Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!cbosgd!ucbvax!hplabs!hplabsc!garfinke
From: garfinke@hplabsc.UUCP
Newsgroups: comp.windows.x
Subject: Re: Applications Hanging? Does XLine work?
Message-ID: <69500008@hplabsc.UUCP>
Date: Thu, 4-Jun-87 15:43:00 EDT
Article-I.D.: hplabsc.69500008
Posted: Thu Jun  4 15:43:00 1987
Date-Received: Sat, 6-Jun-87 09:40:51 EDT
References: <8706040216.AA08131@bu-cs.BU.EDU>
Organization: Hewlett-Packard Laboratories - Palo Alto, CA
Lines: 27

# Subject: Applications Hanging? Does XLine work?

> I get on another terminal and kill -QUIT (just before the core dump
> the window update finishes as I expected, like the buffers finally get
> flushed.) Inspecting the core file it seems to be in write() called
> from _XFlush() called from XNextEvent().

> Anyone know what's causing this? I'm looking but it all seems ok, the
> only thing that comes to mind is that the buffer is full and we've
> entered some sort of deadlock with the server, but that's just a wild
> guess.

I've seen this happen before and the cause is that the client is writing
to the server and blocks when the socket is full.  At the same time, the
server is sending packets back to the client (in this case maybe an expose
event) and the since the client is not consuming these events, the server
blocks when the socket is full. Poof, deadlock.  When the process is killed,
X wakes up from the write and processes the remaining requests.

The key here is not to keep on doing operetions which cause events to 
be generated without consuming them.  Maybe setting a timer before doing
an XFlush will help.  I suppose it is the server responsibility to not
hang the system, but since thats not how they're written, you'll have to 
put some type of workaround in your client.


Dan Garfinkel, HP Labs