GCC 4.0 vs. 3.5 -- don't decide now!

[Nathanael Nerode]

I think we should postpone discussion of this until after stage 1,
and possibly even after stage 2.

It's possible that the changes will be compatibility-breaking enough
to justify a major version increment on that basis; it's also possible
that we will fix most of the compatibility points (if we're really really
good).

It's possible that the performance improvement will be enough to justify
a major version increment on that basis; it's also possible that it won't,
or indeed that it will regress for lots of people.  If it isn't a significant
improvement (even if it paves the way for great improvements later), we don't
want to saddle GCC 4 with the same reputation as GCC 3!

It's possible that even *more* major changes will be present in 3.5, in
addition to gfortran and tree-ssa, which would be a definite point in
favor of calling it "4.0" -- but we don't know yet.  :-)  For instance,
the question of how much cleanup/removal of the RTL passes is done
after tree-ssa goes in makes a big difference.  Also, if new-ra ever gets
ready to be used by default, that would be a major change too.

However, "3.5" is a fairly decent version number from a "big change"
perspective; it's halfway to 4.0.

I think that after stage 1, and certainly after stage 2, we will have a
clearer sense of what the correct version number is.  So I strongly support
putting off the version number decision.  :-)

-- 
There are none so blind as those who will not see.