This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Bootstrap is STILL broken...
- From: Dale Johannesen <dalej at apple dot com>
- To: Andrew Pinski <pinskia at physics dot uc dot edu>
- Cc: Caroline Tice <ctice at apple dot com>, "gcc-patches at gcc dot gnu dot org Patches" <gcc-patches at gcc dot gnu dot org>, David Edelsohn <dje at watson dot ibm dot com>, Geoff Keating <geoffk at geoffk dot org>, Dale Johannesen <dalej at apple dot com>, Zack Weinberg <zack at codesourcery dot com>, Kelley Cook <kelley dot cook at sbcglobal dot net>
- Date: Wed, 31 Mar 2004 16:38:05 -0800
- Subject: Re: Bootstrap is STILL broken...
- References: <20040327133243.1538.qmail@web41505.mail.yahoo.com> <BB065F00-8004-11D8-9789-000393A6D2F2@physics.uc.edu> <860B2C28-82A4-11D8-A608-000393BB90B6@apple.com> <878yhhpd9f.fsf@codesourcery.com> <200403311546.i2VFkOT26240@makai.watson.ibm.com> <13F4C11E-836F-11D8-8F7D-000A95D7CD40@apple.com> <839E484E-8370-11D8-AFAD-000393A6D2F2@physics.uc.edu>
On Mar 31, 2004, at 4:07 PM, Andrew Pinski wrote:
On Mar 31, 2004, at 18:56, Dale Johannesen wrote:
(I am not a Darwin maintainer, but OK. Perhaps I should be at this
point?)
I do not think this is the right general approach to the problem, as
I said, but
in the interests of getting bootstrap going I'll go along with it.
These routines are the same ones that have been used in Apple's gcc
for years,
so they're technically OK. I've verified that they were written at
Apple and we
are willing to contribute them, so there's no legal problem.
In reality, it is the right approach as you can now link code
generated by Apple's
compiler and the FSF compiler which you could not do always before.
You can if you use the Apple compiler to do the link.
Changing libiberty to be bootstrap able will not fix that problem,
just the bootstrap problem.
I agree that this is a good idea for other reasons, but the general
bootstrap problem will still
be there waiting for somebody else to trip over it. If somebody puts
long double arithmetic in
libiberty, for example, I bet the same problem will occur on several
hosts. (There is long double
code there now, but all it does is copy them around, which doesn't
usually require a library call.)
If it is impractical to rebuild libiberty at each stage, why not use
the installed system compiler to
do all the links?