Assertion in win32::WaitForSingleObject()

[robert.stewart@…]

I get the following assertion when I used Ctrl-C on an app:

Assertion failed: win32::WaitForSingleObject( sem,::boost::detail::win32::infinite)==0, file ...\boost\thread\win32\basic_timed_mutex.hpp, line 79

Examination of that line of code reveals that the error handling is insufficient. WaitForSingleObject() can return WAIT_ABANDONED if the lock is held when the owning thread is released. In non-debug builds, this means the code falls through to the next statement in such cases, whereas in debug BOOST_VERIFY reports an error. Either way, WAIT_ABANDONED should be handled.

(Refer to ​http://msdn.microsoft.com/en-us/library/windows/desktop/ms687032.aspx for details on the API.)

[viboes]

What would be the expected behavior when WaitForSingleObject() returns WAIT_ABANDONED?

[viboes]

ping?

[robert.stewart@…]

Since that return value means the calling thread owns the mutex by default of the previous owner, it should be treated as equivalent to WAIT_OBJECT0. Thus, the following seems appropriate:

DWORD const retval(win32::WaitForSingleObject(sem, ::boost::detail::win32::infinite));
BOOST_VERIFY(WAIT_OBJECT0 == retval || WAIT_ABANDONED == retval);

You might prefer to repackage WAIT_ABANDONED like WAIT_TIMEOUT is, so the following may be more in keeping with the current code:

DWORD const retval(win32::WaitForSingleObject(sem, ::boost::detail::win32::infinite));
BOOST_VERIFY(0 == retval || ::boost::detail::win32::abandoned == retval);

[viboes]

Yes, this can be changed. I will do it soon.

[viboes]

Committed in trunk [84710]

[viboes]

Committed revision [84750].