On Sun, Dec 26, 2021 at 09:56:32AM +0000, Visa Hankala wrote:
> On Sun, Dec 26, 2021 at 09:34:30AM +0100, Anton Lindqvist wrote:
> > On Sat, Dec 25, 2021 at 04:27:19AM -0700, Visa Hankala wrote:
> > > CVSROOT:  /cvs
> > > Module name:      src
> > > Changes by:       [email protected]    2021/12/25 04:27:19
> > > 
> > > Modified files:
> > >   regress/sys/kern/poll: poll_iocond.c 
> > > 
> > > Log message:
> > > Adjust pty case for kqueue-based poll(2)
> > > 
> > > Unlike the old code, kqueue-based poll(2) sets POLLIN when the other
> > > end of a pty has been closed. Calling read(2) on such a pty returns
> > > zero anyway. FreeBSD sets POLLIN too.
> > 
> > Seeing a consistent failure here, is that expected?
> > 
> > > sys/kern/poll:
> > Exit: 1
> > Duration: 00:00:03
> > Log: 104-sys-kern-poll.log
> > 
> > ==== run-regress-poll_iocond-pty ====
> > ./poll_iocond pty
> > assertion "pfd[0].revents == (POLLIN | POLLHUP)" failed: file 
> > "/home/src/regress/sys/kern/poll/poll_iocond.c", line 807, function 
> > "proc_parent"
> > *** Signal SIGABRT in . (Makefile:15 'run-regress-poll_iocond-pty')
> > FAILED
> 
> This is expected with the old poll code. As the commit message says,
> there will be a change in behaviour, unless the change turns out bad
> or someone objects.

Thanks for the explanation. Consider using REGRESS_EXPECTED_FAILURES as
having broken tests enabled makes it harder to track down actual
regressions.

I'm also seeing an occasional failure:

==== run-regress-poll_iocond-socket-tcp ====
./poll_iocond socket-tcp
assertion "ret == 2" failed: file 
"/home/src/regress/sys/kern/poll/poll_iocond.c", line 769, function 
"proc_parent"
*** Signal SIGABRT in . (Makefile:15 'run-regress-poll_iocond-socket-tcp')
FAILED

Reply via email to