On Sun, 2003-01-26 at 10:40, Tracy S. Ruggles wrote:
> My webware servers were dying, too, without a mention of it in the
> error logs, and only when I wasn't watching. It turned out that I had
> a stack size limit, too, of 24K and my server was slowly growing in
> size (I have a bug somewhere that I haven't found yet). And, if my
> pages ever raised an exception when it was over the limit, I would get
> a segmentation fault because webware would try to do its pretty
> traceback page, but the gc module (garbage collection) would just die
> with an Out of Memory error. I was finally able to catch the server
> doing it and see the traceback and also using gdb was able to see that
> the gc module caused the seg fault.
Hmm... this brings up the note in Future.html about making a backup
error handler, i.e., a small-as-possible handler that catches any errors
from ExceptionHandler. That might be a good thing to get in 0.8 if we
can, since it's important to debugging otherwise mysterious problems.
I suppose that it might just be enough to put a try:except: around the
call to the ErrorHandler, and then write the exception report to some
sort of error log. If it's out of memory even that might not work, but
maybe it would, I don't know.
Ian Bicking Colorstudy Web Development
PGP: gpg --keyserver pgp.mit.edu --recv-keys 0x9B9E28B7
4869 N Talman Ave, Chicago, IL 60625 / (773) 275-7241