[PATCH v2 01/11] kernel32/tests: Add more invalid handle tests.

[Jacek Caban]

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com>
---
v2: Don't depend on ioctl to set proper last error, it should fix 
failures on TestBot. The first patch demonstrates the problem without 
depending on the way tests are launched.

  dlls/kernel32/tests/console.c | 63 +++++++++++++++++++++++++++++++++++
  1 file changed, 63 insertions(+)

[Zebediah Figura]

Hello Jacek,

On 7/9/20 12:27 PM, Jacek Caban wrote:
> Signed-off-by: Jacek Caban <jacek(a)codeweavers.com>
> ---
> v2: Don't depend on ioctl to set proper last error, it should fix
> failures on TestBot. The first patch demonstrates the problem without
> depending on the way tests are launched.
>
>  dlls/kernel32/tests/console.c | 63 +++++++++++++++++++++++++++++++++++
>  1 file changed, 63 insertions(+)

I probably should have mentioned this earlier, but in truth it took
until noticing this patch series to remember.

In bug 46635 an application passes standard handles to NtQueryObject(),
expecting it to return failure if the handle is a console handle. This
leads me to believe that console handles should not be NT kernel handles
at all, which means that this patch series' approach to checking invalid
handles may not be quite correct. Possibly instead we should use a local
table...?

ἔρρωσο,
Zeb

[Jacek Caban]

On 09.07.2020 19:47, Zebediah Figura wrote:
> Hello Jacek,
>
> On 7/9/20 12:27 PM, Jacek Caban wrote:
> > Signed-off-by: Jacek Caban <jacek(a)codeweavers.com>
> > ---
> > v2: Don't depend on ioctl to set proper last error, it should fix
> > failures on TestBot. The first patch demonstrates the problem without
> > depending on the way tests are launched.
> >
> >  dlls/kernel32/tests/console.c | 63 +++++++++++++++++++++++++++++++++++
> >   1 file changed, 63 insertions(+)
>
> I probably should have mentioned this earlier, but in truth it took
> until noticing this patch series to remember.
>
> In bug 46635 an application passes standard handles to NtQueryObject(),
> expecting it to return failure if the handle is a console handle. This
> leads me to believe that console handles should not be NT kernel handles
> at all, which means that this patch series' approach to checking invalid
> handles may not be quite correct. Possibly instead we should use a local
> table...?

They are kernel handles since Windows 8. I have tests in a queue showing 
that ObjectNameInformation should return "\Device\ConDrv" for console 
handles.


Thanks,

Jacek