[Bug 38440] New: valgrind shows a definite leak in comdlg32/filedlg

https://bugs.winehq.org/show_bug.cgi?id=38440

Bug ID: 38440
           Summary: valgrind shows a definite leak in comdlg32/filedlg
           Product: Wine
           Version: 1.7.41
          Hardware: x86
                OS: Linux
            Status: NEW
          Keywords: download, source, testcase, valgrind
          Severity: minor
          Priority: P2
         Component: comdlg32
          Assignee: [e-mail]
          Reporter: [e-mail]
                CC: [e-mail]
      Distribution: ---

==6025== 28 bytes in 1 blocks are definitely lost in loss record 237 of 1,085
==6025==    at 0x7BC49E0E: notify_alloc (heap.c:254)
==6025==    by 0x7BC4E044: RtlAllocateHeap (heap.c:1715)
==6025==    by 0x7BC43059: alloc_fileio (file.c:381)
==6025==    by 0x7BC45096: server_ioctl_file (file.c:1404)
==6025==    by 0x7BC45A43: NtFsControlFile (file.c:1684)
==6025==    by 0x7B872671: ConnectNamedPipe (sync.c:1615)
==6025==    by 0x4FEA6EF: listen_thread (rpc_transport.c:132)
==6025==    by 0x7BC8023F: ??? (signal_i386.c:2682)
==6025==    by 0x7BC80286: call_thread_func (signal_i386.c:2741)
==6025==    by 0x7BC8021D: ??? (signal_i386.c:2682)
==6025==    by 0x7BC8726D: start_thread (thread.c:443)
==6025==    by 0x4214EFA: start_thread (pthread_create.c:309)
==6025==    by 0x431362D: clone (clone.S:129)
==6025==

[Bug 38440] valgrind shows a definite leak in comdlg32/filedlg

https://bugs.winehq.org/show_bug.cgi?id=38440

--- Comment #1 from Austin English  ---
Also seen in itemdlg:
==6055== 28 bytes in 1 blocks are definitely lost in loss record 255 of 1,238
==6055==    at 0x7BC49E0E: notify_alloc (heap.c:254)
==6055==    by 0x7BC4E044: RtlAllocateHeap (heap.c:1715)
==6055==    by 0x7BC43059: alloc_fileio (file.c:381)
==6055==    by 0x7BC45096: server_ioctl_file (file.c:1404)
==6055==    by 0x7BC45A43: NtFsControlFile (file.c:1684)
==6055==    by 0x7B872671: ConnectNamedPipe (sync.c:1615)
==6055==    by 0x4FEA6EF: listen_thread (rpc_transport.c:132)
==6055==    by 0x7BC8023F: ??? (signal_i386.c:2682)
==6055==    by 0x7BC80286: call_thread_func (signal_i386.c:2741)
==6055==    by 0x7BC8021D: ??? (signal_i386.c:2682)
==6055==    by 0x7BC8726D: start_thread (thread.c:443)
==6055==    by 0x4214EFA: start_thread (pthread_create.c:309)
==6055==    by 0x431362D: clone (clone.S:129)
==6055==

and several other tests, it seems..

[Bug 38440] valgrind shows a definite leak in comdlg32/filedlg

https://bugs.winehq.org/show_bug.cgi?id=38440

Austin English  changed:

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

[Bug 38440] valgrind shows a definite leak in comdlg32/filedlg

https://bugs.winehq.org/show_bug.cgi?id=38440

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
       Distribution|---                         |Debian

[Bug 38440] valgrind shows a definite leak in comdlg32/filedlg

https://bugs.winehq.org/show_bug.cgi?id=38440

--- Comment #2 from Austin English  ---
A similar bug in programs/wscript/tests/run.c:
==20417== 28 bytes in 1 blocks are definitely lost in loss record 68 of 325
==20417==    at 0x7BC49E0E: notify_alloc (heap.c:254)
==20417==    by 0x7BC4E044: RtlAllocateHeap (heap.c:1715)
==20417==    by 0x7BC43059: alloc_fileio (file.c:381)
==20417==    by 0x7BC45096: server_ioctl_file (file.c:1404)
==20417==    by 0x7BC45A43: NtFsControlFile (file.c:1684)
==20417==    by 0x7B872671: ConnectNamedPipe (sync.c:1615)
==20417==    by 0x48CE7D3: local_server_thread (rpc.c:1917)
==20417==    by 0x7BC8023F: ??? (signal_i386.c:2682)
==20417==    by 0x7BC80286: call_thread_func (signal_i386.c:2741)
==20417==    by 0x7BC8021D: ??? (signal_i386.c:2682)
==20417==    by 0x7BC8726D: start_thread (thread.c:443)
==20417==    by 0x4214EFA: start_thread (pthread_create.c:309)
==20417==    by 0x431362D: clone (clone.S:129)
==20417==

[Bug 38440] valgrind shows a definite leak in comdlg32/filedlg

https://bugs.winehq.org/show_bug.cgi?id=38440

Sebastian Lackner  changed:

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

--- Comment #3 from Sebastian Lackner  ---
I've looked at the problem in wscript and there is nothing to fix. Valgrind
thinks the memory is lost because there is no pointer to the allocated memory
in the process itself, but the wineserver still has a pointer (because of the
queued APC). So basically a limitation of Valgrind.

I am not sure whats the best way to get rid of these warnings, blacklisting
them all manually seems a bit complicated, and similar problems might also
occur in ws2_32 with async io structures. On the other hand, updating something
like a linked list with pending async IOs is a bit complicated, when it just
has the purpose to make Valgrind happy. Anyone has a better idea?