[3478] in testers
Re: sgi_62/8.2 login crashes
daemon@ATHENA.MIT.EDU (Greg Hudson)
Sat May 23 16:50:33 1998
Date: Sat, 23 May 1998 16:50:24 -0400
From: Greg Hudson <ghudson@MIT.EDU>
To: Greg Hudson <ghudson@MIT.EDU>
Cc: Robert A Basch <rbasch@MIT.EDU>, testers@MIT.EDU
In-Reply-To: "[3477] in testers"
> (Incidentally, copying binaries to /var/tmp and stripping them there
> doesn't help, so this is probably not an AFS issue.)
That was kind of a dumb conclusion. It could easily be an AFS problem
with ld writing out the executables, especially since it cropped up
when I built the packs into an AFS build tree and not before.
I tried building zephyr into AFS to see if I could reproduce the
problem on zaway, but didn't see the problem, so it isn't all that
easy to reproduce. I also couldn't find any relevant defects in the
AFS README files, but that isn't too surprising since I think we're
using the most recent patch release. I'll send mail to afshelp and
see if they know anything about this for the forthcoming patch
release. In the meantime, I guess I'll rebuild the packs using a
local build directory.
> This problem does not appear to be reliably reproduceable at the
> compiler level (e.g. building zephyr doesn't yield an unstrippable
> zaway binary). I can only conclude that the IRIX 6.2 cc is not
> deterministic. I'm going to switch to building the release with gcc
> now.
Well, that was a nice theory, but it's not trivial to do. (It's also
unclear that cc is really nondeterministic, as opposed to AFS.)
