RE: 3.3 target date?

RE: 3.3 target date?

[S. Bosscher]

Gerald Pfeifer wrote:
> On Wed, 19 Feb 2003, Andreas Jaeger wrote:
> > Our testing at SuSE showed that 3.3 is in a pretty good shape.
> 
> That's interesting, because we still have 77 regressions in 3.3 listed
> in GNATS:
---- 8< ----
> And some 26 regressions in 3.2 (some of which might also apply to 3.3);
> plus a significant compile-time performance problem.

Fortunately all of those 26 are either fixed for 3.3 or included in those 77
open PRs you mentioned.

Greetz
Steven

Re: 3.3 target date?

[Joel Sherrill]

"S. Bosscher" wrote:
> 
> Gerald Pfeifer wrote:
> > On Wed, 19 Feb 2003, Andreas Jaeger wrote:
> > > Our testing at SuSE showed that 3.3 is in a pretty good shape.
> >
> > That's interesting, because we still have 77 regressions in 3.3 listed
> > in GNATS:
> ---- 8< ----
> > And some 26 regressions in 3.2 (some of which might also apply to 3.3);
> > plus a significant compile-time performance problem.
> 
> Fortunately all of those 26 are either fixed for 3.3 or included in those 77
> open PRs you mentioned.

PR9255 is a regression from 3.2 which prevents m68k targets from
compiling.

Overnight, i386-rtems, sh-rtems, and sh-rtemself wouldn't complete a
build
on the 3.3 branch.  I have updated my tree and am going to see if these
failures were fixed overnight and are repeatable on i386-elf, sh-coff,
and
sh-elf respectively.

> Greetz
> Steven

--joel sherrill

Re: 3.3 target date?

[Peter Barada]

>PR9255 is a regression from 3.2 which prevents m68k targets from
>compiling.
>
>Overnight, i386-rtems, sh-rtems, and sh-rtemself wouldn't complete a
>build on the 3.3 branch.  I have updated my tree and am going to see
>if these failures were fixed overnight and are repeatable on i386-elf,
>sh-coff, and sh-elf respectively.

Joal,

Try this patch.  It works on Uberbaum:

Index: combine.c
===================================================================
RCS file: /cvs/uberbaum/gcc/combine.c,v
retrieving revision 1.340
diff -c -r1.340 combine.c
*** combine.c	14 Feb 2003 07:35:44 -0000	1.340
--- combine.c	19 Feb 2003 13:56:53 -0000
***************
*** 8068,8073 ****
--- 8068,8079 ----
    unsigned HOST_WIDE_INT nonzero;
    int i;
  
+ #if 1
+   /* Can only simplify integer modes  */
+   if (!(GET_MODE_CLASS (mode) == MODE_INT || GET_MODE_CLASS (mode) == MODE_PARTIAL_INT))
+     return x;
+ #endif
+ 
    /* Simplify VAROP knowing that we will be only looking at some of the
       bits in it.

-- 
Peter Barada
peter@baradas.org

m68k regression (PR9255) was Re: 3.3 target date?

[Joel Sherrill]

Peter Barada wrote:
> 
> >PR9255 is a regression from 3.2 which prevents m68k targets from
> >compiling.
> >
> >Overnight, i386-rtems, sh-rtems, and sh-rtemself wouldn't complete a
> >build on the 3.3 branch.  I have updated my tree and am going to see
> >if these failures were fixed overnight and are repeatable on i386-elf,
> >sh-coff, and sh-elf respectively.
> 
> Joal,
> 
> Try this patch.  It works on Uberbaum:

It works on the 3.3 branch and allows m68k-elf to build to completion.  
m68k-rtems and m68k-coff have not been checked with this patch.

Since it is to a common file, someone very knowledgeable will have
to review it.  

--joel

> Index: combine.c
> ===================================================================
> RCS file: /cvs/uberbaum/gcc/combine.c,v
> retrieving revision 1.340
> diff -c -r1.340 combine.c
> *** combine.c   14 Feb 2003 07:35:44 -0000      1.340
> --- combine.c   19 Feb 2003 13:56:53 -0000
> ***************
> *** 8068,8073 ****
> --- 8068,8079 ----
>     unsigned HOST_WIDE_INT nonzero;
>     int i;
> 
> + #if 1
> +   /* Can only simplify integer modes  */
> +   if (!(GET_MODE_CLASS (mode) == MODE_INT || GET_MODE_CLASS (mode) == MODE_PARTIAL_INT))
> +     return x;
> + #endif
> +
>     /* Simplify VAROP knowing that we will be only looking at some of the
>        bits in it.
> 
> --
> Peter Barada
> peter@baradas.org

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel@OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
Support Available                (256) 722-9985

RE: 3.3 target date?

[S. Bosscher]

When I looked yesterday, there were 78 (!) high-priority PRs for 3.3, half
of them (38) for c++.

For some PRs there may be patches pending (hint: PR c/8828 ;-), but most PRs
have not been analyzed beyond "confirmed" and are not even assigned to
anyone. There's not that much bug fixing going for 3.3, especially on c++
where everybody is focusing on the new parser.

As a reference, this morning there are at least 87 high-priority PRs for
3.4, with 56 of them for C++.  This includes the PRs for 3.3 that are still
present on the mainline...

There's also the compile time issue.  Apple seems to still be working on
3.3, and it would be nice to have some of that go in before the release.
But we're only 12 days away from March 1...

Greetz
Steven



-----Original Message-----
From: Andreas Jaeger
To: Steven Bosscher
Cc: Paolo Carlini; David Rasmussen; gcc@gcc.gnu.org
Sent: 19-2-03 9:52
Subject: Re: 3.3 target date?

Steven Bosscher <s.bosscher@student.tudelft.nl> writes:

> Op di 18-02-2003, om 16:05 schreef Paolo Carlini:
>> David Rasmussen wrote:
>> 
>> >When is the planned release date for 3.3 (and 3.4 for
>> >that matter)? Where can I see such things in general?
>> >
>>     http://gcc.gnu.org/develop.html
>> 
>
> ... but March 1 doesn't sound like a very realistic release date now,
> does it?

Our testing at SuSE showed that 3.3 is in a pretty good shape.
There're - for x86 and x86-64 only two critical bugs:
- miscompilation of openoffice (just fixed last night by Mark Mitchell
  - thanks!)
- boost library does not compile

(I don't have the PRs handy)

For other platforms I'm aware of a miscompilation on powerpc that Olaf
reported yesterday but that's it for s390 and powerpc.

I'd like to see a March 1 release,
Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj

3.3 target date?

[David Rasmussen]

When is the planned release date for 3.3 (and 3.4 for
that matter)? Where can I see such things in general?

/David

__________________________________________________
Do you Yahoo!?
Yahoo! Shopping - Send Flowers for Valentine's Day
http://shopping.yahoo.com

Re: 3.3 target date?

[Paolo Carlini]

David Rasmussen wrote:

>When is the planned release date for 3.3 (and 3.4 for
>that matter)? Where can I see such things in general?
>
    http://gcc.gnu.org/develop.html

Paolo.

Re: 3.3 target date?

[Steven Bosscher]

Op di 18-02-2003, om 16:05 schreef Paolo Carlini:
> David Rasmussen wrote:
> 
> >When is the planned release date for 3.3 (and 3.4 for
> >that matter)? Where can I see such things in general?
> >
>     http://gcc.gnu.org/develop.html
> 

... but March 1 doesn't sound like a very realistic release date now,
does it?

Greetz
Steven

Re: 3.3 target date?

[Andreas Jaeger]

Steven Bosscher <s.bosscher@student.tudelft.nl> writes:

> Op di 18-02-2003, om 16:05 schreef Paolo Carlini:
>> David Rasmussen wrote:
>> 
>> >When is the planned release date for 3.3 (and 3.4 for
>> >that matter)? Where can I see such things in general?
>> >
>>     http://gcc.gnu.org/develop.html
>> 
>
> ... but March 1 doesn't sound like a very realistic release date now,
> does it?

Our testing at SuSE showed that 3.3 is in a pretty good shape.
There're - for x86 and x86-64 only two critical bugs:
- miscompilation of openoffice (just fixed last night by Mark Mitchell
  - thanks!)
- boost library does not compile

(I don't have the PRs handy)

For other platforms I'm aware of a miscompilation on powerpc that Olaf
reported yesterday but that's it for s390 and powerpc.

I'd like to see a March 1 release,
Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj

Re: 3.3 target date?

[Gerald Pfeifer]

On Wed, 19 Feb 2003, Andreas Jaeger wrote:
> Our testing at SuSE showed that 3.3 is in a pretty good shape.

That's interesting, because we still have 77 regressions in 3.3 listed
in GNATS:

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?synopsis=3.3&priority=high&responsible=&state=&ignoreclosed=Ignore%20Closed&class=&submitter_id=&originator=&release=&columns=category&columns=state&columns=class&columns=responsible&columns=synopsis&displaydate=Display%20Current%20Date&cmd=submit%20query&sortby=Responsible&.cgifields=category&.cgifields=state&.cgifields=confidential&.cgifields=severity&.cgifields=ignoreclosed&.cgifields=responsible&.cgifields=class&.cgifields=submitter_id&.cgifields=displaydate&.cgifields=columns&.cgifields=priority

And some 26 regressions in 3.2 (some of which might also apply to 3.3);
plus a significant compile-time performance problem.

> I'd like to see a March 1 release,

I suppose that all those 77(+26) regressions are problems real users had
with their code bases, so we should at least trim this down significantly.

Gerald
-- 
Gerald "Jerry"   pfeifer@dbai.tuwien.ac.at   http://www.pfeifer.com/gerald/

Re: 3.3 target date?

[Andreas Jaeger]

Gerald Pfeifer <pfeifer@dbai.tuwien.ac.at> writes:

> On Wed, 19 Feb 2003, Andreas Jaeger wrote:
>> Our testing at SuSE showed that 3.3 is in a pretty good shape.
>
> That's interesting, because we still have 77 regressions in 3.3 listed
> in GNATS:
> [...]

Those 77 regressions are sometimes also against GCC 3.2, about 25 are
unique to 3.3 - but I fear more testing will show more ;-).

> And some 26 regressions in 3.2 (some of which might also apply to 3.3);
> plus a significant compile-time performance problem.
>
>> I'd like to see a March 1 release,
>
> I suppose that all those 77(+26) regressions are problems real users had
> with their code bases, so we should at least trim this down significantly.

I agree!  Even if our testing inside SuSE with compilation of around
3000 RPMs has not stumbled over those bugs, it does not mean that
they're not real or unimportant - and I hope that all these bugs will
soon be fixed and we'll also help there.

Note that when we looked first into this in December, we had more than
30 ICEs in different code on x86, now we're down to just 1.

Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj

Re: 3.3 target date?

[Eric Botcazou]

> Note that when we looked first into this in December, we had more than
> 30 ICEs in different code on x86, now we're down to just 1.

For obvious reasons, x86 is not in too bad shape on the 3.3 branch. But I 
don't think we can say the same thing for all other architectures (Sparc 
especially comes to mind).

-- 
Eric Botcazou

Re: 3.3 target date?

[Svein E. Seldal]

Gerald Pfeifer wrote:
> That's interesting, because we still have 77 regressions in 3.3 listed
> in GNATS:

Hmm. I've committed a PR (PR/9387) for the gcc-3.3. And I see now that 
I've failed to mark the synopsis correctly, so it fails your search. 
(And I was uncertain about marking this as a high or a medium bug, so I 
marked it as medium.) -- Maybe thats why nobody have looked at it yet....

Maybe someone should evaluate the PR and at least give it a correct 
status and priority...

Thanks,
Svein Seldal

Re: 3.3 target date?

[Gerald Pfeifer]

On Wed, 19 Feb 2003, Svein E. Seldal wrote:
> Hmm. I've committed a PR (PR/9387) for the gcc-3.3. And I see now that
> I've failed to mark the synopsis correctly, so it fails your search.

From the PR it's not clear whether this is
 (a) a bug on the gcc-3.3 branch, or
 (b) a regression, that is, a bug on the gcc-3.3 branch that is new
     (in the sense that earlier versions of GCC got it right)

Only in the latter case should it be marked "high" and labled as a
regression.

Also, do you see the same on mainline?

(Depending on the answers to these two questions, please go ahead and
update the priority and synopsis of the PR.)

Thanks,
Gerald
-- 
Gerald "Jerry"   pfeifer@dbai.tuwien.ac.at   http://www.pfeifer.com/gerald/

Re: 3.3 target date?

[Dan Kegel]

Gerald Pfeifer wrote:
 > ... we still have 77 regressions in 3.3 listed
> in GNATS:
> 
>  http://gcc.gnu.org/cgi-bin/gnatsweb.pl?synopsis=3.3&priority=high&responsible=&state=&ignoreclosed=Ignore%20Closed&class=&submitter_id=&originator=&release=&columns=category&columns=state&columns=class&columns=responsible&columns=synopsis&displaydate=Display%20Current%20Date&cmd=submit%20query&sortby=Responsible&.cgifields=category&.cgifields=state&.cgifields=confidential&.cgifields=severity&.cgifields=ignoreclosed&.cgifields=responsible&.cgifields=class&.cgifields=submitter_id&.cgifields=displaydate&.cgifields=columns&.cgifields=priority
> 
> And some 26 regressions in 3.2 (some of which might also apply to 3.3);
> plus a significant compile-time performance problem. ...
> 
> I suppose that all those 77(+26) regressions are problems real users had
> with their code bases, so we should at least trim this down significantly.

That's for sure.  Fixing these regressions would go a long
way towards healing gcc3.x's reputation, which is a bit
tattered in some circles.

(I'm going to start running gcc 3.2.2's regression test suite
here soon, I hope, so maybe I can be part of the solution
instead of just a ranting onlooker.)
- Dan

Re: 3.3 target date?

[Mark Mitchell]

--On Wednesday, February 19, 2003 09:52:01 AM +0100 Andreas Jaeger 
<aj@suse.de> wrote:

> I'd like to see a March 1 release,

I don't think we're going to get there.  I'm spending today trying to fix 
3.3 PRs, but there are too many remaining.  I think we need to fix many 
more regressions before the release, and I'm not sure how quickly we can 
get there.

We are making progress, and I'd really like to thank the people who are 
taking bug reports and making small test cases out of them in GNATS. 
That's making it *tons* easier to deal with the problems.

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: 3.3 target date?

[Dan Kegel]

Mark Mitchell wrote:
> We are making progress, and I'd really like to thank the people who are 
> taking bug reports and making small test cases out of them in GNATS. 
> That's making it *tons* easier to deal with the problems.

Is somebody sweeping those up into the regression test suite?
- Dan

Re: 3.3 target date?

[Mark Mitchell]

--On Thursday, February 20, 2003 02:35:27 PM -0800 Dan Kegel 
<dkegel@ixiacom.com> wrote:

> Mark Mitchell wrote:
>> We are making progress, and I'd really like to thank the people who are
>> taking bug reports and making small test cases out of them in GNATS.
>> That's making it *tons* easier to deal with the problems.
>
> Is somebody sweeping those up into the regression test suite?

They go there when the bugs are fixed, which makes sense.

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com