Re: INQUIRY: Inconsistent gcc compile failures

[fhirsch at tse dot com]

Thanks for the help, this FAQ pointed out a couple of key hardware issues
that I should look at. I've long suspected the memory was at fault, but
this helps me to possibly fix the issue in a couple of ways without
replacing memory. Additionally, the FAQ informs of intermittant problems
with Cyrix 6x86 processors, which may be another possible issue.

One last question related to this, I've done a few compiles today as I am
installing a new version of the O/S (Mandrake Linux 7.2), I've found that
with a few code modules that a compile will fail several times in a row on
the same piece, and then I can eventually smash my way through on the third
or fourth attempt. The make might fail somewhere else in the build (same
reason), but usually does fine. However, if I pull "make" into "gdb" and
type "run", it always works fine. Might this symptom indicate better where
the problem might be?

Fred





Joe Buck <jbuck@racerx.synopsys.com> on 12/21/2000 02:35:51 PM

To:   pedwards@disaster.jaj.com (Phil Edwards)
cc:   fhirsch@tse.com, gcc@gcc.gnu.org
Subject:  Re: INQUIRY: Inconsistent gcc compile failures


On Thu, Dec 21, 2000 at 02:19:00PM -0500, fhirsch@tse.com wrote:
> > Errors do not
> > always occur in the same module of code. Sometimes the code element
will
> > compile fine, and fail on a subsequent element. Eventually the code
will
> > compile all the way through in most cases.
> >
> > Enclosed is a sample report of the error (received when compiling Perl
DBD
> > mysql module). I believe I had one of these in gdb once, but no
backtrace
> > was actually available. gdb reported a normal exit state.
> >
> > cc: Internal compiler error: program cc1 got fatal signal 11

Phil Edwards writes:
> This is almost always a sign of hardware failure.  There is no source of
> randomness in the compiler itself that would sometimes create good code
and
> sometimes fail when given the same input.  I suspect your RAM is going
bad.
>
> Somebody has a really good webpage somewhere about this, but I don't
> recall where.

The web page is

http://www.bitwizard.nl/sig11/

Generally if the problem is in the compiler, the crash will be repeatable,
you will always see the message or never see it on a particular input.
If it is intermittent and random, then it is probably bad memory.