GCC version "consistency" (was Re: making and running Swarm after GCC upgrade

[Paul E. Johnson]

Alex Lancaster wrote:
> 
> >>>>> "JB" == John B Barnett, Ph D <address@hidden> writes:
> 
> JB> Alex I noted that it was suggested that the newest version of gcc
> JB> (2.95) be used for compiling Swarm apps.  I downloaded gcc-2.95
> JB> and using the directions on the web, configured, built and
> JB> installed the new version on my machine with RH6.0.  The whole
> JB> process seemed to go OK, i.e., no obvious messages to the
> JB> contrary.  To test the installation, I "made" the SimpleCBug which
> JB> also seemed to go fine, but when I went to run "bug", it gave me
> JB> the "Segmentation fault (core dumped)' error message.  I also had
> JB> a previously compiled version of SimpleObjCBug and when I tried to
> JB> run it, it gave the same error message.  Any idea where I went
> JB> astray?  John
> 
> I'd need a gdb backtrace at the very least to diagnose it (seg faults
> can result from one of many sources).  It's quite common that an app
> (a Swarm app or otherwise) will build, but fail at runtime.  Try this
> (`bt' is short for `backtrace').

Perhaps we could have a clarification on recommendations for changing
the compiler.  

The last time this came up, I think it was Marcus or perhaps Glen who
said that when you change the compiler, it might be necessary to
recompile lots of packages, as well as swarm.  THe whole "toolchain"
might be affected.  I'm certain that, when tclobjc was a separate
package, that it was one that had to be redone.  Maybe libffi as well. 
And I have a faint memory of some problems in a version of
blt8.0unofficial that were solved by recompiling that after a compiler
change.   

After one experience like that, I  got conservative and vowed that I'd
never swap out the c library or the compiler unless I had a very good
reason.  Of course, that was complicated at the time by the separation
of gcc (needed to compile kernel) and egcs (used for swarm), so maybe it
is not such a big deal anymore.  

But if the troubles described in this note are any indication, it seems
to me we ought to fiercely advise people to leave gcc/egcs alone, or
else map out all of the implications of changing the compiler to warn
them.

-- 
Paul E. Johnson                         email: address@hidden
Dept. of Political Science              http://lark.cc.ukans.edu/~pauljohn
University of Kansas                    Office: (785) 864-9086
Lawrence, Kansas 66045                  FAX: (785) 864-5700

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.