Re: [ros-dev] [ros-diffs] [jimtabor] 37783: - Fixed hooking bug for Firefox, a plus, new SEH found a bug!

[Timo Kreuzer]

This looks like you're creating a mem leak.
> Author: jimtabor
> Date: Sun Nov 30 17:39:20 2008
> New Revision: 37783
>
> URL: http://svn.reactos.org/svn/reactos?rev=37783&view=rev http://svn.reactos.org/svn/reactos?rev=37783&view=rev
> Log:
> Fixed hooking bug for Firefox, a plus, new SEH found a bug!
>
> Modified:
>     trunk/reactos/subsystems/win32/win32k/ntuser/callback.c
>
> Modified: trunk/reactos/subsystems/win32/win32k/ntuser/callback.c
> URL: http://svn.reactos.org/svn/reactos/trunk/reactos/subsystems/win32/win32k/ntu... http://svn.reactos.org/svn/reactos/trunk/reactos/subsystems/win32/win32k/ntuser/callback.c?rev=37783&r1=37782&r2=37783&view=diff
> ==============================================================================
> --- trunk/reactos/subsystems/win32/win32k/ntuser/callback.c [iso-8859-1] (original)
> +++ trunk/reactos/subsystems/win32/win32k/ntuser/callback.c [iso-8859-1] Sun Nov 30 17:39:20 2008
> @@ -507,12 +507,12 @@
>     }
>     _SEH2_END;
> IntCbFreeMemory(Argument);
> if (!NT_SUCCESS(Status))
>   {
>      return 0;
>   }
>
> if (Argument) IntCbFreeMemory(Argument);
>
> return Result;
>
> }

[James Tabor]

Hi,
It looks like Firefox and friends return back from user callback more
than once. This safeguards the return so it does not crash. My guess
we are doing something wrong again. It reminds me of the AbiWord
GdiFlush abuse, maybe this is the same where their developers assume
something and it was backed up by someone from M$ as a joke. It's all
about performance, the GdiFlush calls slow down AbiWord maybe this is
the same with the erroneous callbacks for hooks. Technically it is a
call from user space to kernel space. I need more time later to
research this.
Thanks,
James

On Thu, Dec 4, 2008 at 6:00 AM, Timo Kreuzer [e-mail] wrote:
> This looks like you're creating a mem leak.

Ref: http://bugzilla.abisource.com/show_bug.cgi?id=11282