Plan for GCC-3.2.3

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

Hi,

   Here is a plan for GCC-3.2.3.

I believe there are needs for GCC-3.2.3:

   1) the lastest release (3.2.2) exhibited packaging problems
      (missing info files);

   2) the lastest release (3.2.2) failed to bootstrapped on some
      plateforms.  The regression was subsequent to a another
      regression fix.

   3) a stable variant of 3.2.x series before the next major 3.3.0
      release. 

I think cases #1 and #2 speak for themselves.

Before GCC-3.3.0 (which we all expect to be better than 3.2.x) goes
out, I think we need a stable enough variant of 3.2.x series that
can act as both a fallback and a reference point.  

I also believe that somehow we need to give ourselves some time to fix
the regressions (compile-time, semantics, optimizations, ...) that are
pending in 3.3.x.  Realistically speaking, those regressions won't be
fixed in the next two or three weeks.  On the other hand, I'm not
convinced we should get 3.3.0 out under the pressure.  All that is to
say that there is room to release a usable enough 3.2.3 compiler  so
as to give ourselves enough latitude to work on the 3.3 regression.

After 3.2.3 is out, I'll consider the gcc-3_2-branch closed and will
invite everybody to concentrate resources on 3.3 and 3.4, i.e. help Mark.

I don't want to release 3.2.3 in the hurry (like last time).  On the
other hand, my agenda is going to be a bit squezed near the end of
this month and the first days of April (from 1st to 11th I'll be in
Oxford for the ACCU conference and the ISO C++ committee meeting).
I'll be able (at least I hope so) to read mails and such during that
period.  Given that, I think 15th of April is a good candidate.
Therefore, I consider the period 

   from 2003-04-01 to 2003-04-14 inclusive

to be semi-closed for gcc-3_2-branch in the sense that any
non-documentation patch ought to be routed by me and an appropriate
maintainer; such a patch shall fall under the constraints 2) explained
below.  This period will be for intensive testing.  


The period

   from 2003-03-10 to 2003-03-31 inclusive

is open for gcc-3_2-branch in the following sense:

  a patch can be applied to gcc-3_2-branch if and only if:
    
     1) it fixes a regression from previous GCC-3.x releases or from
        GCC-2.95.x; otherwise 

     2) it ought to have a very strong justification, be important to
        multiple users if fix a bug in some significant free software
        package, and be low risk.
  

Comments and suggestions are mostly welcome.

-- Gaby