Chris Bartram wrote:

> We've had (and apparently still have) a problem on our MPE/iX 5.0
> system with tcp connection resources not being released fully. A while
> back someone at HP gave me the following debug script [snip]

This was fixed in NSTED54 on 5.0.  There was a control buffer pool
problem in the 5.5 beta but was fixed in X.55.20 and incorporated into
the general release.

All problems appear to be cgi related, so I would change focus to the
always-elusive "stdin/stdout <-> socket" interface mechanics.  The cgi
script will be simply "writing to stdout" with probably little error
checking, so there is probably some not-quite-so-robust piece of the
socket<>stdout "glue" at fault here and you'd need a definitive answer
from the lab.

While they fix this, perhaps then they can improve on the
MPE-friendliness of the socket<>stdout interface so you can have
non-posix, non-C cgi programs (but I'd settle for the current problem to
be fixed for now).

Have any of you checked httpd/OMI $stdlist for errors/stack
traces/tombstones/etc?

Jeff Kell <[log in to unmask]>