RE: release candidate 5 available--Solaris now passes all tests

[DeBerry, Marshall]

I'm using gcc 3.4.2, running under Solaris 5.9 Generic_117171-09 sun4u
sparc SUNW,Sun-Fire-V240.

Yes, memory bugs are some of the hardest to figure out, particulary if
they "appear and disappear" depending on small changes--like this
appears to be.  Any other ideas you or John may have on trying to debug
this I'd be happy to try out.

 Marshall 

-----Original Message-----
From: address@hidden [mailto:address@hidden 
Sent: Tuesday, August 02, 2005 2:23 PM
To: DeBerry, Marshall
Cc: address@hidden; address@hidden
Subject: Re: release candidate 5 available--Solaris now passes all tests


"DeBerry, Marshall" <address@hidden> writes:

> Jut a slight update.  I reran the configure without the the -O0 flag, 
> but with -g, and had no problem with the check test running on 
> examine.sh.  Just running configure with no CFLAGS set, causes the 
> Abort to happen.
>
> Generally, when quirkies like that happen, doesn't it point to some 
> sort of memory leak somewhere?

It may indicate a memory-related bug, simply because those are some of
the most common hard-to-find bugs in practice.  It is probably not a
memory leak, because a memory leak just means that memory is wasted.

Strangely, I couldn't reproduce this bug on Solaris 8 using any of GCC
3.3, 4.0, or the Sun C compiler.  (Actually the Sun compiler gives some
warnings and errors we should fix, but they're not related to this
problem.)

What version of GCC are you using?
-- 
"Long noun chains don't automatically imply security."
--Bruce Schneier