Martin Koegler <e9925248@stud4.tuwien.ac.at>; Date: Sun, 12 Jan 2003 14:48:03 UTC; 

I use xfs in my font-path for Xtightvnc. If i restart xfs (/etc/init.d/xfs reload), the Xtightvnc-process starts using all available cpu-time (Idle 0%)

To reproduce run (each command at a different shell):
xfs -port 8000 
Xtightvnc :30 -fp tcp/localhost:8000
Then send Signal 1 to xfs (kill -1).
A top will show after a few seconds, that Xtightvnc uses all of the free cpu-time.


The problem is, that an reconnect is started, but the reconnect is never done.
Xvnc tries to reopen the connection, but the reconnect-code is never reached.
Before the select, it notices that an reconnect is pending and starts immediately the next try.

A simple solution is to insert the reconnect-hook at another position:
In Xvnc/lib/font/fc/fserve.c needs to be changed:
--- fserve.c~	Tue Mar 19 11:29:49 2002
+++ fserve.c	Sun Jan 12 14:41:37 2003
@@ -1232,6 +1232,9 @@
     int         err;
     fsGenericReply rep;
 
+    if (awaiting_reconnect) {
+	_fs_try_reconnect();
+    }
     /* see if there's any data to be read */
 
     /* 

Maybe the reconnect-code at the end of the function could be removed, but it causes no problem.


Xrealvnc has the same problem and contains the same file. So the patch should also work for it, but i have not tried it there.