[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: nasty DCOPServer bug
From:       George Staikos <staikos () kde ! org>
Date:       2001-03-03 0:02:58
[Download RAW message or body]

On Friday 02 March 2001 18:35, Waldo Bastian wrote:
> On Friday 02 March 2001 13:32, George Staikos wrote:
> >    I checked /proc/pid/maps and I know that it is not using the wrong
> > libraries or binaries.  This is pure CVS code, updated within the past 2
> > hours.  I have 4 processes running when it dies.  They are:
> >
> > 22697 kdeinit: Starting up...
> > 22700 kdeinit: dcopserver --nosid
> > 22701 kdeinit: klauncher
> > 22702 kdeinit: klauncher        <--- this one is using 100% cpu
>
> 22701 is the process started to launch klauncher. 22701 is waiting for
> klauncher (Pid 22702) to return from a call to newInstance().
>
> Appearantly 22702 is severly confused.

   Yes it is very confused, and it is dying inside dcopclient.cpp.  I'm 
certain it's a race condition because when I put fprintf() statements in 
various places in dcopclient/server, the place it freezes in moves.  The last 
place I had it narrowed down to was two return; statements.  Yes, "return;".  
It was in dcopclient or server under the ProcessMessage (?) method  and 
static method (or was it just a function?).  Anyhow this was at 3 am and I 
still haven't slept (7pm now).  I gave up because fprintf() is no way to 
debug race conditions, especially when one doesn't know the code one is 
trying to debug.  gdb couldn't seem to resolve more than 1 or 2 symbols even 
after I compiled with -g and explicitly loaded in the files.  *shrug*

-- 

George Staikos

[prev in list] [next in list] [prev in thread] [next in thread]