Re: [AOLSERVER] crashes related to insufficient stack space

[Jim Davidson]

In a message dated 11/1/02 6:33:35 PM, [EMAIL PROTECTED] writes:

Except in my recent experiences the ONLY clearly insufficient stack space
errors I have received come back to one of two things... 1)PHP and 2) the
glibc upgrade in redhat doing something funky to the memory requirements
causing DNS lookups (al la hosts.allow in nsperm or gethostbyaddr)...

This tells me that writing bad code in TCL isn't the only way, or even the
most prolific way that happens in the "real world" that I have seen.. thus
it would be a good idea to at least attempt to signal that the stack space
is insufficient..� heck, even a "high water mark" procedure that you can
turn on that could be accessed the same way other stuff is in
nstelemetry.adp would be a benifit.. because then we could just keep an eye
on that value as a guideline when to consider upping the value.

Heck, even Borland Pascal from the dos age tried (emphasis on tried) to
handle stack issues gracefully... so why not aolserver?

Well, I think the answer to "why not aolserver" is, like you said, the overflow can occur in any code so aolserver can't protect against all cases.  I think fiddling around with guard sizes may help out, at least generating a better core dump (note core dumps don't work very well on Linux).  Newer pthread books talk about thread stacks and guard size and I found this on the Solaris site:

http://docs.sun.com/db/doc/805-5080/6j4q7emgj?a=view#attrib-22

-Jim