Re: [LTP] [PATCH 2/2] open_posix_testsuite/../mq_timedsend/12-1: fix race

[Jan S. <jst...@re...>]

----- Original Message -----
> From: ch...@su...
> To: "Jan Stancek" <jst...@re...>
> Cc: ltp...@li...
> Sent: Tuesday, 30 July, 2013 7:35:17 PM
> Subject: Re: [LTP] [PATCH 2/2] open_posix_testsuite/../mq_timedsend/12-1: fix race
> 
> Hi!
> > -#define INTHREAD 0
> > -#define INMAIN 1
> > +#define SAFE(op) \
> > +	do {\
> > +		if ((op) == -1) {\
> > +			perror("unexpected failure"); \
> > +			exit(PTS_UNRESOLVED); \
> > +		} \
> > +	} while (0)
> 
> I do not like this approach, this does not tell what call has failed. It
> would be really hard to figure out what exact line of code has failed
> from the test error messages.

True, I'll remove that. I took if from other testcase, so we may
want to fix that one too (some day).

> 
> > -/* manual semaphore */
> > -int sem;
> > +/* variable to indicate how many times signal handler was called */
> > +static volatile sig_atomic_t in_handler;
> >  
> > -/* flag to indicate signal handler was called */
> > -int in_handler;
> > +/* errno returned by mq_timedsend() */
> > +static volatile int mq_timedsend_errno = -1;
> >  
> > -/* flag to indicate that errno was set to eintr when mq_timedsend()
> > - * was interruped. */
> > -int errno_eintr;
> > +/* pipes to synchronize thread calling mq_timedsend */
> > +static int pipefd_start[2];
> > +static int pipefd_exit[2];
> 
> What about using mutexes? (it's multithread application anyway).

Pipes seemed like less typing, but I can change that.

Regards,
Jan

> 
> --
> Cyril Hrubis
> ch...@su...
>