Re: GCC 3.1 Issues

Re: GCC 3.1 Issues

[David Ronis]

How about this PR?

5200  Store Motion is broken


David

Re: GCC 3.1 Issues

[Rainer Orth]

Mark Mitchell <mark@codesourcery.com> writes:

> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.

* PR libgcj/5944: libgcj fails to bootstrap on alpha-dec-osf4.0f and
  sparc-sun-solaris2.5.1, a regression from 3.0.5

* PR target/5949: GCJ gets an ICE bootstrapping libgcj on alpha-dec-osf5.1,
  a regression from 3.0.5

* PR bootstrap/5948: bootstrap failure on i386-pc-solaris2.8, a regression
  from 3.0.5

* PR target/5966: many EH failures on alpha-dec-osf5.1, a regression from
  3.0.5

Btw., who may mark those PRs high priority: anyone with GNATS write access
or only global write privs or other maintainers?

Besides, I've started bootstrapping from the tops of the 3.0 and 3.1
branches on various platforms

	http://gcc.gnu.org/ml/java/2002-03/msg00289.html

and started investigating testsuite failures.  How should I best handle
those that I can analyse?  File individual PRs for the various failure
modes found?

	Rainer

-- 
-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University

Re: GCC 3.1 Issues

[Daniel Egger]

Am Mon, 2002-03-11 um 07.04 schrieb Fergus Henderson:

> optimization/5878 "GCC 3.1 breaks on C code generated by the Mercury compiler"
> is a regression from 2.95.3.  This is an ICE in verify_local_live_at_start,
> at flow.c:606; "register 3 died unexpectedly".

I'm seeing lots of ICEs in verify_local_live_at_start in nowadays
gcc on powerpc-linux-gnu when compiling with >-O1. Just FWIW.

-- 
Servus,
       Daniel

Re: GCC 3.1 Issues

[Reichelt]

Hello,

I just filed PR 5921, which is a regression from 3.0.x.

The bug is an ice-on-illegal-code, but it shares the legal part
with PR 4286. Therefore, the real problem might be related to the
common legal part of the code.

In this case, the bug would be quite serious, since this part of the
code was distilled from iostream, so that there's a big chance for every
program to suffer from related problems.

Because of the potential harm, I suggest to put PR 5921/4286 on the
"3.1 issues list".

Greetings,
Volker Reichelt

Re: GCC 3.1 Issues

[Joel Sherrill]

Fergus Henderson wrote:
> 
> On 06-Mar-2002, Mark Mitchell <mark@codesourcery.com> wrote:
> > If you know of regressions in GCC 3.1 relative to previous releases,
> 
> optimization/5878 "GCC 3.1 breaks on C code generated by the Mercury compiler"
> is a regression from 2.95.3.  This is an ICE in verify_local_live_at_start,
> at flow.c:606; "register 3 died unexpectedly".

I need to produce a cut down test can but I have seen a failure in the
same routine for sh-* (coff, elf, rtems, rtemself) when compiling
newlib.

> --
> Fergus Henderson <fjh@cs.mu.oz.au>  |  "I have always known that the pursuit
> The University of Melbourne         |  of excellence is a lethal habit"
> WWW: <http://www.cs.mu.oz.au/~fjh>  |     -- the last words of T. S. Garp.

-- 
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: GCC 3.1 Issues

[Reichelt]

Hi,

the following PRs show regressions relative to previous releases
(checked with gcc 3.1 20020304):

* Regressions from 3.0.4:
5078      ice-on-legal-code
5565      ice-on-illegal-code
5656      ice-on-illegal-code
5657      ice-on-illegal-code
5666      ice-on-illegal-code

* Regressions from 2.95.3 that haven't been fixed in 3.0.x and are
  still present in the 3.1 branch:
4757      rejects-legal
4934      ice-on-legal-code
4954/5052 ice-on-legal-code
4979      ice-on-legal-code
5189      ice-on-legal-code
5571      ice-on-legal-code

In addition there's PR 5665. The code snippet in the audit trail shows a
regression from 3.0.4. But there might be more problems hidden in the
original test example with does not compile with 3.0.x.

Greetings,
Volker Reichelt

Re: GCC 3.1 Issues

[Fergus Henderson]

On 06-Mar-2002, Mark Mitchell <mark@codesourcery.com> wrote:
> If you know of regressions in GCC 3.1 relative to previous releases,

optimization/5878 "GCC 3.1 breaks on C code generated by the Mercury compiler"
is a regression from 2.95.3.  This is an ICE in verify_local_live_at_start,
at flow.c:606; "register 3 died unexpectedly".

-- 
Fergus Henderson <fjh@cs.mu.oz.au>  |  "I have always known that the pursuit
The University of Melbourne         |  of excellence is a lethal habit"
WWW: <http://www.cs.mu.oz.au/~fjh>  |     -- the last words of T. S. Garp.

Re: GCC 3.1 Issues

[Andreas Schwab]

Mark Mitchell <mark@codesourcery.com> writes:

|> If you know of regressions in GCC 3.1 relative to previous releases,
|> or major issues that stand between us and a GCC 3.0 release, please
|> file PRs in GNATS and send the PR numbers to me.

Another regression relative to gcc 3.0: target/5901, miscompiling the
Linux kernel.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE GmbH, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: GCC 3.1 Issues

[Andreas Schwab]

Mark Mitchell <mark@codesourcery.com> writes:

|> If you know of regressions in GCC 3.1 relative to previous releases,
|> or major issues that stand between us and a GCC 3.0 release, please
|> file PRs in GNATS and send the PR numbers to me.

preprocessor/5899 is a regression wrt. gcc 3.0, which breaks building
glibc.  It contains a command where both the dependencies and the macro
list should be produced, for generating the contents of <bits/syscall.h>.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE GmbH, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: GCC 3.1 Issues

[Janis Johnson]

On Wed, Mar 06, 2002 at 10:20:26AM -0800, Mark Mitchell wrote:
> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.

PR 5892: ICE in gen_nop_type during 2.4.17 ia64 kernel build

This is a regression from 3.0.4.

Janis  

Re: GCC 3.1 Issues

[Janis Johnson]

On Wed, Mar 06, 2002 at 10:20:26AM -0800, Mark Mitchell wrote:
> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.

PR 5891: ICE compiling SPEC fp test with -funroll-loops on ia64

This is a regression from GCC 3.0.4.

Janis

Re: GCC 3.1 Issues

[Janis Johnson]

On Wed, Mar 06, 2002 at 10:20:26AM -0800, Mark Mitchell wrote:
> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.

PR 5877: negative offset to array index leads to inefficient code

This happens on ia64, perhaps other targets as well.  It's a regression
from 3.0.x.

Janis

Re: GCC 3.1 Issues

[Andreas Jaeger]

Mark Mitchell <mark@codesourcery.com> writes:

> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.

I have two PRs that might be related:
- PR 5172: A miscompilation of glibc on i686, this is fixed if store
  motion is disable.
- PR 5200: Store motion is broken (the PR is a pointer to some
  messages so that this topic get reviewed before the 3.1 release).

Store motion seems to have some serious bugs (generating wrong code)
and if those are not getting fixed, I propose to disable store motion
completly.

Since store motion is a new feature relative to 3.0, any bugs in store
motion are regressions.

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

Re: GCC 3.1 Issues

[Robert Dewar]

<<Directing bug reports to gcc-bugs@gcc.gnu.org is not optimal since such
bug reports are not filed in a bug tracking system.  Instead users should
be told "See <URL:http://www.gnu.org/software/gcc/bugs.html> for
instructions.", as is done elsewhere in the compiler (and that URL
describes the available means of submitting bug reports, and will be
updated if they change).  Perhaps you could contribute some instructions
to that page, describing how Ada bug reporting differs from that for C?
>>

OK, we will make sure that the sources are appropriately adjusted in both
respects.

Re: GCC 3.1 Issues

[Joseph S. Myers]

On Fri, 8 Mar 2002, Robert Dewar wrote:

> As for the bug box, I am confused by this, the current sources should
> generate a message saying
> 
>                   Write_Str
>                     ("| Please submit bug report by email " &
>                     "to gcc-bugs@gcc.gnu.org.");
>                   End_Line;

As far as the 3.1 release process is concerned, the "current sources" are
those on the GCC 3.1 branch in CVS.  This source code you quote does not
appear there.

Directing bug reports to gcc-bugs@gcc.gnu.org is not optimal since such
bug reports are not filed in a bug tracking system.  Instead users should
be told "See <URL:http://www.gnu.org/software/gcc/bugs.html> for
instructions.", as is done elsewhere in the compiler (and that URL
describes the available means of submitting bug reports, and will be
updated if they change).  Perhaps you could contribute some instructions
to that page, describing how Ada bug reporting differs from that for C?

-- 
Joseph S. Myers
jsm28@cam.ac.uk

Re: GCC 3.1 Issues

[Arati Dikey]

> From: Mark Mitchell <mark at codesourcery dot com> 
 > 
 > If you know of regressions in GCC 3.1 relative to previous releases,
 > or major issues that stand between us and a GCC 3.0 release, please
 > file PRs in GNATS and send the PR numbers to me.

PR:5768

optimization/5768: sh codegen leaves stuff on stack


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Free download of GNUSH tool chain for Hitachi's SH Series.
The following site also offers free support to European customers.
Read more at http://www.kpit.com/products/support.htm.
Latest version of GNUSH is released on Jan 1, 2002.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 

Re: GCC 3.1 Issues

[Robert Dewar]

I agree that the lack of a user manual is blocking to including the GNAT
sources. I will try to address this immediately.

As for the bug box, I am confused by this, the current sources should
generate a message saying

                  Write_Str
                    ("| Please submit bug report by email " &
                    "to gcc-bugs@gcc.gnu.org.");
                  End_Line;

providing that Gnat_Version_Type is set to "FSF    " as it should be for the
sources at the FSF. Perhaps somehow this did not get done. I will check

Re: GCC 3.1 Issues

[Kaveh R. Ghazi]

 > From: Mark Mitchell <mark at codesourcery dot com> 
 > 
 > If you know of regressions in GCC 3.1 relative to previous releases,
 > or major issues that stand between us and a GCC 3.0 release, please
 > file PRs in GNATS and send the PR numbers to me.

PR4128:

RTL checking bootstrap on solaris2.7 fails compiling
libf2c/libF77/l_gt.c

--
Kaveh R. Ghazi			Director of Systems Architecture
ghazi@caip.rutgers.edu		Qwest Global Services

Re: GCC 3.1 Issues

[Toon Moene]

Mark Mitchell wrote:

> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.

This one:

5731	middle-end	g77 SIGSEGVs on simple routine mixing complex, 
			real arithmetic (from BLAS) using -O -ffast-math

Fails on i686, powerpc and alphaev6.

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)

RE: GCC 3.1 Issues

[Hans-Peter Nilsson]

On Fri, 8 Mar 2002, Danny Smith wrote:

> other/5620: [3.1] GCC -save-temps foo.c fails to build foo.o
> http://gcc.gnu.org/ml/gcc-bugs/2002-02/msg00221.html
>
> is a regression from 3.0 and 2.95 on mingw.  Nobody else seems to have
> reported this problem so it maybe it target-specific.

I see it on i686-pc-linux-gnu (to various crosses) too, every
time I do -save-temps and try to inspect the object file. ;-(

Perhaps the synopsis should say "erroneously removes foo.o", not
"fails to build foo.o", since -v says it's built (and I trust
that).

brgds, H-P

RE: GCC 3.1 Issues

[Danny Smith]

other/5620: [3.1] GCC -save-temps foo.c fails to build foo.o
http://gcc.gnu.org/ml/gcc-bugs/2002-02/msg00221.html

is a regression from 3.0 and 2.95 on mingw.  Nobody else seems to have
reported this problem so it maybe it target-specific.

Danny

http://movies.yahoo.com.au - Yahoo! Movies
- Vote for your nominees in our online Oscars pool.

Re: GCC 3.1 Issues

[Hans-Peter Nilsson]

On Thu, 7 Mar 2002, Joel Sherrill wrote:
>   ??     - d30v-elf fails to compile crtbegin.o

PR 4191.

There's a similar failure ("simple config tweak needed") for
sparclite-elf which might be of embedded use.

>   ??     - fr30-elf ICE on crtbegin.o

PR 4192.

>   PR5676 and 3591 - mcore-elf can't build libstdc++ (DUPE PR)

Not really duplicates; one refer to 3.0 and the other 3.1 (or
the trunk if you will).

> ?? indicates ones I haven't dug through the log enough yet to formally
> report and can't find a match by searching for the target in the
> synopsis.

I found those above with
"nquery-pr -H gcc.gnu.org -d gcc -x -y d30v"
"nquery-pr -H gcc.gnu.org -d gcc -x -y fr30"
("what's a web interface?" :-)

brgds, H-P

Re: GCC 3.1 Issues

[Stan Shebs]

Joel Sherrill wrote:
> 
> law@redhat.com wrote:
> >
> > In message <200203072002.PAA27142@makai.watson.ibm.com>, David Edelsohn writes:
> >  >      arc-elf
> >  >      avr-elf
> >  >      d30v-elf
> >  >      fr30-elf
> >  >      mcore-elf
> >  >      mn10200-elf
> >  >      v850-elf
> >  >
> >  > Are these targets ever built in any other configuration?  In other words,
> >  > are the failures ELF-specific for those targets or those targets simply
> >  > have not been maintained in GCC-3?
> > I'm pretty sure the d30, fr30, mn10200 and v850 are all ELF only.
> 
> I don't think they are ELF specific.  Each appears to be target
> specific and a sign of not being maintained.  We had hoped for
> better embedded target support in 3.1 than 3.0 which is why I
> am mentioning these.

Can we *please* retire the d30v, or at least take it off the
try-to-compile list?  I don't know of any evidence that the
chip ever even taped out, let alone got used by anybody.

> I didn't think it worth mentioning but the a29k-coff
> doesn't compile either.

In case people don't know, the a29k has recently gone onto
GDB's obsolete list.  Even AMD now disavows any rumors of
possible association with this architecture... 1/2 :-)

Stan

Re: GCC 3.1 Issues

[Joel Sherrill]

law@redhat.com wrote:
> 
> In message <200203072002.PAA27142@makai.watson.ibm.com>, David Edelsohn writes:
>  >      arc-elf
>  >      avr-elf
>  >      d30v-elf
>  >      fr30-elf
>  >      mcore-elf
>  >      mn10200-elf
>  >      v850-elf
>  >
>  > Are these targets ever built in any other configuration?  In other words,
>  > are the failures ELF-specific for those targets or those targets simply
>  > have not been maintained in GCC-3?
> I'm pretty sure the d30, fr30, mn10200 and v850 are all ELF only.

I don't think they are ELF specific.  Each appears to be target 
specific and a sign of not being maintained.  We had hoped for
better embedded target support in 3.1 than 3.0 which is why I
am mentioning these. 

I have been compiling 39 cross targets and most of the failures
appear to be no more general than to all targets for a CPU. 

I didn't think it worth mentioning but the a29k-coff 
doesn't compile either.

> jeff

-- 
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: GCC 3.1 Issues

[law]

In message <200203072002.PAA27142@makai.watson.ibm.com>, David Edelsohn writes:
 > 	arc-elf
 > 	avr-elf
 > 	d30v-elf
 > 	fr30-elf
 > 	mcore-elf
 > 	mn10200-elf
 > 	v850-elf
 > 
 > Are these targets ever built in any other configuration?  In other words,
 > are the failures ELF-specific for those targets or those targets simply
 > have not been maintained in GCC-3?
I'm pretty sure the d30, fr30, mn10200 and v850 are all ELF only.

jeff

Re: GCC 3.1 Issues

[David Edelsohn]

	arc-elf
	avr-elf
	d30v-elf
	fr30-elf
	mcore-elf
	mn10200-elf
	v850-elf

Are these targets ever built in any other configuration?  In other words,
are the failures ELF-specific for those targets or those targets simply
have not been maintained in GCC-3?

Thanks, David

Re: GCC 3.1 Issues

[Joel Sherrill]

Hi,

I don't know how critical these are but they are BUILD failures
for a handful of embedded targets when only C and C++ are
enabled.   In this email, I am ignoring failures for building 
other languages and test failures.

  PR4792 - ICE building libgcc for arc-elf
  PR3931 - avr-elf can't build libstdc++
  ??     - d30v-elf fails to compile crtbegin.o
  ??     - fr30-elf ICE on crtbegin.o
  PR3589 - hppa1.1-proelf illegal instructions
  PR5676 and 3591 - mcore-elf can't build libstdc++ (DUPE PR)
  PR4195 - mn10200-elf fails building libgcc __ffssi2
           There are two other bootstraps filed on this
           target which might be applicable if it got further. 
  ??      - v850-elf ICE segfault compiling libiberty/regex.c 

?? indicates ones I haven't dug through the log enough yet to formally
report and can't find a match by searching for the target in the
synopsis.

Like Mark said.. filing and fixing 'em are different issues. :)

-- 
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: GCC 3.1 Issues

[Stephane Carrez]

Hi Mark,

For m6811-elf, I've logged the following regression:

5854 initialize_uninitialized_subregs() breaks HC11/HC12 port
     http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5854


It is a showstopper for m6811-elf because it prevents building libgcc.

	Stephane

-----------------------------------------------------------------------
         Home                               Office
E-mail: Stephane.Carrez@worldnet.fr        Stephane.Carrez@sun.com
WWW:    http://stephane.carrez.free.fr     http://www.sun.com

RE: GCC 3.1 Issues

[Billinghurst, David (CRTS)]

This is a g77 regression on irix.

>Number:         5651
>Category:       fortran
>Synopsis:       Optimization (-funroll-loops) error with LAPACK sstebz.f

-----Original Message-----
From: Mark Mitchell [mailto:mark@codesourcery.com]
Sent: Thursday, 7 March 2002 4:20 
To: gcc@gcc.gnu.org
Subject: GCC 3.1 Issues


If you know of regressions in GCC 3.1 relative to previous releases,
or major issues that stand between us and a GCC 3.0 release, please
file PRs in GNATS and send the PR numbers to me.

I would like to start keeping track of these issues -- and fixing them --
as soon as possible.

Thanks!

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

Re: GCC 3.1 Issues

[Eric Botcazou]

> This is a regression from gcc 3.0:
>
> "ICE in emit_move_insn, at expr.c:2748, regression from gcc 3.0"
> http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5656

I posted the description of the problem some time ago, as well as a remedy:
http://gcc.gnu.org/ml/gcc-patches/2002-02/msg01508.html

--
Eric Botcazou
ebotcazou@multimania.com

Re: GCC 3.1 Issues

[Nathan Sidwell]

The C++ regressions that I know of are,

5682 ICE in build_secondary_vtable
5659 default access for class/struct bug
5658 Regression on redefinition of a type in a derived class.
5449 internal error #20000524
4460 seg faults instead of throwing in a complicated inheritance hierarchy
4381 Exceptions virtually inherited from a class cause segmentation fault at run tim
4377 more errors with multiple non-type template parameters
4026 bad code generated with optimization
4003 ICE on template instantiation including friendship declaration.
3882 error referencing a variable in its initializer expression in template code
3870 bogus error specializing a template function
3948 Two destructors are called when no copy destructor is defined
3935 const static class member access
3230 False warning for 'returning reference to temporary'

I've marked these as high priority in gnats.
5682, 4460, 4381 look like ABI bugs.

nathan
-- 
Dr Nathan Sidwell :: Computer Science Department :: Bristol University
           The voices in my head told me to say this
nathan@acm.org  http://www.cs.bris.ac.uk/~nathan/  nathan@cs.bris.ac.uk

Re: GCC 3.1 Issues

[Franz Sirl]

At 19:20 06.03.2002, Mark Mitchell wrote:
>If you know of regressions in GCC 3.1 relative to previous releases,
>or major issues that stand between us and a GCC 3.0 release, please
>file PRs in GNATS and send the PR numbers to me.
>
>I would like to start keeping track of these issues -- and fixing them --
>as soon as possible.

For powerpc-linux-gnu there are currently 2 showstopper bugs (both 
regressions from 2.95.x and gcc-3.0.x):

PR 5693 <http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5693>
PR 5863 <http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5863>

Franz.

Re: GCC 3.1 Issues

[Craig Rodrigues]

Hi,

This is a regression from gcc 3.0:

"ICE in emit_move_insn, at expr.c:2748, regression from gcc 3.0"
http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5656

-- 
Craig Rodrigues        
http://www.gis.net/~craigr    
rodrigc@attbi.com

Re: GCC 3.1 issues

[Brad Lucier]

Re:

> There is the sparcv9 ICE when running the libstdc++ testsuite:
> http://gcc.gnu.org/ml/libstdc++/2002-02/msg00175.html
> 
> I think Brad Lucier opened a GNATS report on it but I don't know the 
> number and can't find it at the moment.

It's 5740.  Your mail to perdue.edu got lost.  (That's a joke, son, I say,
that's a joke...)

3.1 hasn't bootstrapped on sparcv9 since before February 28.

Brad

Re: GCC 3.1 Issues

[Benjamin Kosnik]

There is the sparcv9 ICE when running the libstdc++ testsuite:
http://gcc.gnu.org/ml/libstdc++/2002-02/msg00175.html

I think Brad Lucier opened a GNATS report on it but I don't know the 
number and can't find it at the moment.

From Peter Schmid
http://gcc.gnu.org/ml/libstdc++/2002-03/msg00018.html
PR c++/5504

There also appear to a large number of ICE's in the C++ category of GNATS...

-benjamin

Re: GCC 3.1 Issues

[Craig Rodrigues]

Hi,

The following issues in GNATS are listed as regressions from 3.0:

"Duplicate typedef causes ICE (gcc 3.0 regression)" 
http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5817

"Regression on redefinition of a type in a derived class."
http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5658



The following is listed as a regression from 2.95.3
"Initialization of flexible char array member"
http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&database=gcc&pr=5597
-- 
Craig Rodrigues        
http://www.gis.net/~craigr    
rodrigc@attbi.com

Re: GCC 3.1 Issues

[Tom Tromey]

>>>>> "Joe" == Joe Buck <Joe.Buck@synopsys.com> writes:

Joe> I believe that a number of ports are missing needed pieces for
Joe> libjava, which wasn't part of 2.95.3.  Not sure what other
Joe> factors might be an issue.

I don't understand.  Do you mean there are platforms where libgcj 3.0
worked but 3.1 does not?  I don't know of any like this.

Tom

Re: GCC 3.1 Issues

[Joseph S. Myers]

On Wed, 6 Mar 2002, Mark Mitchell wrote:

> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.
> 
> I would like to start keeping track of these issues -- and fixing them --
> as soon as possible.

I think the GNAT bug box - directing bug reports to the ACT bug tracking
system rather than the public GCC one - must be fixed if the Ada front end
is included in the release (even if not built by default).  ada/5856.

Perhaps the lack of a GNAT user manual also needs fixing.  I haven't
submitted a PR for that.  It was supposed to be fixed in the big GNAT
merge.

-- 
Joseph S. Myers
jsm28@cam.ac.uk

Re: GCC 3.1 Issues

[Mark Mitchell]

> Mark.. Do you care about regressions on cross embedded targets for
> languages other than C/C++.  I ask because I could build almost all
> of the RTEMS targets with all languages enabled with 2.95.3.  But
> so many targets fail to build with 3.0.x when all languages are
> enabled that I quit even trying to build languages other than C/C++.

Yes, I care.  Probably not as much about a code-gen failure on a major
native target, but yes, we still care.

The real question is who is going to fix what you find.  If it's not
you, then I will keep track of the problems, but I'm not sure who will
be able to fix them.

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

Re: GCC 3.1 Issues

[Joel Sherrill]

Joe Buck wrote:
> 
> Joel Sherrill writes:
> 
> > Mark.. Do you care about regressions on cross embedded targets for
> > languages other than C/C++.  I ask because I could build almost all
> > of the RTEMS targets with all languages enabled with 2.95.3.  But
> > so many targets fail to build with 3.0.x when all languages are
> > enabled that I quit even trying to build languages other than C/C++.
> 
> I believe that a number of ports are missing needed pieces for libjava,
> which wasn't part of 2.95.3.  Not sure what other factors might be an issue.

I should have mentioned that "all languages" for 2.95.3 included 
Objective-C, Fortran, Chill, and gcj. The interest and breakage 
factors for each of those could vary.  

> > I am happy to try on the 3.1 branch to generate non-C/C++ reports
> > but if no one cares, I don't want to waste the time. :)
> 
> I care; I'd like to see as much data as possible, because it's the only
> way we know what we need to fix, or what we need to warn the users about.

OK.  I will add this to my todo queue. :)

-- 
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: GCC 3.1 Issues

[Joe Buck]

Joel Sherrill writes:

> Mark.. Do you care about regressions on cross embedded targets for 
> languages other than C/C++.  I ask because I could build almost all
> of the RTEMS targets with all languages enabled with 2.95.3.  But
> so many targets fail to build with 3.0.x when all languages are 
> enabled that I quit even trying to build languages other than C/C++.

I believe that a number of ports are missing needed pieces for libjava,
which wasn't part of 2.95.3.  Not sure what other factors might be an issue.

> I am happy to try on the 3.1 branch to generate non-C/C++ reports
> but if no one cares, I don't want to waste the time. :)

I care; I'd like to see as much data as possible, because it's the only
way we know what we need to fix, or what we need to warn the users about.

Re: GCC 3.1 Issues

[Joel Sherrill]

Mark Mitchell wrote:
> 
> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.
> 
> I would like to start keeping track of these issues -- and fixing them --
> as soon as possible.

Mark.. Do you care about regressions on cross embedded targets for 
languages other than C/C++.  I ask because I could build almost all
of the RTEMS targets with all languages enabled with 2.95.3.  But
so many targets fail to build with 3.0.x when all languages are 
enabled that I quit even trying to build languages other than C/C++.

I am happy to try on the 3.1 branch to generate non-C/C++ reports
but if no one cares, I don't want to waste the time. :)

> Thanks!

No problem.
 
> --
> Mark Mitchell                   mark@codesourcery.com
> CodeSourcery, LLC               http://www.codesourcery.com

-- 
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: GCC 3.1 Issues

[H . J . Lu]

On Wed, Mar 06, 2002 at 10:20:26AM -0800, Mark Mitchell wrote:
> If you know of regressions in GCC 3.1 relative to previous releases,
> or major issues that stand between us and a GCC 3.0 release, please
> file PRs in GNATS and send the PR numbers to me.
> 
> I would like to start keeping track of these issues -- and fixing them --
> as soon as possible.
> 

optimization/5844. See

http://gcc.gnu.org/ml/gcc-bugs/2002-03/msg00192.html



H.J.

GCC 3.1 Issues

[Mark Mitchell]

If you know of regressions in GCC 3.1 relative to previous releases,
or major issues that stand between us and a GCC 3.0 release, please
file PRs in GNATS and send the PR numbers to me.

I would like to start keeping track of these issues -- and fixing them --
as soon as possible.

Thanks!

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