msg 0171 wParam 
> 00001234 lParam 00005678 instead of TRUE/00000224/WM_USER+1/0x1234/0x5678
> msg.c:11628: Test failed: 2: hwnd 00000001 expected 1, got 0
> msg.c:11632: Test failed: 2: got ret 0 hwnd ABABABAB msg abababab wParam 
> abababab lParam abababab instead of TRUE/0/WM_USER+2/0x5678/0x1234
>
> So it's HWND_TOP that has the wrong message 0171 vs. WM_USER+1
> and
> HWND_BOTTOM is the one that fails.

Looks like win9x behaviour in this test is completely broken, it's better
IMO to detect win9x and skip the test.

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Paul Vriens]

Dmitry Timoshkov wrote:
> "Paul Vriens"  wrote:
> > > Perhaps PeekMessage() sets a reasonable last error code in that case
> > > which you can test for, and 'continue;'.
> >
> > Well actually, I've improved/corrected the ok() messages and it shows:
> >
> > msg.c:11637: Test failed: 0: got ret 1 hwnd 00000224 msg 0171 wParam 
> > 00001234 lParam 00005678 instead of TRUE/00000224/WM_USER+1/0x1234/0x5678
> > msg.c:11628: Test failed: 2: hwnd 00000001 expected 1, got 0
> > msg.c:11632: Test failed: 2: got ret 0 hwnd ABABABAB msg abababab 
> > wParam abababab lParam abababab instead of TRUE/0/WM_USER+2/0x5678/0x1234
> >
> > So it's HWND_TOP that has the wrong message 0171 vs. WM_USER+1
> > and
> > HWND_BOTTOM is the one that fails.
>
> Looks like win9x behaviour in this test is completely broken, it's better
> IMO to detect win9x and skip the test.

I'm not too much into this, so I take your word for it.

There is nothing (apart from the failures of course) that indicates 
Win9x (last error doesn't change).

We could of course use CreateWindowExW instead of A ?

user32: Add a test for PeekMessage((HWND)-1), make it pass under Wine.

[Dmitry Timoshkov]

"Paul Vriens"  wrote:
> There is nothing (apart from the failures of course) that indicates 
> Win9x (last error doesn't change).
>
> We could of course use CreateWindowExW instead of A ?

That's an easiest solution, yes.