Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!usc!wuarchive!texbell!texsun!convex!convex.COM
From: tchrist@convex.COM (Tom Christiansen)
Newsgroups: comp.lang.perl
Subject: Re: itimers
Message-ID: <-1773786417@convex.convex.com>
Date: 15 Feb 90 16:28:14 GMT
References: <5334@convex.convex.com> <7077@jpl-devvax.JPL.NASA.GOV>
Sender: news@convex.com
Reply-To: tchrist@convex.COM (Tom Christiansen)
Organization: CONVEX Software Development, Richardson, TX
Lines: 35

YARG, I was using 2 timeval's instead of 2 itimerval's.  Chastisement
accepted.

The routines work now with one minor change to compensate for
what *seems* to be a perl bug regarding return values from syscall.
The last statement in setitimer() had to be changed from :

    !syscall(&SYS_setitimer, $which, $_[0], $_[1]);
to:
    syscall(&SYS_setitimer, $which, $_[0], $_[1]) != -1;

I found that in the test case posted, syscall was returning 1, 3, and 1, 
although errno was OK and everything worked.  This is true on both
a Sun and a Convex.  I coded up the same thing in C using the 
indirect syscall() interface, and it was just fine, returning 0's 
not 1, 3, and 1.  Looking at do_syscall() in perl/doarg.c, I put
this debugging line in perl itself right after the call:

	printf("perl: syscall returned %d\n", retval);

and confirmed that the syscall()s really are working fine.  BUT
by the time the interpreted code gets a hold of it, it's been 
changed into 1, 3, and 1.  

I can't help wondering whether my array-sensitive &itimer function
is messing with perl's stack (the one it keeps of my code, not
its own stack.)


--tom
--

    Tom Christiansen                       {uunet,uiucdcs,sun}!convex!tchrist 
    Convex Computer Corporation                            tchrist@convex.COM
		 "EMACS belongs in <sys/errno.h>: Editor too big!"