Re: Adding zlib to GCC

[Tom Tromey <tromey at cygnus dot com>]

>>>>> "Mark" == Mark Mitchell <mark@codesourcery.com> writes:

Mark> The whole point of component architectures is to provide stable,
Mark> independent components.  Something like zlib should be a
Mark> different component, and it should be used unmodified in GCC.
Mark> If it needs modification, we should work to put together a new
Mark> zlib release, and we should use that release.

In this particular case:

We don't own zlib.  Someone else does.  We'd be including it in the
gcc tree for convenience, since it is required by various Java
components.  (And the goal of getting Java into the tree is to make
sure people actually build and test it.)

The net zlib doesn't use configure and can't be dropped into our
source tree as-is.

This isn't the only case like this for Java support -- just the first
one.  We also will want to include the Boehm GC (also doesn't use
configure) and fastjar (maintained elsewhere).  Our policy for libgcj
has been to minimize the number of local changes (in most cases,
long-term divergences are limited to configury), and to closely track
the upstream sources using `cvs import'.

Mark> Trust me -- I'm finding it very hard to get a GCC 3.0 release
Mark> done, largely because I'm waiting for people to stabilize
Mark> various pieces of the tree, and complete little bits of
Mark> implementation.  The more intertwined things are, the harder
Mark> that will be.

I think this is a different problem, namely that there are too many
changes going into gcc 3.0.

Tom