problems with select ONLY under Eunice

[Don Libes]

I have some code that works fine under 4.2 & 4.3 but not under
Eunice (the 4.3 version).  The problem is with select().  (Yes, I
am correctly handling the difference in fd_set properly.)

The first problem is that it doesn't detect pending connects
properly.  After the listener accepts a connect request, select
insists the old connect request is still pending.

Wollongong has confirmed that select has this bug, and it is not to
be fixed in the near future.  They suggested using the raw QIO
calls as the way to work around it.

Has anyone done this?  It would save me a lot of pain if I don't
have to learn about TWG's lovely QIOs.

Does anyone have the source to select, which might explain why they
can't do it right?  They refused to send it to me, and said it was
something related to VMS-braindamage.

I have some other questions about IPC programming with Eunice.
Does anyone have a list of hints/tips or some general experience
about the differences between it and raw Berkeley UNIX.

For instance, I notice that running under the debugger, socket()
generates integers much larger than when not under the debugger.
In particular, they don't fit in the fd_set bit mask!!!

Anyway, I have lots of questions like these.  Wollongong always
seems so surprised when I call them - like I'm the first to
complain.  That just can't be, can it?

Don Libes          cme-durer.arpa      ...!uunet!cme-durer!libes