gamedevlists-windows

[GD-Windows] RE: [Algorithms] Finding driver faults

[Jon W. <hp...@mi...>]

I'm moving this to gamedevlists-windows (if you're not there,
then please subscribe)

If it's Windows XP, you can configure the machine to write a
minidump, which you can have them mail to you. Then open that
minidump using "Open Minidump" in the WinDbg debugger; open
the symbol file for the version of the app that they're running,
and start looking up the recorded stack to find an address
inside where your app was loaded -- that's your return address,
most likely. Do a "kv" from there (or similar).

If you can connect WinDbg using FireWire (also requires WinXP)
and boot the client machine with debugging mode turned on, that
will let you catch the problem in the act and debug while it's
there. The nice thing is it doesn't require installing any
debugger anything on the target machine!

If you're on Win2k, WinDbg still works, but usually is run with
serial cables (SLOOOOOOOOOW) and can't do minidumps.

If you're on 16-bit Windows, there's RTERM and DEBUGGER.EXE for
cross-debugging kernel mode, but it's really quite primitive
tools and they're really hard to set up and use. I've found one
(1) bug in our app that way. (of course, I wouldn't have found
it at all any other way, so I guess it's still a good thing)

Cheers,

			/ h+

> -----Original Message-----
> From: gda...@li...
> [mailto:gda...@li...]On Behalf Of Phil
> Scadden
> Sent: Sunday, October 20, 2002 12:44 PM
> To: gda...@li...
> Subject: [Algorithms] Finding driver faults
>
>
> Having hassle with rare but annoying crashes on an nVidea driver.
>
>    Access violation at address 01775844 in module 'nvoglnt.dll'
>    Write of address 00000000
>
> Only on client machines, never in office. What I would like to
> know is what
> the call sequence in my code is that triggers the error to see if there is
> a possibility for a workaround. Cant give nVidea any meaningful
> information
> without it. Anyone got ideas on how to trace such an error?
>
>
> ----------------------------------------------------------
> Phil Scadden, Institute of Geological and Nuclear Sciences
> 41 Bell Rd South, PO Box 30368, Lower Hutt, New Zealand
> Ph +64 4 5704821, fax +64 4 5704603
>
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:
> Access Your PC Securely with GoToMyPC. Try Free Now
> https://www.gotomypc.com/s/OSND/DD
> _______________________________________________
> GDAlgorithms-list mailing list
> GDA...@li...
> https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
> Archives:
> http://sourceforge.net/mailarchive/forum.php?forum_id=6188
>

RE: [GD-Windows] RE: Finding driver faults

[Daniel V. <vo...@ep...>]

I don't want to move it again but here are some OpenGL specific hints:

> >    Write of address 00000000

gl*Pointer could be NULL (driver can't check for that)
pixels argument of glCompressedTexImage*DARB could be NULL (driver can't
check for that)

Seeing it is a write those two cases might be unlikely but there are a
couple of places in OpenGL where NULL is a valid argument the driver can't
check for because the spec doesn't say so... and in theory you could come up
with a stride/ index/ ... scenario where NULL would map into the client
address space :)

If you ask the client to run a program like glTrace to find out exactly in
which OpenGL call it dies it should be much easier to track down.

-- Daniel, Epic Games Inc.

> -----Original Message-----
> From: gam...@li...
> [mailto:gam...@li...]On Behalf Of
> Jon Watte
> Sent: Monday, October 21, 2002 1:28 AM
> To: Phil Scadden
> Cc: gam...@li...
> Subject: [GD-Windows] RE: [Algorithms] Finding driver faults
>
>
>
> I'm moving this to gamedevlists-windows (if you're not there,
> then please subscribe)
>
> If it's Windows XP, you can configure the machine to write a
> minidump, which you can have them mail to you. Then open that
> minidump using "Open Minidump" in the WinDbg debugger; open
> the symbol file for the version of the app that they're running,
> and start looking up the recorded stack to find an address
> inside where your app was loaded -- that's your return address,
> most likely. Do a "kv" from there (or similar).
>
> If you can connect WinDbg using FireWire (also requires WinXP)
> and boot the client machine with debugging mode turned on, that
> will let you catch the problem in the act and debug while it's
> there. The nice thing is it doesn't require installing any
> debugger anything on the target machine!
>
> If you're on Win2k, WinDbg still works, but usually is run with
> serial cables (SLOOOOOOOOOW) and can't do minidumps.
>
> If you're on 16-bit Windows, there's RTERM and DEBUGGER.EXE for
> cross-debugging kernel mode, but it's really quite primitive
> tools and they're really hard to set up and use. I've found one
> (1) bug in our app that way. (of course, I wouldn't have found
> it at all any other way, so I guess it's still a good thing)
>
> Cheers,
>
> 			/ h+
>
> > -----Original Message-----
> > From: gda...@li...
> > [mailto:gda...@li...]On Behalf Of Phil
> > Scadden
> > Sent: Sunday, October 20, 2002 12:44 PM
> > To: gda...@li...
> > Subject: [Algorithms] Finding driver faults
> >
> >
> > Having hassle with rare but annoying crashes on an nVidea driver.
> >
> >    Access violation at address 01775844 in module 'nvoglnt.dll'
> >    Write of address 00000000
> >
> > Only on client machines, never in office. What I would like to
> > know is what
> > the call sequence in my code is that triggers the error to see
> if there is
> > a possibility for a workaround. Cant give nVidea any meaningful
> > information
> > without it. Anyone got ideas on how to trace such an error?
> >
> >
> > ----------------------------------------------------------
> > Phil Scadden, Institute of Geological and Nuclear Sciences
> > 41 Bell Rd South, PO Box 30368, Lower Hutt, New Zealand
> > Ph +64 4 5704821, fax +64 4 5704603
> >
> >
> >
> > -------------------------------------------------------
> > This sf.net email is sponsored by:
> > Access Your PC Securely with GoToMyPC. Try Free Now
> > https://www.gotomypc.com/s/OSND/DD
> > _______________________________________________
> > GDAlgorithms-list mailing list
> > GDA...@li...
> > https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
> > Archives:
> > http://sourceforge.net/mailarchive/forum.php?forum_id=6188
> >
>
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Gamedevlists-windows mailing list
> Gam...@li...
> https://lists.sourceforge.net/lists/listinfo/gamedevlists-windows
> Archives:
> http://sourceforge.net/mailarchive/forum.php?forum_id=555
>