Strange core dump of jade when building /usr/doc

[Simon L. Nielsen]

On 2003.09.04 18:00:35 -0500, Manuel Rabade - MiG wrote:
> On Thu, Sep 04, 2003 at 10:55:11PM +0200, Simon L. Nielsen wrote:
> > On 2003.09.02 17:27:08 -0500, Manuel Rabade - MiG wrote:
> > > Hi, i have updated /usr/doc (with only the documents in english and spanish) and
> > > when i do a 'make DOC_LANG=en_US.ISO8859-1 FORMATS=html' jade crashes in
> > > /usr/doc/en_US.ISO8859-1/articles/hubs, sometimes as a 'Bus error' and other
> > > times as a 'Segmentation fault'.
> > 
> > I can't remeber that I have ever experienced jade coredumping... The
> > fact that it dies with different errors make me suspect hardware error.
> > It could also be non-standard CFLAGS in make.conf.
> > 
> > I don't really have any other ideas to why it dies.
> 
> Humm .. i think that it isn't a hardware error, is my desktop machine and i can
> compile and run everything fine except *that* :'(.
> 
> I "fix" it yesterday installing jade from a package (jade-1.2.1_1), so i think
> that my CFLAGS ('CPUTYPE=p4' and 'CFLAGS=-O2 -pipe') are breaking jade, very
> strange if it's the problem because that flags are not too much agressive.

Actually they are a bit agressive. '-O2' have produced broken code for
varius versions of GCC, and at least one version of gcc 3.x has produced
broken code for CPUTYPE=p4 (though this should be fixed now).  More
optimizations sometimes trigger coding bugs, which lower -O levels hide.

From make.conf(5) :

     CFLAGS   (str) Controls the compiler setting when compiling C code.
              Optimization levels above -O (-O2, ...) are not supported.

In short, while higher -O levels might work, they are rarely worth all
the problems they create.

-- 
Simon L. Nielsen
FreeBSD Documentation Team
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-doc/attachments/20030905/5147fa54/attachment.sig>