gcc/libm floating-point bug?

[Ian Freislich]

Morten Rodal wrote:
> On Wed, May 21, 2003 at 06:10:14PM -0700, David O'Brien wrote:
> > Honest question of you -- I'll assume you're subscribed to
> > freebsd-current at .  How have you missed all the warnings from myself and
> > others not to trust the -march=3Dpentium4 optimizations?  I honestly want
> > to know so we can figure out a better way of getting the word out.
> >=20
> 
> I am not going to discuss the removal of the -march=3Dpentium4, but a
> warning would be nice.  Either in the form of the release notes or in
> the make.conf, preferrable both.
> 
> As for the "missed all warnings [...]" I think this just proves that
> it is necessary to document things like this.  Also see the post about
> a lot of new people trying out 5.0 for the first time and the troubles
> they are having with getting help.  Search for

It seems the following comment pertaining to CPUTYPE was removed
from make.conf:

# If you experience any problems after setting this flag, please unset
# it again before submitting a bug report or attempting to modify code.
# It may be that certain types of software will become unstable after being
# compiled with processor-specific (or higher - see below) optimization flags.
# If in doubt, do not set CPUTYPE or CFLAGS to non-default values.

It was enough of a warning to make me hesitate to change CPUTYPE
and look there first when problems cropped up.  Perhaps it needs
to go back.

Ian