Re: Building on RH 8.0 using UML

On Mon, 24 Mar 2003, Robert Spier wrote:

> > > > I tried a test run of 5.8.x on it and all combinations failed  on 
> > > > op/alarm.t.
> > > This is not too surprising.  UML has to do some weird things with
> > > timers to function.  Can you isolate which test is failing?
> > Yes, one of the runs said:
> >  Tests start here:
> > PERLIO = stdio  # Failed at op/alarm.t line 31
> > #      got ''
> > # expected 'ALARM!
> > # '
> > # Failed at op/alarm.t line 32
> 
> What happens if you run op/alarm.t manually?
> 
> ./perl -Ilib t/op/alarm.t

[mike@rh8-uml perl-5.8.x]$ ./perl -Ilib t/op/alarm.t
1..4
not ok 1 - alarm w/$SIG{ALRM} vs inf loop
# Failed at t/op/alarm.t line 31
#      got ''
# expected 'ALARM!
# '
not ok 2 -    right time
# Failed at t/op/alarm.t line 32
ok 3 - alarm w/$SIG{ALRM} vs system()
ok 4 -    right time (waited 3 secs for 3-sec alarm)

 
> The test is doing a busy wait, 1 while (time - $start_time < 6), and I
> guess it's not actually waiting.

It is actually waiting, I added a few debugging prints and it says

my $start_time = time;
print "start at $start_time\n";
eval {
    local $SIG{ALRM} = sub { die "ALARM!\n" };
    alarm 3;

    # perlfunc recommends against using sleep in combination with alarm.
    1 while (time - $start_time < 6);
};
print "out of eval at ", scalar(time), "\n";
alarm 0;
my $diff = time - $start_time;
print "diff is $diff\n";

(good enough for a rough indicator...) and I get 

start at 1048601190
out of eval at 1048601196
diff is 6

So it looks like the alarm isn't killing the eval before the while 
terminates.

I'll try running the smoke test on my "real" RH 8 now that the UML one has 
stopped...

Mike 

-- 
[email protected]                    |           The "`Stok' disclaimers" apply.
http://www.stok.co.uk/~mike/       | GPG PGP Key      1024D/059913DA 
[email protected]                  | Fingerprint      0570 71CD 6790 7C28 3D60
http://www.exegenix.com/           |                  75D2 9EC4 C1C0 0599 13DA