GCC Status Report (2004-02-29)

[Mark Mitchell <mark at codesourcery dot com>]

GCC 3.4
=======

We have 57 regressions open against GCC 3.4.  This number does *not*
include those regressions that I have delayed until future releases
because they affected non-primary problems, etc.  I will pick over the
remaining 57 to some degree to winnow the number even more -- but
there are some pretty serious bugs remaining.

We have 8 wrong-code regressions.

We have 16 ice-on-valid-code regressions, most of which are in plain
C.  We seem to be having particular problems with cselib_record_set.
Jan, can you help with that?  See, for example, PR 14345.

There are a number of problems relating to Zack's unfinished c-decl.c
rework.  The changes to date seem to have fixed one class of problems,
but created another class.  Some of these problems can be fixed
without finishing Zack's rewrite -- but others cannot.  Zack and I
have talked about this, and concluded that the rewrite needs to
happen: but for 3.5, not 3.4.  Most of the regressions in 3.4
resulting from these changes are not terribly serious, and we'll be
fixing the ones that *are* serious shortly.

There are a number of compile-time performance problems, which, again,
will probably not get solved before the release due to the risk
associated with the likely fixes.

In order to remain on track for March 15th, we need to make some good
headway in this coming week.

As of Monday, March 8, all non-documentation check-ins to the branch
will require my explicit approval.

GCC 3.5
=======

We continue to await eagerly the tree-ssa merge.  Until that happens,
it's very hard to say anything more about the state of GCC 3.5.

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com