Signed-off-by: Jacek Caban jacek@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(+)
Hello Jacek,
On 7/9/20 12:27 PM, Jacek Caban wrote:
Signed-off-by: Jacek Caban jacek@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
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@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
-
Jacek Caban -
Zebediah Figura