Am Sonntag, den 07.09.2014, 11:16 -0400 schrieb Rich Felker:
> On Sun, Sep 07, 2014 at 04:45:01PM +0200, Jens Gustedt wrote:
> > Am Sonntag, den 07.09.2014, 07:32 -0400 schrieb Rich Felker:
> > > I'm not clear on whether the assignment is well-defined in pthreads,
> > > but actually attempting to use the mutex (by passing it to any of the
> > > pthread_mutex_* functions) would be UB. The same should be true for
> > > C11 threads; if not, it's a defect.
> > 
> > It is.
> 
> It is a defect? Or..?

sorry, I meant it is a defect. Yet another proof of the severe lack of
specification and semantic that this whole C thread thing has in the
standard.

> > sure, we all (should) know that, but the average user wouldn't
> > 
> > > I don't think the committee intended to forbid any of the above types
> > > of implementation; on the contrary it seems they went out of their way
> > > to support crazy types of implementations, e.g. by omitting
> > > initializers.
> > 
> > No, unfortunately for the later, the lack of a definition for default
> > initialization and initializers seems to be intentional. There are
> > people on the committee who defend the interdiction of statically
> > initialized mutexes, seemingly because some oldish windows thread
> > implementation didn't have it.
> 
> That's what I mean. By refusing to support static initialization of
> mutexes, they seem to be supporting the possibility of implementations
> for which static initialization is impractical, much like some of the
> crazy ideas I mentioned above.

<rant>
Even with those crazy ideas it is easily possible to have the
corresponding function do such a lacking initialization based on a
default 0 initialized field, in the same way pthread_once_t
works. This is a bit of an overhead at each call, but I wouldn't mind
at all penalizing any implementation that deviates from the
"all-default-initialization-is-0" rule.

That's already something the standard has for years for pointers and
floating point. A platform may have different representations for null
pointers or for 0.0. But it is the problem of the platform provider to
do everything that 0 initialization does the right thing, and not to
leave such crazy thing to the user of the type.
</rant>

Jens

-- 
:: INRIA Nancy Grand Est ::: AlGorille ::: ICube/ICPS :::
:: ::::::::::::::: office Strasbourg : +33 368854536   ::
:: :::::::::::::::::::::: gsm France : +33 651400183   ::
:: ::::::::::::::: gsm international : +49 15737185122 ::
:: http://icube-icps.unistra.fr/index.php/Jens_Gustedt ::