This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: patches to move to gcc-3_2-branch
- From: Mark Mitchell <mark at codesourcery dot com>
- To: Benjamin Kosnik <bkoz at redhat dot com>
- Cc: "libstdc++ at gcc dot gnu dot org" <libstdc++ at gcc dot gnu dot org>
- Date: Thu, 28 Nov 2002 11:01:11 -0800
- Subject: Re: patches to move to gcc-3_2-branch
--On Thursday, November 28, 2002 12:01:43 PM -0600 Benjamin Kosnik
<bkoz@redhat.com> wrote:
We're using this policy for every other part of the system; why not
V3?
The locale bits weren't final in 3.1, so "regression" means nothing.
Right -- and that's the point. We got by without thoses bits in GCC 3.1,
and we can get by without them now.
I'm trying to keep 3_2-branch in sync with gcc-3_2-rhl8-branch, which is
getting bug fixes as well as regression fixes (both of which I agree
with). I feel that the "regression-only" policy is stymieing my best
intentions to keep a level playing field for all.
It's not your job to keep a level playing field. If another distributor
wants to use the same bits that are on the Red Hat branch they can do
that. It's nice that those bits are accessible.
The 3.2 branch is supposed to be a stable branch for releases that are
used by people on sall sorts of systems. Most people are going to be
totally unaffected by new features because they're not using those
features in GCC 3.2.1. When they upgrade to GCC 3.2.2, they'd like to
see bugs fixed, and nothing broken.
You're right that the branch diverges from the mainline quickly; this
happens across the compiler. It's true that this makes maintaining
the branch a bit tricky, but it is also means that, for example, we
delivered a 3.2.1 that fixed tons of bugs -- and I believed introduced
very few new problems.
I understand your frustruation, but I think it's important to treat V3
like everything else.
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com