[basic-improvements] bootstrap failure on x86

[basic-improvements] bootstrap failure on x86

[Diego Novillo]

While bootstrapping a bib -> tree-ssa merge yesterday, I ran into
a build failure on libjava.  Zack didn't run into this failure,
but I have now reproduced it on the b-i branch.

Host CC:  gcc (GCC) 3.1.1
Host CXX: g++ (GCC) 3.1.1

Configured with:
$ /home/dnovillo/gcc-3_4-basic-improvements-branch/src/configure --prefix=/home/dnovillo/gcc-3_4-basic-improvements-branch/native --srcdir=/home/dnovillo/gcc-3_4-basic-improvements-branch/src

The failure is in building java/io/ObjectInputStream.o:

-----------------------------------------------------------------------------
$ /notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/gcc/gcj -B/notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/i686-pc-linux-gnu/libjava/ -B/notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/gcc/ --encoding=UTF-8 -fclasspath= -fbootclasspath=/notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/i686-pc-linux-gnu/libjava -ffloat-store -g -O2 -MD -MT java/io/ObjectInputStream.lo -MF java/io/ObjectInputStream.d -c /home/dnovillo/gcc-3_4-basic-improvements-branch/src/libjava/java/io/ObjectInputStream.java -o java/io/ObjectInputStream.o -v -save-temps
Reading specs from /notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/gcc/specs
Reading specs from /notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/i686-pc-linux-gnu/libjava/libgcj.spec
rename spec lib to liborig
Configured with: /home/dnovillo/gcc-3_4-basic-improvements-branch/src/configure --prefix=/home/dnovillo/gcc-3_4-basic-improvements-branch/native --srcdir=/home/dnovillo/gcc-3_4-basic-improvements-branch/src
Thread model: posix
gcc version 3.4-bi 20020930 (experimental)
 /notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/gcc/jc1 /home/dnovillo/gcc-3_4-basic-improvements-branch/src/libjava/java/io/ObjectInputStream.java -fhash-synchronization -fno-use-divide-subroutine -fuse-boehm-gc -fnon-call-exceptions -fkeep-inline-functions -quiet -dumpbase ObjectInputStream.java -auxbase-strip java/io/ObjectInputStream.o -g -O2 -version -fencoding=UTF-8 -fclasspath= -fbootclasspath=/notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/i686-pc-linux-gnu/libjava -ffloat-store -MD -MT java/io/ObjectInputStream.lo -MF java/io/ObjectInputStream.d -o ObjectInputStream.s
GNU Java version 3.4-bi 20020930 (experimental) (i686-pc-linux-gnu)
        compiled by GNU C version 3.4-bi 20020930 (experimental).
Class path starts here:
    ./
    /notnfs/dnovillo/BLD-gcc-3_4-basic-improvements-branch-native.tobiano/i686-pc-linux-gnu/libjava/ (system)
/home/dnovillo/gcc-3_4-basic-improvements-branch/src/libjava/java/io/ObjectInputStream.java: In class `java.io.ObjectInputStream':
/home/dnovillo/gcc-3_4-basic-improvements-branch/src/libjava/java/io/ObjectInputStream.java: In method `java.io.ObjectInputStream.callReadMethod(java.lang.Object,java.lang.Class)':
/home/dnovillo/gcc-3_4-basic-improvements-branch/src/libjava/java/io/ObjectInputStream.java:1390: error: Attempt to delete prologue/epilogue insn:
(insn 556 555 557 27 (nil) (set (reg:SI 5 edi)
        (mem:SI (plus:SI (reg/f:SI 6 ebp)
                (const_int -4 [0xfffffffc])) [0 S4 A8])) -1 (nil)
    (nil))
/home/dnovillo/gcc-3_4-basic-improvements-branch/src/libjava/java/io/ObjectInputStream.java:1390: internal compiler error: in propagate_one_insn, at flow.c:1634
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
-----------------------------------------------------------------------------

Can anybody else reproduce this failure?


Thanks.  Diego.

Re: [basic-improvements] bootstrap failure on x86

[Zack Weinberg]

On Thu, Oct 03, 2002 at 12:55:01PM -0400, Diego Novillo wrote:
> While bootstrapping a bib -> tree-ssa merge yesterday, I ran into
> a build failure on libjava.  Zack didn't run into this failure,
> but I have now reproduced it on the b-i branch.

I'm now getting this failure too.  If no one beats me to it I will
have a look eventually; however, I want to concentrate on regression
fixes in 3.2 and 3.3 for the moment.

zw

Re: [basic-improvements] bootstrap failure on x86

[Diego Novillo]

On Fri, 04 Oct 2002, Zack Weinberg wrote:

> On Thu, Oct 03, 2002 at 12:55:01PM -0400, Diego Novillo wrote:
> > While bootstrapping a bib -> tree-ssa merge yesterday, I ran into
> > a build failure on libjava.  Zack didn't run into this failure,
> > but I have now reproduced it on the b-i branch.
> 
> I'm now getting this failure too.  If no one beats me to it I will
> have a look eventually; however, I want to concentrate on regression
> fixes in 3.2 and 3.3 for the moment.
> 
I hadn't built b-i-b before, but since I didn't get this failure
on mainline, maybe this could be cured by another merge.

I have moved the SPEC2000 daily builds to my public page.  Expect
to see daily results for b-i-b and tree-ssa in the next couple of
days at http://people.redhat.com/dnovillo/spec2000/


Diego.

Re: [basic-improvements] bootstrap failure on x86

[Zack Weinberg]

On Fri, Oct 04, 2002 at 04:14:05PM -0400, Diego Novillo wrote:
> On Fri, 04 Oct 2002, Zack Weinberg wrote:
> > I'm now getting this failure too.  If no one beats me to it I will
> > have a look eventually; however, I want to concentrate on regression
> > fixes in 3.2 and 3.3 for the moment.
> > 
> I hadn't built b-i-b before, but since I didn't get this failure
> on mainline, maybe this could be cured by another merge.

I'll try that after the current libstdc++ build failure in mainline
gets addressed.

> I have moved the SPEC2000 daily builds to my public page.  Expect
> to see daily results for b-i-b and tree-ssa in the next couple of
> days at http://people.redhat.com/dnovillo/spec2000/

May I suggest that you change the graph generation scripts to put the
legend outside the graph area? Right now, some of the interesting
lines get lost under the legend.

zw

Re: [basic-improvements] bootstrap failure on x86

[Gabriel Dos Reis]

Zack Weinberg <zack@codesourcery.com> writes:

| On Fri, Oct 04, 2002 at 04:14:05PM -0400, Diego Novillo wrote:
| > On Fri, 04 Oct 2002, Zack Weinberg wrote:
| > > I'm now getting this failure too.  If no one beats me to it I will
| > > have a look eventually; however, I want to concentrate on regression
| > > fixes in 3.2 and 3.3 for the moment.
| > > 
| > I hadn't built b-i-b before, but since I didn't get this failure
| > on mainline, maybe this could be cured by another merge.
| 
| I'll try that after the current libstdc++ build failure in mainline
| gets addressed.

More accurately, it is the front-end that is broken.  We need approval
from another global-write privs maintainer to back out Mark's recent
patch. 

-- Gaby

Re: [basic-improvements] bootstrap failure on x86

[Zack Weinberg]

On Sun, Oct 06, 2002 at 08:52:11PM +0200, Gabriel Dos Reis wrote:
> Zack Weinberg <zack@codesourcery.com> writes:
> 
> | On Fri, Oct 04, 2002 at 04:14:05PM -0400, Diego Novillo wrote:
> | > On Fri, 04 Oct 2002, Zack Weinberg wrote:
> | > > I'm now getting this failure too.  If no one beats me to it I will
> | > > have a look eventually; however, I want to concentrate on regression
> | > > fixes in 3.2 and 3.3 for the moment.
> | > > 
> | > I hadn't built b-i-b before, but since I didn't get this failure
> | > on mainline, maybe this could be cured by another merge.
> | 
> | I'll try that after the current libstdc++ build failure in mainline
> | gets addressed.
> 
> More accurately, it is the front-end that is broken.  We need approval
> from another global-write privs maintainer to back out Mark's recent
> patch. 

I know.  The point is that it would be silly to do a merge now and
propagate the problem to another branch.

zw

Re: [basic-improvements] bootstrap failure on x86

[Gabriel Dos Reis]

Zack Weinberg <zack@codesourcery.com> writes:

| On Sun, Oct 06, 2002 at 08:52:11PM +0200, Gabriel Dos Reis wrote:
| > Zack Weinberg <zack@codesourcery.com> writes:
| > 
| > | On Fri, Oct 04, 2002 at 04:14:05PM -0400, Diego Novillo wrote:
| > | > On Fri, 04 Oct 2002, Zack Weinberg wrote:
| > | > > I'm now getting this failure too.  If no one beats me to it I will
| > | > > have a look eventually; however, I want to concentrate on regression
| > | > > fixes in 3.2 and 3.3 for the moment.
| > | > > 
| > | > I hadn't built b-i-b before, but since I didn't get this failure
| > | > on mainline, maybe this could be cured by another merge.
| > | 
| > | I'll try that after the current libstdc++ build failure in mainline
| > | gets addressed.
| > 
| > More accurately, it is the front-end that is broken.  We need approval
| > from another global-write privs maintainer to back out Mark's recent
| > patch. 
| 
| I know.  The point is that it would be silly to do a merge now and
| propagate the problem to another branch.

I'm not suggesting a silly merge; my point is that the problem isn't a
libstdc++ build failure: It is a badly broken cc1plus problem.

-- Gaby

Re: [basic-improvements] bootstrap failure on x86

[Richard Henderson]

On Sun, Oct 06, 2002 at 08:52:11PM +0200, Gabriel Dos Reis wrote:
> More accurately, it is the front-end that is broken.  We need approval
> from another global-write privs maintainer to back out Mark's recent
> patch. 

I was hoping that one of the other C++ folks would be able
to look at the failure and figure out what was wrong with
the small patch.

Technically Mark's got like 12 hours left on the 48 to 
fix the patch before auto-reversion, but it sounds like he
hasn't checked in since he left on his trip.  I guess it's
probably best to go ahead and revert the thing now so that
folks can get work done.


r~

Re: [basic-improvements] bootstrap failure on x86

[Diego Novillo]

On Sun, 06 Oct 2002, Zack Weinberg wrote:

> May I suggest that you change the graph generation scripts to put the
> legend outside the graph area? Right now, some of the interesting
> lines get lost under the legend.
> 
Good point.  They're now generated below the graph.  Thanks.


Diego.

Reverting Mark's cp patch (Was: Re: [basic-improvements] bootstrap failure on x86)

[Andreas Jaeger]

Richard Henderson <rth@redhat.com> writes:

> On Sun, Oct 06, 2002 at 08:52:11PM +0200, Gabriel Dos Reis wrote:
>> More accurately, it is the front-end that is broken.  We need approval
>> from another global-write privs maintainer to back out Mark's recent
>> patch. 
>
> I was hoping that one of the other C++ folks would be able
> to look at the failure and figure out what was wrong with
> the small patch.
>
> Technically Mark's got like 12 hours left on the 48 to 
> fix the patch before auto-reversion, but it sounds like he
> hasn't checked in since he left on his trip.  I guess it's
> probably best to go ahead and revert the thing now so that
> folks can get work done.

Gaby, Richard,

Can either of you revert the patch, please ?

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

Re: Reverting Mark's cp patch (Was: Re: [basic-improvements] bootstrap failure on x86)

[Richard Henderson]

On Mon, Oct 07, 2002 at 06:49:32AM +0200, Andreas Jaeger wrote:
> Can either of you revert the patch, please ?

Done.

        * decl2.c, pt.c: Revert c++/7754 fix.

I've also fixed up the 3.2 branch wrt HJ's unauthorized commit.
He reverted only part of Mark's patch, which doesn't strike me
as particularly safe.


r~