This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: VAX Ultrix bootstrap failure with gcc-2.96
- To: bkorb at sco dot COM
- Subject: Re: VAX Ultrix bootstrap failure with gcc-2.96
- From: "Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>
- Date: Thu, 15 Jun 2000 16:41:31 -0400 (EDT)
- Cc: dave at hiauly1 dot hia dot nrc dot ca, gcc-patches at gcc dot gnu dot org
> From: Bruce Korb <bkorb@sco.COM>
>
> "Kaveh R. Ghazi" wrote:
> > I would argue against this. When we have one autoconf/configure
> > mechanism for gcc and one for cc, then the cc one tends to bit rot
> > over time. Its just better software engineering to use the same code
> > for both cases unless the benefits far outweigh the negatives.
> >
> > IMHO, I don't see a large benefit here worth teaching gcc about
> > __GETGROUPS_T like we do for __SIZE_TYPE__.
>
> OK. Then I want to stay out of the decision process.
> I confess to being unclear whether we are talking about
> getting the GCC built or compiling end user programs.
We are talking about compiling end user programs. But gcc really
falls into that category too. As soon as we start stage2, actually as
soon as we start building libgcc.a in stage1, we are building with the
fixed headers and must get this right. Though I don't think gcc
source code calls getgroups anywhere as of today, its possible to
create a broken unistd.h which won't allow compiling anything at all
which attempts to include it. (This is what happened last time.)
> Fixinc is only about the latter. If sys-protos.h is all that
> is of concern and it is only used to build GCC, then the
> result of the configure test can be used, as long as
> $BUILD = $HOST. No need to reference $TARGET and
> no need to mess with __GETGROUPS_T either.
sys-protos.h is processed and eventually used by *fixproto/fix-header*
to add prototypes to K&R headers. The fixproto/fix-header process is
analogous and complementary to fixinc (but I'm pretty sure you already
knew that.) So I believe the results of the autoconf test should only
be used when BUILD==TARGET. The trick IMO is to get it to actually
work given the implementation problems I listed against the last
attempt.
http://gcc.gnu.org/ml/gcc-patches/2000-06/msg00412.html
--Kaveh
--
Kaveh R. Ghazi Engagement Manager / Project Services
ghazi@caip.rutgers.edu Qwest Internet Solutions