We have lots of programs written using gnustep-base (1.12.0 at the
moment). These are all daemons -- no UI, Foundation only -- running
on Intel-architecture boxes running linux 2.4. There are hundreds of
instances of about twenty different programs on hundreds of machines,
and they'll run fine for months, and then out of the blue one of them
will throw an exception like this one:
Uncaught exception NSInvalidArgumentException, reason: NSException
(instance) does not recognize reason
It's not always -[NSException reason] -- sometimes it's a method in
one of my classes -- but it's always a method that does exist in the
named class. I don't think it's some random memory sprayer, because
if it were, I'd expect other funny behaviors from time to time --
segfaults, argument errors, corrupted output files, whatever, and I
get none of that. It's definitely not a memory leak; process sizes
remain stable for months. The only unexplained error they ever get is
this one.
Not all my programs do this; only two have -- but the problem is so
infrequent it might be that they're all susceptible, and only these
two have been unlucky. The ones that have succumbed are running an
HTTP server that I wrote, and they use multiple threads, one for each
request. Also they use OpenSSL, for HTTPS support. I do provide
locking methods to OpenSSL, which is otherwise not thread safe. But
there's at least one other HTTPS server using the same implementation
that has never died this way, so this might not be relevant.
Has anyone else ever run into anything like this?
- lc
_______________________________________________
Discuss-gnustep mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/discuss-gnustep
