Shared library annoyance with gcc-3_0-branch

Bo Thorsen bo@sonofthor.dk
Sun Feb 18 07:50:00 GMT 2001


On Sun, 18 Feb 2001 dewar@gnat.com wrote:
> <<I thought the Open Source (or Free Software) philosophy was "Release
> early, release often" and "Get a partial solution out there, we can
> perfect it later."
>
> This may be some people's philosophy, but to say that all Free Software
> follows this philosophy is wrong, and in fact damaging. It implies that
> all Free Software is in a constantly changing and half-baked state.
> Certainly our viewpoint is diametrically opposed.
>
> Frequent releases are a menace to real production use. Of course there
> is no objection to a fluid development environment where frequent beta
> versions are available, that is certainly desirable for ongoing testing
> and development purposes.

The Linux kernel has been proving this statement wrong for a very long
time and glibc has been proving it right. You can not state this as a
black and white fact since it depends on wether or not new releases will
break stuff made for the previous ones. If the abi for c and c++ really
are cast in stone after a GCC 3.0 is out, it should not be a problem to
increase the release schedule - of course with the exception that it
really is a lot of work to do so. However, if people do not accept that
until a GCC 4, no matter what the problem the abi can not be changed, then
you have exactly the situation you describe.

> But my view is that releases should be treated with the same intense
> insistence on quality and perfection that applies to any software that
> is intended to be used in a production environment.

Noone can argue against this, but the line gets blurred when the current
released compiler is too buggy to be usable. This is IMHO not the case for
i386, but for other ports it could be - I have been hearing all sorts of
curses from the alpha people over GCC. Also, not releasing anything is
what made RedHat include the bogus 2.96. If at least bugfix releases had
come out, this would probably not have happened.

My main point here is that as long as binary and source compatibility is
maintained, there will not be a problem in releasing more often. But if
people can't resist fixing things that breaks compatibility I would tend
to agree with you.

Bo.

-- 

     Bo Thorsen                 |   Lahnsgade 31, st.
     Free software developer    |   5000 Odense C
     SuSE Labs                  |   Denmark



More information about the Gcc mailing list