Re: Subtle bug with GCC optimization?

[Marius Vollmer <mvo at zagadka dot ping dot de>]

Gary Houston <ghouston@arglist.com> writes:

> > From: Marius Vollmer <mvo@zagadka.ping.de>
> > Date: 16 Aug 2000 19:55:00 +0200
> > 
> > Gary Houston <ghouston@arglist.com> writes:
> > 
> > > It looks like the same GCC bug we ran into a few months ago: try adding
> > > -fno-gcse.
> > 
> > Yes, indeed.  Using -fno-gcse produces correct code.  What did you do
> > to work around your instance of the bug?
> 
> It was the "Crash in scm_boot_guile if stdin isn't open" problem: you
> helped find a fix for it.  The gcse bug caused a SEGV instead of
> abort when the problem occured, so the work-around was to make the
> problem not occur.

Ah, yes, I remember.  I have now fixed my problem by inserting this

    #ifdef __GNUC__
      /* GCC 2.95.2 has a bug in its optimizer that makes it generate
         incorrect code sometimes.  This barrier stops it from being too
         clever. */
      asm volatile ("" : "=g" (winds));
    #endif

at a strategic place in scm_ithrow.  I don't need to turn off gcse and
the rest of Guile is unaffected.  The code for scm_ithrow seems to be
just as optimal, but I don't really care.