Re: merge-with-binutils documentation is wrong/incomplete

["Joseph S. Myers" <jsm28 at cam dot ac dot uk>]

On Wed, 21 Mar 2001, Markus Werle wrote:

> Please fill in the following form:
> ------------------------------------------------

I take it you're volunteering to update the online documentation using the
answers to this?

>    [  ] is possible and can be achieved without any conflicts
>    using the following cvs commands (login omitted):

Try using /cvs/uberbaum (I haven't tried this; see discussion of different
philosophies below).

There is a range of different philosophies on the unification of GCC and
binutils CVS trees, from

(a) everything should be unified in one tree (so cvs HEAD checkouts test
development GCC with development binutils);

through

(b) GCC and binutils should be separate, but all front ends should be
unified in the one GCC tree;

to

(c) there should be documented and fixed interfaces between the different
parts of the tree, which should be developed and frozen on their own
schedules (so, one version of a front end could be used with another of a
back end); so no such unification of either binutils or different front
ends would be needed.

I prefer (b) - that is, the interfaces to allow binutils to be built in a
single tree should be documented (and I hope you'll be able to assist with
this documentation), but if GCC starts linking against BFD (say) then it
should use public interfaces to allow linking against any BFD in a given
version range, either an installed BFD or a stable BFD release imported
into the GCC tree, but not CVS HEAD BFD unless a specific action is taken
to use that.  People working on both GCC and binutils together may prefer
(a).  Some people prefer in principle the better discipline and
scalability of (c), e.g.
<URL:http://gcc.gnu.org/ml/gcc/2000-08/msg00675.html>.

-- 
Joseph S. Myers
jsm28@cam.ac.uk