stack overflow limit

stack overflow limit

[Drew Adams]

Not a request or a suggestion.  I'm just wondering about the stack size limit
(e.g. for regexp search, search.c).

Would it make sense to make it any bigger, given that machines nowadays are more
powerful and have more memory, or do you consider that pretty much all such
stack overflows (e.g. for regexp matching) are just due to poorly chosen
regexps?

Again, just wondering if it might be appropriate (time) to bump the stack limit.
(No, I don't have a particular use case or regexp in mind.)

Re: stack overflow limit

[Stefan Monnier]

> Not a request or a suggestion.  I'm just wondering about the stack size limit
> (e.g. for regexp search, search.c).
> Would it make sense to make it any bigger, given that machines
> nowadays are more powerful and have more memory, or do you consider
> that pretty much all such stack overflows (e.g. for regexp matching)
> are just due to poorly chosen regexps?

IIRC this depends on the OS stack size, which hasn't grown nearly as
fast as hardware resources.


        Stefan

Re: stack overflow limit

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Date: Wed, 04 May 2011 15:51:22 -0300
> Cc: emacs-devel@gnu.org
> 
> > Not a request or a suggestion.  I'm just wondering about the stack size limit
> > (e.g. for regexp search, search.c).
> > Would it make sense to make it any bigger, given that machines
> > nowadays are more powerful and have more memory, or do you consider
> > that pretty much all such stack overflows (e.g. for regexp matching)
> > are just due to poorly chosen regexps?
> 
> IIRC this depends on the OS stack size, which hasn't grown nearly as
> fast as hardware resources.

How much stack does Emacs have on a typical Unix or GNU machine these
days?

The value of re_max_failures we use now needs 4MB of stack on a 32-but
machine, twice as much on a 64-bit machine.  We also need stack space
for GC.  The result should be compared to what Emacs has and what it
can have.

`ulimit' seems to indicate we get 8MB of stack on an x86_64 GNU/Linux
system.  On Windows, we tell the linker to reserve 8MB as well (but
the Windows build is a 32-bit build.)

Ultimately, the question is: should we increase the value of
re_max_failures?

Re: stack overflow limit

[Andy Wingo]

On Wed 04 May 2011 22:43, Eli Zaretskii <eliz@gnu.org> writes:

> `ulimit' seems to indicate we get 8MB of stack on an x86_64 GNU/Linux
> system.

That varies per-system.  Some distributions do not set a stack limit via
`ulimit'.

Andy
-- 
http://wingolog.org/