GCC Release Status (2003-10-03)

GCC Release Status (2003-10-03)

[Mark Mitchell]

GCC 3.3.2
=========

GCC 3.3.2 was tentatively scheduled for October 1, 2003, which is -2
days away.  Well, I guess we're not going to get *that* release out on
time, unless I can find a time machine on E-Bay.  Part of this is due
to the fact that we've had major deliverables that have taken away
from my ability to do volunteer work on GCC.

There are still a lot of open 3.3.2 regressions: 69 at last count,
which is down from 96 the last time I sent a status report.  So,
that's progress!  On the other hand, we're not getting this down to
anything near zero anytime before December at this rate.

A lot of these PRs are assigned to people -- but it doesn't seem like
much progress is being made. A lot of these PRs have been ignored for
months.  

I see no choice but to do the release anyhow; I can't see how having
no release is better for the world than having a release with a lot of
bug fixes, even if it does have some regressions.

Therefore, I intend to make a test release by next Wednesday, and a
final release a week after that, assuming no absolute show-stoppers.

If there are 3.3.2 PRs for which a patch has been posted, but not
reviewed, please (a) put a pointer to the patch in the PR, and (b) add
me to the Cc list for the PR.  I will try to review and/or apply all
that I get.  Please don't send me mail directly; that will only make
it harder for me to keep track of what's going on.

GCC 3.4
=======

GCC 3.4 is in Stage 2.  GCC 3.4 will enter Stage 3 (bugfixes only) on
October 15th, which is 12 days away.

There are 150 bugs targeted for GCC 3.4 (up from 143); 46 are C++ bugs
(up from 44).

Previous Status Report
======================

http://gcc.gnu.org/ml/gcc/2003-09/msg00284.html

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

Re: GCC Release Status (2003-10-03)

[Gerald Pfeifer]

On Fri, 3 Oct 2003, Mark Mitchell wrote:
> There are still a lot of open 3.3.2 regressions: 69 at last count,
> which is down from 96 the last time I sent a status report.
> [...]
> There are 150 bugs targeted for GCC 3.4 (up from 143); 46 are C++ bugs
> (up from 44).

Comparing numbers, GCC 3.4 seems in surprisingly good shape (or GCC 3.3
in relatively bad shape).  What's your feeling?

Gerald

PS: I applied the patch below to our main page.

Index: index.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/index.html,v
retrieving revision 1.397
diff -u -3 -p -r1.397 index.html
--- index.html	22 Sep 2003 09:41:01 -0000	1.397
+++ index.html	3 Oct 2003 22:56:39 -0000
@@ -71,7 +71,7 @@ to maintain and improve quality.</p>
 <!-- This is a fifth, permanent item. -->

 <dt><strong>Release status:</strong>
-  <a href="http://gcc.gnu.org/ml/gcc/2003-09/msg00284.html">2003-09-05</a>
+  <a href="http://gcc.gnu.org/ml/gcc/2003-10/msg00119.html">2003-10-03</a>
 </dt><dd>
 </dd>

ping, RFA (try 2). DEBUG mainainters: Where to put location_t info?

[Carlo Wood]

On Fri, Oct 03, 2003 at 03:50:55PM -0700, Mark Mitchell wrote:
> GCC 3.4 is in Stage 2.  GCC 3.4 will enter Stage 3 (bugfixes only) on
> October 15th, which is 12 days away.

Can one of main maintainers please answer my question
about how/where to add a location_t *somehow* to a CALL_EXP?

See http://gcc.gnu.org/ml/gcc/2003-09/msg01180.html

This patch would also be a great asset for the next
version of gdb which intends to support 'virtual'
call traces.  It is better if the debug information
on call-locations for inlined functions will be already fixed
in 3.4.  The patch is really small, I just need to know in
which structure, or whereever, to put the extra data.

Please?

-- 
Carlo Wood <carlo@alinoe.com>

Re: GCC Release Status (2003-10-03)

[Eric Botcazou]

> I see no choice but to do the release anyhow; I can't see how having
> no release is better for the world than having a release with a lot of
> bug fixes, even if it does have some regressions.

Even regressions wrt to the previous release? See PR 12223.

> If there are 3.3.2 PRs for which a patch has been posted, but not
> reviewed, please (a) put a pointer to the patch in the PR, and (b) add
> me to the Cc list for the PR.  I will try to review and/or apply all
> that I get.  Please don't send me mail directly; that will only make
> it harder for me to keep track of what's going on.

Done for 3 PRs.

-- 
Eric Botcazou

Re: GCC Release Status (2003-10-03)

[Gerald Pfeifer]

On Sat, 4 Oct 2003, Eric Botcazou wrote:
> Even regressions wrt to the previous release? See PR 12223.

If PR 12223 really breaks QT and OpenOffice, it looks like a real
show-stopper. :-(

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

Re: GCC Release Status (2003-10-03)

[Jakub Jelinek]

On Sat, Oct 04, 2003 at 09:32:55AM +0200, Gerald Pfeifer wrote:
> On Sat, 4 Oct 2003, Eric Botcazou wrote:
> > Even regressions wrt to the previous release? See PR 12223.
> 
> If PR 12223 really breaks QT and OpenOffice, it looks like a real
> show-stopper. :-(

PR 12223 is most likely fixed, it just needs someone to verify on OOo
(certainly QT/KDE works for us again here).

	Jakub

Re: GCC Release Status (2003-10-03)

[Mark Mitchell]

On Fri, 2003-10-03 at 23:41, Jakub Jelinek wrote:
> On Sat, Oct 04, 2003 at 09:32:55AM +0200, Gerald Pfeifer wrote:
> > On Sat, 4 Oct 2003, Eric Botcazou wrote:
> > > Even regressions wrt to the previous release? See PR 12223.
> > 
> > If PR 12223 really breaks QT and OpenOffice, it looks like a real
> > show-stopper. :-(
> 
> PR 12223 is most likely fixed, it just needs someone to verify on OOo
> (certainly QT/KDE works for us again here).

The test case in the PR is fixed, according to your notes.  Therefore,
I've closed the PR.

Now, there is the broader question of "do QT and OpenOffice work?" 
(This PR addressed a specific code-generation problem, which is now
fixed.)

From what you say, QM and OpenOffice are OK.  Therefore, I don't see a
showstopper here.  If someone finds a problem with those applications, a
new PR should be filed.

Thanks,

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

Re: GCC Release Status (2003-10-03)

[Mark Mitchell]

On Fri, 2003-10-03 at 23:21, Eric Botcazou wrote:
> > I see no choice but to do the release anyhow; I can't see how having
> > no release is better for the world than having a release with a lot of
> > bug fixes, even if it does have some regressions.
> 
> Even regressions wrt to the previous release? See PR 12223.

Yes, those are more significant -- but that particular PR has been
fixed; it just wans't marked that way.

> > If there are 3.3.2 PRs for which a patch has been posted, but not
> > reviewed, please (a) put a pointer to the patch in the PR, and (b) add
> > me to the Cc list for the PR.  I will try to review and/or apply all
> > that I get.  Please don't send me mail directly; that will only make
> > it harder for me to keep track of what's going on.
> 
> Done for 3 PRs.

Thanks!  I've reviewed them.

Yours,

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

Re: GCC Release Status (2003-10-03)

[Michael Elizabeth Chastain]

FYI,

I've been running the gdb test suite with various versions of gcc
and checking the results.  As far as I know, the gdb test suite is
the only way to test the debug info that gcc emits.

gcc gcc-3_3-branch is healthy, no regressions against gcc 3.3.1.

gcc HEAD has some regressions, some of which I have reported as
gcc bugs, some of which I have not gotten around to reducing to
bug reports.

My testing is limited to native i686-pc-linux-gnu with
-gdwarf-2 -g2 and -gstabs+ -g2.

If I see anything bad happen with the 3.3 branch I know what to do
(file a bug report, mark it "regression").  And I send my reports
to gcc-testresults.  I just wanted to put a word in since gcc 3.3.2
is so close to release -- I promise not to post a note every time
Mark Mitchell writes status!

Michael C
GDB QA Guy