Crash in pthread_detach (double free)

[kaliber]

Test case: ​http://posixtest.cvs.sourceforge.net/viewvc/posixtest/posixtestsuite/conformance/interfaces/pthread_create/2-1.c?revision=1.2&view=markup

[v]

Seems like a silly thing to do, calling pthread_detach() after pthread_join(), but if it's a valid action perhaps attached patch is an acceptable solution.
I'd be inclined to return B_BAD_VALUE for a thread that has already been detached, but that doesn't agree with the testcase.

The patch has not been compiled or tested.
I couldn't find any POSIX doc that agreed or disagreed with the testcase, but I haven't looked very hard.

[v]

updated patch

[v]

Updated patch. The previous patch went against the coding guidelines by assigning in an if statement.

[kaliber]

According to specification ​http://www.opengroup.org/onlinepubs/000095399/functions/pthread_detach.html function should return 0 instead of B_OK and EINVAL instead of B_BAD_VALUE.

[anevilyak]

Replying to kaliber:

According to specification ​http://www.opengroup.org/onlinepubs/000095399/functions/pthread_detach.html function should return 0 instead of B_OK and EINVAL instead of B_BAD_VALUE.

Which are precisely the values B_OK and B_BAD_VALUE map to.

[axeld]

They are now, but in POSIX code we should actually use the POSIX error codes as well.

[v]

patch with posix error codes

[v]

How about haiku_ticket4828_try2.diff then? All BeOS/Haiku style error codes in pthread.c are replaced by the POSIX versions (as far as I can see). In addition to modifications to pthread_detach().

[axeld]

Thanks for the patch! I've updated it slightly (and fixed a minor coding style violation) in hrev33775.