On Tue, Jun 24, 2025 at 03:22:21PM +1000, David Gwynne wrote:
> On Fri, Jun 20, 2025 at 12:00:13PM +0000, Job Snijders wrote:
> > Perhaps good to add error checking to the pthread_* calls?
> 
> yes...
> 
> > While there, the rwlock_unlock in auth_insert() seemed somewhat
> > superfluous given that the program errors out right after.
> > 
> > OK?
> 
> sorry to go way back in time on this thread, but the way you're handling
> errors from the pthread api and how i understand that api don't line up.
> 
> my understanding is that pthread functions return error codes
> directly, and they don't set errno. therefore, rather than handling
> errors like this:
> 
> 	if (pthread_create(&writer, NULL, &parse_writer, &fd) != 0)
> 		errx(1, "pthread_create");
> 
> you should be doing this:
> 
> 	int error;
> 
> 	error = pthread_create(&writer, NULL, &parse_writer, &fd);
> 	if (error != 0)
> 		errc(1, error, "pthread_create");

The reason for this is a cheat around a portability issue with errc(),
which isn't provided by glibc and musl. We should fix this in portable.
But automake is terribly annoying.