[Bug 36783] New: valgrind shows two invalid writes in kernel32/tests/thread.c

https://bugs.winehq.org/show_bug.cgi?id=36783

Bug ID: 36783
           Summary: valgrind shows two invalid writes in
                    kernel32/tests/thread.c
           Product: Wine
           Version: 1.7.20
          Hardware: x86
                OS: Linux
            Status: NEW
          Keywords: download, source, testcase, valgrind
          Severity: normal
          Priority: P2
         Component: kernel32
          Assignee: [e-mail]
          Reporter: [e-mail]

==27802== Invalid write of size 4
==27802==    at 0x4E34830: test_SetThreadContext (thread.c:988)
==27802==    by 0x4E37887: func_thread (thread.c:1763)
==27802==    by 0x4E537EE: run_test (test.h:584)
==27802==    by 0x4E53BBA: main (test.h:654)
==27802==  Address 0x6a8e8d4 is on thread 1's stack
==27802==

==27802== Invalid write of size 4
==27802==    at 0x4E34842: test_SetThreadContext (thread.c:989)
==27802==    by 0x4E37887: func_thread (thread.c:1763)
==27802==    by 0x4E537EE: run_test (test.h:584)
==27802==    by 0x4E53BBA: main (test.h:654)
==27802==  Address 0x6a8e8d0 is on thread 1's stack
==27802==

note: I disabled test_thread_fpu_cw() because of a valgrind bug, which you may
need to do to see this.

[Bug 36783] valgrind shows two invalid writes in kernel32/tests/thread.c

http://bugs.winehq.org/show_bug.cgi?id=36783

Sebastian Lackner  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #1 from Sebastian Lackner  ---
I fear we cannot really fix this. For me the corresponding code looks okay:

--- snip ---
986         /* simulate a call to set_test_val(10) */
987         stack = (int *)ctx.Esp;
988         stack[-1] = 10;                // <--
989         stack[-2] = ctx.Eip;           // <--
990         ctx.Esp -= 2 * sizeof(int *);
991         ctx.Eip = (DWORD)set_test_val;
--- snip ---
( http://source.winehq.org/source/dlls/kernel32/tests/thread.c#L986 )

The problem is that Valgrind doesn't understand what we're doing here (manually
adjusting the stack pointer and writing values on the stack to simulate a
function call).

[Bug 36783] valgrind shows two invalid writes in kernel32/tests/thread.c

https://bugs.winehq.org/show_bug.cgi?id=36783

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |WONTFIX

--- Comment #2 from Austin English  ---
(In reply to Sebastian Lackner from comment #1)
> I fear we cannot really fix this. For me the corresponding code looks okay:
>
> --- snip ---
> 986         /* simulate a call to set_test_val(10) */
> 987         stack = (int *)ctx.Esp;
> 988         stack[-1] = 10;                // <--
> 989         stack[-2] = ctx.Eip;           // <--
> 990         ctx.Esp -= 2 * sizeof(int *);
> 991         ctx.Eip = (DWORD)set_test_val;
> --- snip ---
> ( http://source.winehq.org/source/dlls/kernel32/tests/thread.c#L986 )
>
> The problem is that Valgrind doesn't understand what we're doing here
> (manually adjusting the stack pointer and writing values on the stack to
> simulate a function call).

Marked as intentional, thanks.

[Bug 36783] valgrind shows two invalid writes in kernel32/tests/thread.c

https://bugs.winehq.org/show_bug.cgi?id=36783

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #3 from Austin English  ---
Closing.