Re: GCC 3.0.2 and BSD Make

[Loren James Rittle <rittle at latour dot rsch dot comm dot mot dot com>]

> Which parts need GNU make?  Definitely libgcj does, perhaps libstdc++, I
> dont know.  But if --enable-languages=c still works with BSD make, that
> may be good enough for some people.

Since the bootstrap is failing in stage2 within the gcc/fixinc
directory with BSD make, I think even a configuration with
--enable-languages=c is currently broken.  However, I haven't tried a
bootstrap with that constraint.

Unless it broke again while we couldn't test it for other reasons,
libstdc++-v3 no longer requires GNU make.  Unless I misremember, at
the request of the GCC SC itself (or at least a member of said SC), we
fixed all issues with the build process for that language library.

Perhaps, as GNU ada maintainers did, libjava maintainers should place
some documentation on this matter in gcc/doc/install.texi or else
people will not know of your constraint until far into the bootstrap
when it fails with a bizarre error message.

> I think the idea was to avoid bug reports from BSD users who configure GCC
> with all languages enabled.  The truth is a little more complicated, as
> usual.

Yes, I thought it was best to convey what actually works in a release
rather than what we strive to obtain in gcc releases.

The truth of the matter is that we received over 20 PRs in the GCC
GNATS from people that tried to configure 3.0.1 in-source (nothing
related to BSD make).  These people received failures far into the
bootstrap process.  Who knows how many other people encounter the
issue and gave up or otherwise wasted time debugging a known problem?
I was personally responsible for breaking that configuration and/or
not checking someone else's configuration changes in libstdc++-v3
closely enough so I felt particularly bad about the breakage.  Thus,
in addition to all other stated reasons, I wanted to avoid useless PRs
in the GCC GNATS and personal guilt about wasting people's time. ;-)

Regards,
Loren