[Bug 36296] New: valgrind shows a potential leak in msvcp90/tests/ios.c

https://bugs.winehq.org/show_bug.cgi?id=36296

Bug ID: 36296
           Summary: valgrind shows a potential leak in msvcp90/tests/ios.c
           Product: Wine
           Version: 1.7.18
          Hardware: x86
                OS: Linux
            Status: NEW
          Keywords: download, source, testcase
          Severity: normal
          Priority: P2
         Component: msvcp
          Assignee: [e-mail]
          Reporter: [e-mail]

==4993== 12 bytes in 1 blocks are possibly lost in loss record 78 of 270
==4993==    at 0x7BC4C735: notify_alloc (heap.c:255)
==4993==    by 0x7BC50F79: RtlAllocateHeap (heap.c:1716)
==4993==    by 0x5238262: MSVCRT_operator_new (heap.c:63)
==4993==    by 0x5576CF7: locale_facet_register (locale.c:257)
==4993==    by 0x557AE2C: ctype_char_use_facet (locale.c:1763)
==4993==    by 0x5565633: basic_istream_char(intXX_t, long long) (ios.c:7728)
==4993==    by 0x556570E: basic_istream_char_ipfx (ios.c:7757)
==4993==    by 0x55657E9: basic_istream_char_sentry_create (ios.c:7775)
==4993==    by 0x5567D9C: basic_istream_char_read_uint64 (ios.c:8606)
==4993==    by 0x49571BB: test_num_get_get_uint64 (ios.c:1054)
==4993==    by 0x495C89F: func_ios (ios.c:1959)
==4993==    by 0x4965F3E: run_test (test.h:584)
==4993==    by 0x496633B: main (test.h:654)
==4993==

==5033== 12 bytes in 1 blocks are possibly lost in loss record 89 of 270
==5033==    at 0x7BC4C735: notify_alloc (heap.c:255)
==5033==    by 0x7BC50F79: RtlAllocateHeap (heap.c:1716)
==5033==    by 0x5238262: MSVCRT_operator_new (heap.c:63)
==5033==    by 0x5576CF7: locale_facet_register (locale.c:257)
==5033==    by 0x5582135: numpunct_short_use_facet (locale.c:4418)
==5033==    by 0x558C55C: num_put_short__Iput (locale.c:8151)
==5033==    by 0x558CAA9: num_put_short_do_put_ulong (locale.c:8260)
==5033==    by 0x558CB7C: num_put_wchar_put_ulong (locale.c:8277)
==5033==    by 0x55634B7: basic_ostream_print_ushort (ios.c:6974)
==5033==    by 0x55635A0: basic_ostream_short_print_ushort (ios.c:6996)
==5033==    by 0x495C795: test_ostream_print_ushort (ios.c:1943)
==5033==    by 0x495C8D1: func_ios (ios.c:1969)
==5033==    by 0x4965F3E: run_test (test.h:584)
==5033==    by 0x496633B: main (test.h:654)
==5033==

==5054== 12 bytes in 1 blocks are possibly lost in loss record 88 of 270
==5054==    at 0x7BC4C735: notify_alloc (heap.c:255)
==5054==    by 0x7BC50F79: RtlAllocateHeap (heap.c:1716)
==5054==    by 0x5248262: MSVCRT_operator_new (heap.c:63)
==5054==    by 0x5586CF7: locale_facet_register (locale.c:257)
==5054==    by 0x559B659: num_put_short_use_facet (locale.c:7850)
==5054==    by 0x5573589: basic_ostream_short_print_ushort (ios.c:6996)
==5054==    by 0x495C795: test_ostream_print_ushort (ios.c:1943)
==5054==    by 0x495C8D1: func_ios (ios.c:1969)
==5054==    by 0x4965F3E: run_test (test.h:584)
==5054==    by 0x496633B: main (test.h:654)
==5054==

etc.

[Bug 36296] valgrind shows a potential leak in msvcp90/tests/ios.c

http://bugs.winehq.org/show_bug.cgi?id=36296

Piotr Caban  changed:

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

--- Comment #1 from Piotr Caban  ---
This is caused by optimization that skips memory freeing when dll is unloaded
on process exit. You need to disable it before using valgrind. Maybe we should
think about a way of running the tests with cleanup code executed.

The leaks are gone when you comment:
if (lpvReserved) break;
line in DllMain.

[Bug 36296] valgrind shows a potential leak in msvcp90/tests/ios.c

https://bugs.winehq.org/show_bug.cgi?id=36296

--- Comment #2 from Austin English  ---
*** Bug 36295 has been marked as a duplicate of this bug. ***

[Bug 36296] valgrind shows a potential leak in msvcp90/tests/ios.c

https://bugs.winehq.org/show_bug.cgi?id=36296

--- Comment #3 from Austin English  ---
(In reply to Piotr Caban from comment #1)
> This is caused by optimization that skips memory freeing when dll is
> unloaded on process exit. You need to disable it before using valgrind.
> Maybe we should think about a way of running the tests with cleanup code
> executed.

That would be very useful, for Coverity and maybe Clang or other tools as well.

[Bug 36296] valgrind shows a potential leak in msvcp90/tests/ios.c

https://bugs.winehq.org/show_bug.cgi?id=36296

Piotr Caban  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
      Fixed by SHA1|                            |73a996de013264379c1b68e9512
                   |                            |aa127b0b21bfb
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #4 from Piotr Caban  ---
In this case it was quite easy to force dll unload before process exit because
msvcp90.dll is loaded dynamically.

[Bug 36296] valgrind shows a potential leak in msvcp90/tests/ios.c

https://bugs.winehq.org/show_bug.cgi?id=36296

Alexandre Julliard  changed:

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

--- Comment #5 from Alexandre Julliard  ---
Closing bugs fixed in 1.7.19.