This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
GCC 3.0 Status Report
- To: gcc at gcc dot gnu dot org
- Subject: GCC 3.0 Status Report
- From: Mark Mitchell <mark at codesourcery dot com>
- Date: Mon, 14 May 2001 14:28:15 -0700
- Organization: CodeSourcery, LLC
GCC 3.0 Status Report
=====================
Overall
-------
The last major piece of functionality for the release (the
new ABI-compliant exception-handling model) has been installed.
Accompanying this are improvements to the C++ standard library
to work with the new model. Thanks to Richard Henderson
and Benjamin Kosnik for implementing these changes.
We are getting close to zero unexpected FAILs in the testsuites
on many major platforms. We will achieve this goal soon. I have
fixed many, many bugs in the last week or so, and disabled some
tests that do not represent regressions.
If you are a platform maintainer/tester, and you know of a bug that
you believe is essential for the release -- in other words, that is
is a regresssion from GCC 2.95.x -- please make sure it is in GNATS
and marked "high" priority. Remember that the criteria here is a
*regression*. At this point, if it is broken with GCC 2.95.x and
not already fixed, it is going to stay broken. It would take an
earth-shatterring bug of doom to make me think otherwise.
Please do not send me your list of bugs directly. Put them in
GNATS; that is what it is for. If you send them to me, I will
likely delete them. Then, I will deny every having received
them. :-)
Schedule
--------
I will be on vacation May 25th through June 2nd. I was hoping
to make the release before May 25th. Now, I hope to make a test
release before I go.
At this point, I am finally willing to set a schedule.
Previously, I believed that we should wait until certain functional
criteria were met. We are now very close to those requirements.
And, time *does* matter; there really is value in getting this beast
out the door.
Therefore, we're now going to pretend that the company has told us
that we must ship by a certain date, or the copmany will be unable
to raise additional financing, that we will be summarily fired
without severance, and that all of our stock options will be
worthless. If you've worked at a dot-bomb recently, this should
sound familiar.
The release will be made on or before 11:59PM GMT -8 (Pacific
Daylight Time) June 15th, 2001 A.D. Because otherwise I will
commit suicide at 12:00 AM on June 16th, and you will all feel
very, very bad.
Requests for Help
-----------------
Is there anyone willing to work on the release script? This
script builds a tarball from what's in CVS, but also needs to
add generated files not in CVS (e.g., `parse.y'). We need to
test that the generated tarball actually works. This is a
classic release engineering task. It would be very helpful to
me if someone who is skilled in this area would step forward,
thereby freeing me up to continue fixing bugs.
Please track down GNATS bugs that apply to your targets,
front-ends, etc. Fix 'em, post 'em, check 'em in.
Please try to test GCC on as much software as you can. Build
X windows. Build the Linux kernel. Run 'em. File bugs in GNATS.
Repeat.
Volunteer of The Week
---------------------
This weeks winner is tireless in his efforts to clean up
crufty old bits of GCC. Always willing to look for away to
avoid the "expedient hack", he has often gone where angels, and
even fools, fear to tread. Recently, he's tackled issues
involving crufty old headers installed by GCC, ancient Makefile
ugliness, and ancient configury ugliness. Someday, I fully
expect that he will the README.ALTOS and README.ACORN from the
GCC directory itself.
All of this effort will make the GCC 3.1, and every release
thereafter, easier -- so our Volunteer of the Week is ...
... Zack Weinberg!
Congratulations.