Re: 2 x86-64 ABI bugs in gcc 3.3 and 3.4

[Gerald Pfeifer <gerald at pfeifer dot com>]

As usual, Joe did an excellent job in summarizing the issue (also from my
point of few) :-), so I'm only going to add some specific notes.

On Thu, 20 May 2004, Gabriel Dos Reis wrote:
> I believe a variable is forgotten here:  The sequencing of the
> releases.  The fixes are included in consequentive releases.  3.3.3
> comes before 3.4.0 which comes before 3.3.4 which comes before 3.4.1.
> So saying, it is broken in 3.2, fixed in 3.3 and 3.4 is not entirely
> accurate.

Please keep in mind that distributions and many professional users in
general often do not have the opportunity to readily update to the latest
version of a release branch.

> This sets that either I indirectly direct which patch goes in 3.4.x or
> 3.4.x blocks fixes for 3.3.x.  We can examine some specific situations
> on case by case basis, but I think in general that is not good.

>From my personal experience, I feel that this should be the general policy
when backporting fixes:  Bugs on the 3.3 and 3.4 branches should either be
fixed on both, or only on 3.4.

(I can well understand that, from the point of view of the 3.3 release
manager and users of the 3.3 branch, this doesn't look very appealing,
but looking at the overall picture and lessons learned from my day job,
that's my personal understanding.  I can well imagine others disagreeing
quite strongly.)

> To arrive to the point where we fixed those many bugs in 3.3.x I had
> to use rules less stricter than for 3.4.x -- because I think the
> branching was cut too early, and those fixes should have been in.

I believe 3.3.3 has been a good one, and I applaud you (and the other
contributors) for that.  I agree with Joe that, now 3.4.0 is out, we
should move towards more defensive maintenance mode, but I understood
that this has been your strategy anyways, so we may be in violent
agreement here anyways!

Gerald
-- 
Gerald Pfeifer (Jerry)   gerald@pfeifer.com   http://www.pfeifer.com/gerald/