On Apr 20 15:14, Eric Blake wrote:
> Corinna Vinschen <corinna-cygwin <at> cygwin.com> writes:
>
> > > > I'd prefer a testcase in C.
> > >
> > > So would I. I'm not even sure whether perl was using flock or
> > > lockf/fcntl. Do you still need me to try and write a STC, or at least
> > > test how perl behaves with your first patch?
> >
> > I checked in a patch which hopefully solves both problems. The lock.pl
> > testcase now works, at least. Can you test if the original scenario
> > works now as well? If not, I'd really need another testcase.
>
> Nope; with snapshot 20090418 I'm still seeing the 'make -j2' failures. I'll
Too bad.
> see if I can come up with a C program that can be run as two processes to
> simulate the failure as a simpler test case. I did verify via strace that
> perl
> is using flock(fd, LOCK_EX).
Yes, I saw this already while debugging the lock.pl testcase.
Whatever it is, it must be more complicated than this:
#include <stdio.h>
#include <sys/file.h>
int
main ()
{
int fd = open ("flock.c", O_RDWR);
if (fd >= 0)
{
printf ("About to call flock\n");
if (!flock (fd, LOCK_EX))
{
printf ("flock succeeded. Waiting...\n");
getchar ();
close (fd);
}
else
perror ("flock");
}
return 0;
}
Because this testcase works fine. I hope it's not trying to do this:
parent opens file
fork
child calls flock()
exit
fork
second child relies on the lock.
because this is exactly the scenario which doesn't work with flock
right now. I have it on my TODO list, but so far I'm at a loss as
to how to implement it.
Corinna
--
Corinna Vinschen Please, send mails regarding Cygwin to
Cygwin Project Co-Leader cygwin AT cygwin DOT com
Red Hat
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
