Subject: Re: cc1plus lockup? And kernel panic
To: Andrew Doran <ad@NetBSD.org>
From: Bernhard Moellemann <zza@serpens.de>
List: current-users
Date: 08/23/2007 13:13:00
Hi!

You asked:

> I've noticed gcc doing this when it runs out of space in TMPDIR (usually
> /var/tmp), although it will loop trying to open a file getting ENOSPC.

There were no messages on the console or in the syslog saying filesystem full.
I did not look for the amount RAM remaining, I will have a look for it in
the future.

The $WRKOBJDIR is in /usr and had enough space for sure. /tmp is an tmpfs
and may have run out of space. Is gcc writing hardwired to /var/tmp also?

> Was the machine under memory pressure at the time? Was it paging?

Sorry, cannot say this. But it has 1 GB RAM and no swap. (The disk layout
is from old times and the reserved 256 MB swap space is useless when
having 1 GB RAM.)

But no X was running, and I usually do not hit the 1 GB even with X and
running firefox.

Ah, here are the last lines of top in the screen history 8-)

load averages:  1.00,  0.85,  0.49                  up 0 days,  2:50   11:59:59
33 processes:  31 sleeping, 2 on processor
CPU0 states:  3.5% user,  0.0% nice, 54.2% system,  0.0% interrupt, 42.3% idle
CPU1 states:  3.5% user,  0.0% nice, 38.8% system,  0.0% interrupt, 57.7% idle
Memory: 68M Act, 12K Inact, 1108K Wired, 7560K Exec, 39M File, 892M Free
Swap:
Connection to arcanic closed.

So no memory pressure as far as I can see.

    Bernhard  //
            \X/