Date: Thu, 20 Mar 2008 08:32:16 +0100 From: Heiko Wundram <modelnine@modelnine.org> To: freebsd-stable@freebsd.org Subject: pthread_mutexattr_settype non-conformance to man-page and POSIX Message-ID: <200803200832.16340.modelnine@modelnine.org>
next in thread | raw e-mail | index | archive | help
I hit a bug in libthr with pthread_mutexattr_settype which (at least as far as
I understand the POSIX reference and also the man-page) makes it
non-conformant to the specifications.
Quoting:
"""
RETURN VALUES
If successful, these functions return 0. Otherwise, an error number is
returned to indicate the error.
...
ERRORS
...
The pthread_mutexattr_settype() function will fail if:
[EINVAL] Invalid value for attr, or invalid value for type.
"""
This does not happen (at least not in libthr):
"""
int
_pthread_mutexattr_settype(pthread_mutexattr_t *attr, int type)
{
int ret;
if (attr == NULL || *attr == NULL || type >= PTHREAD_MUTEX_TYPE_MAX) {
errno = EINVAL;
ret = -1;
} else {
(*attr)->m_type = type;
ret = 0;
}
return(ret);
}
"""
The error code EINVAL is stored to errno, and -1 is returned, which is wrong
at least according to my understanding of the man-page, which pretty much
says the same thing as POSIX.
I haven't looked yet whether this code has always been this way for FreeBSD,
but at least under glibc[+NPTL], the error-number is returned directly as the
return value (from a quick skimming of the sources), just as I would
understand the specifications.
The code in question is similar to pretty much all other functions in libthr,
which also set errno and return -1, so basically it's a systematic error in
the implementation of libthr, if I'm not completely mistaken.
Anybody else have any more insight on this?
--
Heiko Wundram
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200803200832.16340.modelnine>