new TCL vs. current-6

[Doug White]

On Wed, 8 Dec 2004, Kris Kennaway wrote:

> On Wed, Dec 08, 2004 at 12:05:38PM -0500, Mikhail Teterin wrote:
> > Three of my TCL-based ports fail now on amd64 running current-6. All failures
> > occur, when an attempt is made to use the TCL-interpreter -- either to run
> > the port's self-tests, or to generate the manual pages. The failures are
> > either "Floating point exceptions" or "Segmentation faults" and appear to
> > only happen on amd64 (may be, on ia64 as well):
> >
> > http://people.freebsd.org/~fenner/errorlogs/mi%40aldan.algebra.com.html
> >
> > Can anyone confirm being able to use freshly built TCL on amd64 _at all_?
> >
> > My ports did not change in months -- what could be wrong with TCL and/or
> > amd64?
>
> Peter speculated a problem when the world is built with -O2 (with or
> without -fno-strict-aliasing).  I haven't been able to confirm this by
> testing with a world built with -O, because I only have one working
> amd64 package builder.

A quick Makefile look and gdb run finds it taking SIGFPE in strtod().
The tcl build is also very odd since it seems to apply CFLAGS twice like:

cc $CFLAGS -O $CFLAGS ...

I haven't gone looking for why, just noticed it during the build.

This is on a machine built with -O2.  I'll kick off an -O build now and
see what happens.

-- 
Doug White                    |  FreeBSD: The Power to Serve
dwhite at gumbysoft.com          |  www.FreeBSD.org