Project

General

Profile

Actions

Bug #2763

closed

pthread_mutex_destroy fails with error EINVAL(22) when run from main thread

Added by mneumann almost 7 years ago. Updated almost 7 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
Start date:
01/08/2015
Due date:
% Done:

0%

Estimated time:

Description

The following program when compiled with -pthread on DragonFly fails with assertion 22, while it works on Linux:

#include <assert.h>
#include <pthread.h>

int main() {
pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
assert(pthread_mutex_destroy(&lock) == 0); // returns EINVAL (22)
}

While when it goes through a lock/unlock cycle or pthread_mutex_init() is called before,
the pthread_mutex_destroy() does not fail with EINVAL:

int main() {
pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
pthread_mutex_init(&lock, NULL);
assert(pthread_mutex_destroy(&lock) == 0); // OK
}

I propose the attached patch to return success if pthread_mutex_destroy() is called
for the PTHREAD_MUTEX_INITIALIZER case.


Files

diff-thread.txt (493 Bytes) diff-thread.txt Patch to libthread_xu mneumann, 01/08/2015 02:41 AM
Actions #1

Updated by swildner almost 7 years ago

Correct me if I'm wrong, but it looks as with this patch, pthread_mutex_destroy()ing a mutex twice would no longer result in EINVAL. Do we want that?

Actions #2

Updated by mneumann almost 7 years ago

Am 08.01.2015 um 13:05 schrieb :
> Issue #2763 has been updated by swildner.
>
>
> Correct me if I'm wrong, but it looks as with this patch,
pthread_mutex_destroy()ing a mutex twice would no longer result in
EINVAL. Do we want that?

Agreed! It's better to leave as is unless we want to use a sentinel
value of say (void*)1 to detect an already destroyed mutex. Btw, same
problem occurs for condvar and rwlock.

Regards,

Michael
Actions

Also available in: Atom PDF