[Bug libstdc++/59807] New: mutex misses destructor if non-function call initialization is used

[Bug libstdc++/59807] New: mutex misses destructor if non-function call initialization is used

[ahanins at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

            Bug ID: 59807
           Summary: mutex misses destructor if non-function call
                    initialization is used
           Product: gcc
           Version: 4.8.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ahanins at gmail dot com

Hi,
  Follow up to https://sourceforge.net/p/mingw-w64/bugs/376/

  This is related to GTHR interface to pthread.

  C++11 __mutex_base class does not define a destructor if __GTHREAD_MUTEX_INIT
is defined. It means, underlying implementation (pthread for example) has no
any means to do a resource cleanup when std::mutex is destructed. In
particular, it causes semaphore object resource (handle) leak on Windows in
MinGW winpthread implementation where semaphore object is created during first
pthread_mutex_lock invocation.
  Wouldn't it be more robust to always define a destructor for __mutex_base
which calls __gthread_mutex_destroy, or even more flexibly, introduce a
separate macro like __GTHREAD_MUTEX_DESTROY_FUNCTION which controls whether
destructor should be defined at all or not.

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[ahanins at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

--- Comment #1 from Andrey H. <ahanins at gmail dot com> ---
Simplest code which leaks handles on Windows:

for(;;) {
  std::mutex op_mutex;
  op_mutex.lock();
  op_mutex.unlock();
}

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
It a target's pthread_mutex requires cleanup then it should not define
__GTHREAD_MUTEX_INIT, it should use the init function, and then it gets a
chance to also run a destroy function.

That can be controlled by defining _GTHREAD_USE_MUTEX_INIT_FUNC in the relevant
libstdc++-v3/config/os/xxx/os_defines.h header.

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
In other words, we already have all the machinery in place to handle such
cases, it just needs to be used for the target.

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[ahanins at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

--- Comment #4 from Andrey H. <ahanins at gmail dot com> ---
Thanks Jonathan, I followed your suggestion regarding os_defines.h, recompiled
the whole mingw-w64 toolchain and can confirm that it helps. But doesn't it
meant that it should be fixed in GCC sources itself (I mean in
libstdc++-v3/config/os/mingw32-w64/os_defines.h). Or there is other standard
way how to influence os_defines.h during configuration phase?

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Yes, the copy in the sources needs to be fixed.  Does mingw-w64 always use
winpthread?

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[ktietz at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

Kai Tietz <ktietz at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-01-29
     Ever confirmed|0                           |1

--- Comment #6 from Kai Tietz <ktietz at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #5)
> Yes, the copy in the sources needs to be fixed.  Does mingw-w64 always use
> winpthread?

Sadly, no.  The use of POSIX-threading is optional.  That was actual the reason
why I didn't added it by default.  We would need to probe if pthread-library is
present, and just then activate enable it.

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[ahanins at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

--- Comment #7 from Andrey H. <ahanins at gmail dot com> ---
Kai, so what would be the fix?
IMHO, this bug is quite a show stopper for C++11 threading in Mingw-w64, so
maybe it worth to create a temporary patch for Mingw-w64 compilation with
pthreads support just to have a working toolchain?

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

--- Comment #8 from Jonathan Wakely <redi at gcc dot gnu.org> ---
If the target for this is only mingw-w64 then PR 57440 is a dup

[Bug libstdc++/59807] mutex misses destructor if non-function call initialization is used

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59807

Kai Tietz <ktietz at gcc dot gnu.org> changed:

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

--- Comment #9 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Agreed, this bug is duplicate of PR/57440

*** This bug has been marked as a duplicate of bug 57440 ***