Thread tests failure on Windows

[Paul Vriens]

Hi,

I've not been able to get the kernel thread tests working on windows. The
only thing I've found is that changing thread.c:

/* simulate a call to set_test_val(10) */
#if 0
    stack = (int *)ctx.Esp;
    stack[-1] = 10;
    stack[-2] = ctx.Eip;
    ctx.Esp -= 2 * sizeof(int *);
    ctx.Eip = (DWORD)set_test_val;
#endif

or even

#if 0
    ctx.Eip = (DWORD)set_test_val;
#endif

makes the test run instead of crashing. They still fail, but only for the
last ok().

Unfortunatly this is way over my head. So I'd appreciate if someone could
have a look.

What worries me (maybe needless) is that the test runs OK on Wine. So
either we're way better or we implement things differently.

Cheers,

Paul.

[Eric Pouech]

Paul Vriens wrote:
> Hi,
>
> I've not been able to get the kernel thread tests working on windows. The
> only thing I've found is that changing thread.c:
> /* simulate a call to set_test_val(10) */
>
> #if 0
>     stack = (int *)ctx.Esp;
>     stack[-1] = 10;
>     stack[-2] = ctx.Eip;
>     ctx.Esp -= 2 * sizeof(int *);
>     ctx.Eip = (DWORD)set_test_val;
> #endif
>
> or even
>
> #if 0
>     ctx.Eip = (DWORD)set_test_val;
> #endif
>
> makes the test run instead of crashing. They still fail, but only for the
> last ok().
>
> Unfortunatly this is way over my head. So I'd appreciate if someone could
> have a look.
>
> What worries me (maybe needless) is that the test runs OK on Wine. So
> either we're way better or we implement things differently.
>
> Cheers,
>
> Paul.

is it still the case with latest CVS ?
A+

[Saulius Krasuckas]

* On Thu, 17 Nov 2005, Eric Pouech wrote:
> - Paul Vriens wrote:
> > I've not been able to get the kernel thread tests working on windows. 
> > The only thing I've found is that changing thread.c:
> > /* simulate a call to set_test_val(10) */
> >
> > #if 0
> >     stack = (int *)ctx.Esp;
> >     stack[-1] = 10;
> >     stack[-2] = ctx.Eip;
> >     ctx.Esp -= 2 * sizeof(int *);
> >     ctx.Eip = (DWORD)set_test_val;
> > #endif
> >
> > makes the test run instead of crashing.
> >
> > What worries me (maybe needless) is that the test runs OK on Wine. So
> > either we're way better or we implement things differently.
>
> is it still the case with latest CVS ?

you mean with latest Winetest-*.exe package?  yes it is still present. [1] 
if you look at the "kernel32:thread" line you'll notice all OSes crashing.  
strange enough is the fact, that some win98s don't crash. [2]  maybe that 
are those which runs under Qemu emulator? (w98.ie50.dx61 and w98.ie6.dx81)

[1] http://test.winehq.org/data/200511171000/
[2] http://test.winehq.org/data/200511161000/#Win98

[Paul Vriens]

> - On Thu, 17 Nov 2005, Eric Pouech wrote:
> > - Paul Vriens wrote:
> > > I've not been able to get the kernel thread tests working on windows.
> > > The only thing I've found is that changing thread.c:
> > > /* simulate a call to set_test_val(10) */
> > >
> > > #if 0
> > >     stack = (int *)ctx.Esp;
> > >     stack[-1] = 10;
> > >     stack[-2] = ctx.Eip;
> > >     ctx.Esp -= 2 * sizeof(int *);
> > >     ctx.Eip = (DWORD)set_test_val;
> > > #endif
> > >
> > > makes the test run instead of crashing.
> > >
> > > What worries me (maybe needless) is that the test runs OK on Wine. So
> > > either we're way better or we implement things differently.
> >
> > is it still the case with latest CVS ?
>
> you mean with latest Winetest-*.exe package?  yes it is still present. [1]
> if you look at the "kernel32:thread" line you'll notice all OSes crashing.
> strange enough is the fact, that some win98s don't crash. [2]  maybe that
> are those which runs under Qemu emulator? (w98.ie50.dx61 and w98.ie6.dx81)
>
> [1] http://test.winehq.org/data/200511171000/
> [2] http://test.winehq.org/data/200511161000/#Win98

I first though that maybe it has to do with running the tests on VMware
(most of my test results on test.winehq.org are from VMware images), but
my WinXP box at home also fails.

And as Saulius already mentioned, the current output on test.winehq.org is
with latest CVS (including your fix for the set_test_val routine).

Paul.

[Detlef Riekenberg]

Am Freitag, den 18.11.2005, 02:51 +0200 schrieb Saulius Krasuckas:
> strange enough is the fact, that some win98s don't crash. [2]  maybe that 
> are those which runs under Qemu emulator? (w98.ie50.dx61 and w98.ie6.dx81)

qemu is correct.

[Paul Vriens]

On Thu, 2005-11-17 at 13:51 +0100, Paul Vriens wrote:
> Hi,
>
> I've not been able to get the kernel thread tests working on windows. The
> only thing I've found is that changing thread.c:
> /* simulate a call to set_test_val(10) */
>
> #if 0
>     stack = (int *)ctx.Esp;
>     stack[-1] = 10;
>     stack[-2] = ctx.Eip;
>     ctx.Esp -= 2 * sizeof(int *);
>     ctx.Eip = (DWORD)set_test_val;
> #endif
>
> or even
>
> #if 0
>     ctx.Eip = (DWORD)set_test_val;
> #endif
>
> makes the test run instead of crashing. They still fail, but only for the
> last ok().
>
> Unfortunatly this is way over my head. So I'd appreciate if someone could
> have a look.
>
> What worries me (maybe needless) is that the test runs OK on Wine. So
> either we're way better or we implement things differently.

http://cvs.winehq.org/patch.py?id=21337

so we were better after all :-)

Paul.