Re: Strange lossage.

[RMS@MIT-AI]

I have in RMS;SUPDUP a supdup that uses windows.
It works, with just one bug: that if you send a supdup window
a connect message while it isn't exposed, and then send it an expose message,
it doesn't succeed in starting up.  However, if you send
it a connect message and then give its parallelism a
run-reason, and then expose it (typing from the terminal)
it does succeed.  Why this should matter, I can't figure out.
Maybe it's because the output process can only read the first
character of the greeting and then has to wait for the exposure.
Note that it works to send the connect message while the window
is already exposed.  As a result, in practice the only thing
that doesn't work is the ordinary (SUPDUP) function
which has to send the connect before the expose since the
window it is done in is going under when the expose is done.

When it loses, it's very strange.  The SUPDUP-SEND-TTY-VARIABLES
function does get called, but the telser never sees them arrive.
However, the same number of packets get sent out to the ITS
in either case, I think I saw from the lisp machine peek.

I suspected that it might be the fact that exposing the window
redisplays its box and label, so I turned that completely off.
It didn't change anything.

It may be relevant that I once got into a state where
(HOST-DATA 202) (that is AI) waited forever.
I think that compiling and loading RMS;SUPDUP and LMIO;KBD
and doing (SUPDUP) will reproduce this.  (That's 2026, not 202).