I wrote:
> LIU Hao wrote:
>
>> If someone forgot to initialize their mutex like `static pthread_mutex_t
>> mutex;`, they would have a
>> zero-initialized mutex.
>
> By taking a quick look at cond.c and rwlock.c, they seem to treat such
> zero-initialized object as invalid and fail with EINVAL. The code in mutex.c
> seems to be an exception.
>
>> It is, of course, no good code; but if you revert that commit, such code
>> would be broken.
>
> I think that code which forgets to properly initialize such static object is
> broken and really should fail to signal that code needs to be fixed.
>
> Still, if you think we should continue allowing zero-initialized
> `pthread_mutex_t`, I'll keep this *feature* as-is.
Now that I look at `pthread_mutex_destroy`, it stores `(pthread_mutex_t)NULL`:
```
/* Sabotage attempts to re-use the mutex before initialising it again. */
*m = (pthread_mutex_t)NULL;
```
Treating zero-initialized `pthread_mutex_t` as invalid would also allow to
diagnose use of a mutex after it was destroyed.
- Kirill Makurin
_______________________________________________
Mingw-w64-public mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
