GCC 3.0 Status Report

[Mark Mitchell <mark at codesourcery dot com>]

[Note: if your name appears explicitly in the `Cc' list, there is an
       action item for you at the bottom of this report.]

GCC 3.0 Status Report
=====================

Timing
------

There are ten days until the release of GCC 3.0.

Major Annoucements
------------------

As announced in the May 21st Status Report, GNATS is now frozen.  No
new bugs can be made high priority without my express approval.  If
you wish to mark a bug high priority, you must first send me mail
explaining why fixing the bug is of *absolutely vital* import.  It
must not only be a regression from GCC 2.95 -- it must also be in some
way cataclysmic.  Failing to build the Linux kernel or the X server
for an x86 would be one example.  No optimization issue is likely to
qualify; correctness issues are probably the only likely candidates.
Crashing on illegal code for which we used to give an error mesage is
probably not a candidate.  Even failing to compile a correct, but
corner-case example, might not qualify.

Overview
--------

Unfortunately, there are many more high-priority bugs now than there
were a week ago.  I am saddenned that more of them have not been
fixed.  Many of these bugs are libstdc++ or Java bugs.  Those two
teams should focus on fixing or downgrading these bugs.  It is
important that I have an overview of what people are really trying to
fix, and what we are going to leave for another release.

The high number of open bugs, and the slow rate at which people are
fixing them, is going to force me to spend more time than I had hoped
on bug-fixing.  That means that I will have less time than I hoped for
packaging, documentation, release notes, etc.  If you are at all able
to contribute in those areas, please do so.  Gerald Pfeifer continues
to coordinate the documentation side of things.

Thanks to those who worked on the release script in my absence!  Since
there do not appear to be any packaging problem reports in GNATS, I
will assume that the tarballs work as advertised.  If you are not
contributing to the release in any other way, testing the prerelease
tarballs to make sure that they build and install is an easy and very
useful thing to do.

Fortunately, it looks like we were able to avoid testsuite regressions
for the last week, which is excellent news.

Action Items
------------

V3 Team, Java Team: Fix or downgrade V3 and Java issues.

RTH: Did we fix the ARM conditional-execution problem?
     Or should we go with the patch I proposed that you said disabled
     a lot of conditional-lifetime information?

Everyone else: Continue to fix high-priority bugs.
	       Application testing.
	       Documentation.