Date: Wed, 11 Sep 2002 18:58:39 +0200
From: Thomas Esser <[EMAIL PROTECTED]>
> I found that the following ditty which in earlier incarnations of my
> system just died almost immediately with a Segmentation Fault, will
> under current 2.4 Linux kernels under, say, RedHat's (null) beta,
> cause the machine to more or less freeze:
>
> tex '\def~{\if~}~'
My system (linux with 2.4.19 kernel) get slower, but tex is stopped with
! TeX capacity exceeded, sorry [main memory size=2500001].
after a few seconds.
Which is very much harmless, no issue with that. For small settings
of the stack ulimit, this might also have segfaulted fast, what I
experienced previously and with which I don't have a problem, either.
But if you write \number instead of \if you should get to see the
unmitigated original problem, as I checked now. Please pass this on
to the web2c maintainer, I had just reported this from memory, and \if
seems to have another brake built in. With \number, on _this_ system
(not the one I am writing this mail on, logged in there remotely), the
Linux OOM killer got active and pulled the plug on the TeX process --
good decision. But it might also have decided to pull the plug on the
X Server instead. This will not cause any problem unless the stack
size ulimit is rather large or unlimited, which unfortunately seems to
be the default on my Laptop (newest RedHat).
> That's bad. So perhaps one should let TeX automatically set a stack
> size limit, roughly what
> ulimit -s 1024 or so would do.
There are always ways to do weired things and to bring the system down
unless you set up limits that make the system unusable. I don't think
that adding such a system dependency is worth the trouble.
Well, I would consider it worth the trouble in cases where this may
affect system stability.
Anyway, the maintainer of web2c should decide that (since I aim to
follow him as closely as possible), so I'll forward him your
request.
I agree. Please pass on this letter as well, if it is not too much
trouble and if he is not reading the pretest list.
--
David Kastrup Phone: +49-234-32-25570
Email: [EMAIL PROTECTED] Fax: +49-234-32-14209
Institut für Neuroinformatik, Universitätsstr. 150, 44780 Bochum, Germany
