Hi folks,
is anyone using cmucl on freebsd (5.4/i386) and has experienced the
following?
I'm using cmucl 19a (from ports) and since recently (see below) when I
run "lisp", it segfaults and writes a 512mb corefile.
I know that cmucl does odd things to obtain memory (like mapping all
available address space or somesuch).
I have raised the data segment+stack size limits (MAXDSIZ, MAXSSIZ
etc.) in the kernel from the defaults of 512/64mb to 1024/128mb
because I needed to run some larger programs. These settings now seem
to play havoc with cmucl but I can't figure out why. Does anyone know
of a workaround or how to fix this, short of resetting the limits to
their original values?
mkb.
Matthias Buelow wrote:
> I have raised the data segment+stack size limits (MAXDSIZ, MAXSSIZ
> etc.) in the kernel from the defaults of 512/64mb to 1024/128mb
> because I needed to run some larger programs. These settings now
> seem to play havoc with cmucl but I can't figure out why. Does
> anyone know of a workaround or how to fix this, short of resetting
> the limits to their original values?
This is a "known problem" if I can put it that way. I'm not sure why
it happens.
Are the larger programs you are trying to run lisp programs? If so,
you can just use the "-dynamic-space-size" argument to CMUCL, because
CMUCL mmaps the memory (which doesn't fall under the MAXDSIZ limit).
If they're not lisp programs I don't know what to tell you.
--
Fred Gilham ······@csl.sri.com || Progressive (adj): Value-free;
tolerant; non-judgemental. E.g. traditional archery instruction
methods spent tedious hours teaching the archer to hit a bulls-eye.
Progressive methods achieved better results by telling the student
archer to shoot in the manner he or she found most comfortable, then
calling whatever the arrow hit the bulls-eye.
Fred Gilham <······@snapdragon.csl.sri.com> writes:
>This is a "known problem" if I can put it that way. I'm not sure why
>it happens.
Hmm.. ok. Some googling seems to support the idea that cmucl is
"tuned" at build time for the default values and it'll break when the
actual sizes don't match these...
>Are the larger programs you are trying to run lisp programs? If so,
>you can just use the "-dynamic-space-size" argument to CMUCL, because
>CMUCL mmaps the memory (which doesn't fall under the MAXDSIZ limit).
>If they're not lisp programs I don't know what to tell you.
Nope... I was fumbling around with some larger simulation program
written in Java and some experimenting in other languages.
mkb.
From: Raymond Toy
Subject: Re: cmucl acting up on freebsd after increased kernel limits
Date:
Message-ID: <sxdirz7ke3s.fsf@rtp.ericsson.se>
>>>>> "Matthias" == Matthias Buelow <···@incubus.de> writes:
Matthias> Hi folks,
Matthias> is anyone using cmucl on freebsd (5.4/i386) and has experienced the
Matthias> following?
You'd probably get better answers on the cmucl mailing lists.
Matthias> I'm using cmucl 19a (from ports) and since recently (see below) when I
Matthias> run "lisp", it segfaults and writes a 512mb corefile.
Matthias> I know that cmucl does odd things to obtain memory (like mapping all
Matthias> available address space or somesuch).
Matthias> I have raised the data segment+stack size limits (MAXDSIZ, MAXSSIZ
Matthias> etc.) in the kernel from the defaults of 512/64mb to 1024/128mb
Matthias> because I needed to run some larger programs. These settings now seem
Matthias> to play havoc with cmucl but I can't figure out why. Does anyone know
Matthias> of a workaround or how to fix this, short of resetting the limits to
Matthias> their original values?
I'm not a freebsd user. Perhaps there's a program or /proc filesystem
where you can look to see where shared libs are mapped for something
simple like cat. That might be a clue.
Ray