This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Beyond GCC 3.0: Summing Up
- To: Mark Mitchell <mark at codesourcery dot com>
- Subject: Re: Beyond GCC 3.0: Summing Up
- From: "H . J . Lu" <hjl at lucon dot org>
- Date: Tue, 10 Jul 2001 00:42:58 -0700
- Cc: Alexandre Oliva <aoliva at redhat dot com>,"dewar at gnat dot com" <dewar at gnat dot com>,"kenner at vlsi1 dot ultra dot nyu dot edu" <kenner at vlsi1 dot ultra dot nyu dot edu>,"gcc at gcc dot gnu dot org" <gcc at gcc dot gnu dot org>
- References: <orzoadxvh6.fsf@guarana.lsd.ic.unicamp.br> <333370000.994742904@warlock.codesourcery.com>
On Mon, Jul 09, 2001 at 10:28:24PM -0700, Mark Mitchell wrote:
>
> This argument is common -- but seems to be refuted by the fact that
> literally hundreds of regressions were fixed on the 3.0 branch, and
> that many more remain. Why weren't they fixed earlier? Many of them
> were reported months and months before they were fixed.
It took 10 months to check in a simple patch to fix a testcase. That
is only after my repeated requests. I have to thank Zack for getting
out his way to make it happen. I have sent quite a few Linux/mips
patches. I am still waiting. I don't think it is that hard to review
all the patches sent in and try to fix all the regressions reported.
It is just a matter of priority.
BTW, the longer between releases, the more regressions you will get.
You get what you ask for. If we shouldn't make frequent major releases,
we should at least make more frequent bug-fix releases, like once a
month. No compiler is regression free. We don't have to remove all
regressions in a bug fix release, as long as
1. It does fix some regressions.
2. We are committed to fix all regressions.
You will set yourself up for a failure if you want to make a perfect
release. To me, gcc 3.0 is a decent .0 release, which is buggy by
default. I don't think an incremental bug fix gcc 3.0.x for every
2 or 3 weeks is a bad idea. I don't understand why anyone wants to
fix all regressions in 3.0.1. I can tell you right now that you will
make 3.0.2 if you want to fix the remaining/new regressions in
3.0.1. We should keep making 3.0.x and let users tell us when it is
good enough. Don't make making a release so complicated. I have
done that for the Linux binutils just by myself. Nothing is perfect,
nor is my binutils. But it is ok as long as I can use it to build a
whole Linux distribution.
H.J.