[Bug 18067] New: Large blocks not reported properly to valgrind by RtlAllocateHeap

http://bugs.winehq.org/show_bug.cgi?id=18067

Summary: Large blocks not reported properly to valgrind by
                    RtlAllocateHeap
           Product: Wine
           Version: 1.1.19
          Platform: Other
        OS/Version: other
            Status: NEW
          Keywords: download, patch, source
          Severity: normal
          Priority: P2
         Component: ntdll
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Found while valgrinding chromium unit tests.  Here's a simple test case:

$ cat > foo.c <<_EOF_
#include <malloc.h>
int main()
{
    char *p = malloc(1000000);
    free(p);
}
_EOF_
$ i586-mingw32msvc-gcc foo.c
$ valgrind --trace-children=yes wine a.exe
...
==2953== Invalid free() / delete / delete[]
==2953==    at 0x6C80EDF: RtlFreeHeap dlls/ntdll/heap.c:1464

Looks like Eric missed a code path when he first added the valgrind hooks?

[Bug 18067] Large blocks not reported properly to valgrind by RtlAllocateHeap

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #1 from Dan Kegel   2009-04-15 16:37:50 ---
Patch sent, http://www.winehq.org/pipermail/wine-patches/2009-April/071894.html

[Bug 18067] Large blocks not reported properly to valgrind by RtlAllocateHeap

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #2 from Dan Kegel   2009-04-16 23:38:47 ---
Patch committed, but Rob pointed out a related problem on wine-devel, let's
keep this bug open for the moment.

[Bug 18067] Large blocks not reported properly to valgrind by RtlAllocateHeap

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #3 from Austin English   2009-04-17 11:44:32 ---
(In reply to comment #2)
> Patch committed, but Rob pointed out a related problem on wine-devel, let's
> keep this bug open for the moment.

http://www.winehq.org/pipermail/wine-devel/2009-April/074890.html

2009/4/15 Dan Kegel <dank at kegel.com>:
> Two chromium unit tests reported phony errors
> when run on wine on valgrind because
> RtlAllocateHeap sometimes returned an
> allocation without notifying valgrind about it.

RtlReAllocateHeap is also missing valgrind notifications for the large
block paths. It also appears that the function leaks memory in the
realloc small -> large path.

[Bug 18067] Large blocks not reported properly to valgrind by RtlAllocateHeap

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #4 from Dan Kegel   2009-04-19 00:37:08 ---
I've started on tests for this.  (You have to run them under valgrind
to get the full effect, but that's ok.)

[Bug 18067] Large blocks not reported properly to valgrind by RtlAllocateHeap

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #5 from Dan Kegel   2009-04-19 21:24:32 ---
Sent next patch,
http://www.winehq.org/pipermail/wine-patches/2009-April/072007.html
I probably won't fix the leak, but at least it has a FIXME now

[Bug 18067] RtlAllocateHeap leaks memory in the small->large path

http://bugs.winehq.org/show_bug.cgi?id=18067

Dan Kegel  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Large blocks not reported   |RtlAllocateHeap leaks memory
                   |properly to valgrind by     |in the small->large path
                   |RtlAllocateHeap             |

--- Comment #6 from Dan Kegel   2009-04-27 10:19:04 ---
OK, the rest of the hooks are in, and there's a FIXME comment next to the leak:
http://www.winehq.org/pipermail/wine-cvs/2009-April/054932.html

The only part remaining is the memory leak.
Changing summary.

[Bug 18067] RtlAllocateHeap leaks memory in the small->large path

http://bugs.winehq.org/show_bug.cgi?id=18067

Chris Humbert  changed:

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

--- Comment #7 from Chris Humbert   2009-05-15 21:43:05 ---
*** Bug 18488 has been marked as a duplicate of this bug. ***

[Bug 18067] RtlAllocateHeap leaks memory in the small->large path

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #8 from Austin English   2009-05-18 10:20:07 ---
Probably fixed by
http://source.winehq.org/git/wine.git/?a=commitdiff;h=13a33b73c349530b17347d....

[Bug 18067] RtlAllocateHeap leaks memory in the small->large path

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #9 from Nikolay Sivov   2009-10-19 15:30:38 ---
(In reply to comment #8)
> Probably fixed by
> http://source.winehq.org/git/wine.git/?a=commitdiff;h=13a33b73c349530b17347d....

Hi, Dan. The last FIXME was fixed with this commit. Is this bug fixed?

[Bug 18067] RtlAllocateHeap leaks memory in the small->large path

http://bugs.winehq.org/show_bug.cgi?id=18067

Dan Kegel  changed:

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

--- Comment #10 from Dan Kegel   2009-10-19 15:54:02 ---
Probably.  I'll know for sure next time I run all the tests.

[Bug 18067] RtlAllocateHeap leaks memory in the small->large path

http://bugs.winehq.org/show_bug.cgi?id=18067

--- Comment #11 from Nikolay Sivov   2009-10-19 15:58:13 ---
(In reply to comment #10)
> Probably.  I'll know for sure next time I run all the tests.

Forgot to mention, your example from description doesn't produce any output
with RtlFreeHeap() in it - I'm not skilled in valgrind output interpretation,
but this word doesn't occur.

[Bug 18067] RtlAllocateHeap leaks memory in the small->large path

http://bugs.winehq.org/show_bug.cgi?id=18067

Alexandre Julliard  changed:

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

--- Comment #12 from Alexandre Julliard   2009-10-23 13:19:05 ---
Closing bugs fixed in 1.1.32.