RE: Correction to crash inside RtlCaptureStackBackTrace() + test case

[Janne Hakonen]

Hi again,

I've been working to implement the tests for RtlCaptureStackBackTrace(), but I have encountered a problem. The function, according to MSDN should be available only in Windows XP and newer OSes.
Now, when I ran these tests on Testbot using all base VMs, the tests were run also on W2KPROSP4 and WNT4WSSP6. However, the function seems to buggy in those. It throws access violation whenever it is not skipping frames.

Here's the test patch I used:
http://testbot.winehq.org/GetFile.pl?JobKey=8606&StepKey=1

In both cases, test on patch's line 83 failed and call on line 92 crashed the test.

In other VMs the tests succeeded and none crashed.

Is there some way I could prevent/skip these tests to be run on Windows NT4 and 2000?

Best Regards,
Janne Hakonen

Correction to crash inside RtlCaptureStackBackTrace() + test case

[Janne Hakonen]

> Now, when I ran these tests on Testbot using all base VMs, the tests were run also on W2KPROSP4 and WNT4WSSP6. However, the 
> function seems to buggy in those. It throws access violation whenever it is not skipping frames.
> Is there some way I could prevent/skip these tests to be run on Windows NT4 and 2000?

Ah, no matter, I think I've got it. Adding following lines to start allows the tests to be skipped on NT and 2000:

__TRY
{
    pRtlCaptureStackBackTrace(0, 62, callers, NULL);
}
__EXCEPT_ALL
{
    win_skip("RtlCaptureStackBackTrace threw SEH exception, "
             "most likely running on Windows NT or 2000\n");
    return;
}
__ENDTRY

Thanks,
Janne

Correction to crash inside RtlCaptureStackBackTrace() + test case

[James McKenzie]

On 1/28/11 11:54 AM, Janne Hakonen wrote:
> > Now, when I ran these tests on Testbot using all base VMs, the tests
>
> were run also on W2KPROSP4 and WNT4WSSP6. However, the
> > function seems to buggy in those. It throws access violation
>
> whenever it is not skipping frames.
> > Is there some way I could prevent/skip these tests to be run on
>
> Windows NT4 and 2000?
>
> Ah, no matter, I think I've got it. Adding following lines to start 
> allows the tests to be skipped on NT and 2000:
>
> __TRY
> {
>     pRtlCaptureStackBackTrace(0, 62, callers, NULL);
> }
> __EXCEPT_ALL
> {
>     win_skip("RtlCaptureStackBackTrace threw SEH exception, "
>              "most likely running on Windows NT or 2000\n");
>     return;
> }
> __ENDTRY

Did you try this on WindowsXP or Windows7 to make sure that the win_skip 
did not function on those platforms?  My understanding is that win_skip 
works on all Windows platforms not just WindowsNT or Windows2000.

Also, it is custom to mark your resubmissions with corrections as [try 
x].  Makes it easier to determine what you are doing.

James McKenzie

Correction to crash inside RtlCaptureStackBackTrace() + test case

[Janne Hakonen]

> Did you try this on WindowsXP or Windows7 to make sure that the
    win_skip did not function on those platforms?  My understanding is
    that win_skip works on all Windows platforms not just WindowsNT or
    Windows2000.
Well, I did test that skip doesn't happen on my home computer, which is Windows 7. I also tested that the skip do happen on Windows NT4 and 2000 on Testbot.

>Also, it is custom to mark your resubmissions with corrections as
    [try x].  Makes it easier to determine what you are doing.

I posted my patches to patches mailing list only once so far, so there hasn't been any retries yet. 
But, I shall add that tag if/when I need to adjust those or any future patches.
If I need to make changes to, say, test cases, should I also send bug fix patch again even if there is no changes to that?

Best Regards,
Janne

Correction to crash inside RtlCaptureStackBackTrace() + test case

[Austin English]

On Sun, Jan 30, 2011 at 22:15, Janne Hakonen  wrote:
> > Also, it is custom to mark your resubmissions with corrections as [try x].
> > Makes it easier to determine what you are doing.
>
> I posted my patches to patches mailing list only once so far, so there
> hasn't been any retries yet.
> But, I shall add that tag if/when I need to adjust those or any future
> patches.
> If I need to make changes to, say, test cases, should I also send bug fix
> patch again even if there is no changes to that?

Please. If it is two separate patches, be sure to make it a series
(label them 1/2, 2/2) and that the tests pass before/after each patch.

Correction to crash inside RtlCaptureStackBackTrace() + test case

[Janne Hakonen]

> > If I need to make changes to, say, test cases, should I also send bug fix
> > patch again even if there is no changes to that?
>
> Please. If it is two separate patches, be sure to make it a series
> (label them 1/2, 2/2) and that the tests pass before/after each patch.

Ok, thanks for the info. :)

Here's the job run by Testbot, when I posted the patches to the mailing list:

http://testbot.winehq.org/JobDetails.pl?Key=8716

Looking at the logs only WNT4WSSP6 and W2KPROSP4 shows message:
rtl.c:1340: Tests skipped: RtlCaptureStackBackTrace threw SEH exception, most likely running on Windows NT or 2000

Best Regards,
Janne