Re: what DOES the GPL really say?

[Craig Burley <burley@tweedledumb.cygnus.com>]

Toon Moene <toon@moene.indiv.nluug.nl> writes:

> kenner@lab.ultra.nyu.edu (Richard Kenner) wrote:
> 
> > In article <5u9rdf$r5t$2@news.utrecht.NL.net> Toon Moene 
> <toon@moene.indiv.nluug.nl> writes:
> > >That depends.  The next official release of g77, g77-0.5.21, on which we 
> will 
> > >continue to work as soon as the GNU machines at MIT come on the net again, 
> > >will be against gcc-2.7.2.3.  I don't know if that release contains the 
> new 
> > >build_complex routines.
> 
> > My understanding is that it will.  That's what I meant.  In any event, the
> > "fix" is trivial, though a person trying to build it would have to know
> > about it.
> 
> If that's the case, the best thing to do for someone building g77 + GNAT 
> together is to disregard the g77 backend patch alltogether (note: from 
> g77-0.5.21 onwards !).  That will give you a Fortran compiler that produces 
> suboptimal objects, but they will be as correct as can be, and the Ada 
> compiler isn't compromised (which is the whole point of using Ada).
[...]
> Sorry to be so vague about 2.7.2.3 - I simply haven't got the time yet to 
> look at it (nor the disk space) ....

Actually, right now it looks like g77 0.5.21 will be based on gcc
2.7.2.3, and anyone wanting to combine it with GNAT should ignore
the *GNAT* patches, which g77 includes.  If they want to do the
opposite -- ignore g77's patches and use GNAT's instead -- they'll
have to change a few things in g77.  I thought about changing those
anyway for 0.5.21, but have been convinced otherwise for this
release at least.  0.5.22 might well be another story, whenever
that appears.  The work needed to make g77 build and work okay
with only the GNAT patch set, or no patches to gcc at all, is probably
only slightly more difficult than that to make 0.5.20 and previous
g77 versions work with GNAT's patch set (the incompatible build_complex()
change to the back end).

BTW, 2.7.2.3 as a base seems to be just fine for g77, at least
according to my g77 test suite and the c-torture-1.45 tests.  That
is, no worse than 2.7.2.2 in any way I can see.  Aside from the
screwups between the .tar.gz and .diff.gz distributions that
afflicted 2.7.2.2 and, therefore, 2.7.2.3, that is; g77 will assume
the user somehow got the "correct" 2.7.2.3 ChangeLog file, though
maybe I'll figure out a way to make the g77 patches appear to
apply cleanly regardless, e.g. by deleting the context lines.

(gcc-2.7.2.1-2.7.2.2.diff.gz is "broken" in that it doesn't patch
gcc/ChangeLog; unfortunately both gcc-2.7.2.3.tar.gz and
gcc-2.7.2.2-2.7.2.3.diff.gz are based on the resulting broken
gcc/ChangeLog, which is basically the 2.7.2.1 one.)
-- 

"Practice random senselessness and act kind of beautiful."
James Craig Burley, Software Craftsperson    burley@gnu.ai.mit.edu?