erratic msgrcv() behavior

[Charlie Kester]

I am having a problem with msgrcv() under SCO System V Xenix on a '286.
Most of the time this system call works just fine, but every now and then
it ails with errno set to EINVAL.  The manual says this will be the errno
when (a) the queue id is invalid (b) an attempt is made to receive into
read-only shared memory (c) the message size requested is less than zero
Only the first can be the case in my code, since I'm not using shared memory
and the message size is a #defined constant.  So the tion is: what can
cause a previously-valid queue id to become invalid?   

I already know that if I use the out-of-the-box defaults for MSGMAP, etc.,
a queue will "disappear" and msgrcv will fail when two or more messages are
placed on the queue before the call to msgrcv. (When this occurs, ipcs -qo
shows 0 messages but NOT 0 bytes on the queue. And ipcrm doesn't reenable
the queue -- the system has to be rebooted.)  I have corrected the tunable
parameters to avoid this fault, but I still lose the queue id every now and
then.

Since there is this undocumented failure mode, I have to think there are others.

Please reply by email.  


-- 
Charlie Kester      |  cgk at chinet.chi.il.us   | "Wretches, utter wretches!
1503 Larson Street  |                         |  keep your hands from beans!"
Sycamore IL 60178   |                         |       (Empedocles)