Re: kernel32/tests: Convert fiber back to thread even when IsThreadAFiber is not available.

[Nikolay Sivov]

On 06.08.2016 15:40, Bernhard Übelacker wrote:
> Testbot shows the fiber test failing for Windows 2003 recently.
> https://test.winehq.org/data/tests/kernel32:fiber.html
>
> This is probably related to the last change in
> http://source.winehq.org/git/wine.git/commit/556fef3dd184e71e1f0345bd8f37259...
>
> As Windows 2003 seems to have no IsThreadAFiber the test never
> calls test_ConvertFiberToThread.
> This leads probably sometimes to the crash now visible.
>
> By calling test_ConvertFiberToThread before return the crash disappears
> on Windows 2003.
>
> Signed-off-by: Bernhard Übelacker <bernhardu(a)mailbox.org>
> ---
>  dlls/kernel32/tests/fiber.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/dlls/kernel32/tests/fiber.c b/dlls/kernel32/tests/fiber.c
> index 9ff9366..0c989af 100644
> --- a/dlls/kernel32/tests/fiber.c
> +++ b/dlls/kernel32/tests/fiber.c
> @@ -158,6 +158,7 @@ static void test_FiberHandling(void)
>      if (!pCreateFiberEx)
>      {
>          win_skip( "CreateFiberEx not present\n" );
> +        test_ConvertFiberToThread();
>          return;
>      }
>
> @@ -171,6 +172,7 @@ static void test_FiberHandling(void)
>      if (!pIsThreadAFiber)
>      {
>          win_skip( "IsThreadAFiber not present\n" );
> +        test_ConvertFiberToThread();
>          return;
>      }

It's better to have a helper for such cleanup, and not to call one test
from another.

kernel32/tests: Convert fiber back to thread even when IsThreadAFiber is not available.

[Bernhard Übelacker]

Am 06.08.2016 um 14:49 schrieb Nikolay Sivov:
> On 06.08.2016 15:40, Bernhard Übelacker wrote:
> > Testbot shows the fiber test failing for Windows 2003 recently.
> > https://test.winehq.org/data/tests/kernel32:fiber.html
> >
> > This is probably related to the last change in
> > http://source.winehq.org/git/wine.git/commit/556fef3dd184e71e1f0345bd8f37259...
> >
> > As Windows 2003 seems to have no IsThreadAFiber the test never
> > calls test_ConvertFiberToThread.
> > This leads probably sometimes to the crash now visible.
> >
> > By calling test_ConvertFiberToThread before return the crash disappears
> > on Windows 2003.
> >
> > Signed-off-by: Bernhard Übelacker <bernhardu(a)mailbox.org>
> > ---
> >  dlls/kernel32/tests/fiber.c | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/dlls/kernel32/tests/fiber.c b/dlls/kernel32/tests/fiber.c
> > index 9ff9366..0c989af 100644
> > --- a/dlls/kernel32/tests/fiber.c
> > +++ b/dlls/kernel32/tests/fiber.c
> > @@ -158,6 +158,7 @@ static void test_FiberHandling(void)
> >      if (!pCreateFiberEx)
> >      {
> >          win_skip( "CreateFiberEx not present\n" );
> > +        test_ConvertFiberToThread();
> >          return;
> >      }
> >
> > @@ -171,6 +172,7 @@ static void test_FiberHandling(void)
> >      if (!pIsThreadAFiber)
> >      {
> >          win_skip( "IsThreadAFiber not present\n" );
> > +        test_ConvertFiberToThread();
> >          return;
> >      }
>
> It's better to have a helper for such cleanup, and not to call one test
> from another.

Hello,
in this case "test_FiberHandling" does call "test_ConvertFiberToThread"
some lines later already today.

Function test_ConvertFiberToThread looks to me already as such a
helper, as it is basically just the call to pConvertFiberToThread.

Kind regards,
Bernhard