Re: AIX bootstrap failure

Re: AIX bootstrap failure

[John David Anglin]

I see exactly the same error under hppa-unknown-linux-gnu:

../../../../gcc/libstdc++-v3/src/localename.cc: In member function `void
   std::locale::_Impl::_M_replace_categories(const std::locale::_Impl*,
   unsigned int)':
../../../../gcc/libstdc++-v3/src/localename.cc:219: internal compiler error: in
   int_mode_for_mode, at stor-layout.c:295

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: AIX bootstrap failure

[Loren James Rittle]

In article <200210012304.g91N4lhu028835@hiauly1.hia.nrc.ca> you write:
>I see exactly the same error under hppa-unknown-linux-gnu:

>../../../../gcc/libstdc++-v3/src/localename.cc: In member function `void
>   std::locale::_Impl::_M_replace_categories(const std::locale::_Impl*,
>   unsigned int)':
>../../../../gcc/libstdc++-v3/src/localename.cc:219: internal compiler error: in
>   int_mode_for_mode, at stor-layout.c:295

And exact same on sparc-sun-solaris2.7 yesterday and today.
-- 
Loren J. Rittle, Principal Staff Engineer, Motorola Labs (IL02/2240)
rittle@labs.mot.com, KeyID: 2048/ADCE34A5, FDC0292446937F2A240BC07D42763672

AIX bootstrap failure

[David Edelsohn]

LAST_UPDATED:
Wed May 18 00:10:41 EDT 2005
Wed May 18 04:10:41 UTC 2005

/tmp/powerpc-ibm-aix5.2.0.0-20050518/./gcc/xgcc -B/tmp/powerpc-ibm-aix5.2.0.0-20050518/./gcc/ -B/farm/dje/install/powerpc-ibm-aix5.2.0.0-20050518/powerpc-ibm-aix5.2.0.0/bin/ -B/farm/dje/install/powerpc-ibm-aix5.2.0.0-20050518/powerpc-ibm-aix5.2.0.0/lib/ -isystem /farm/dje/install/powerpc-ibm-aix5.2.0.0-20050518/powerpc-ibm-aix5.2.0.0/include -isystem /farm/dje/install/powerpc-ibm-aix5.2.0.0-20050518/powerpc-ibm-aix5.2.0.0/sys-include -DHAVE_CONFIG_H -I. -I/farm/dje/src/src/libgfortran -I. -iquote/farm/dje/src/src/libgfortran/io -std=gnu99 -Wall -Wstrict-prototypes -Wmissing-prototypes -Wold-style-definition -Wextra -Wwrite-strings -O2 -g -O2 -c /farm/dje/src/src/libgfortran/runtime/main.c   -DPIC -o .libs/main.o
/farm/dje/src/src/libgfortran/runtime/main.c: In function '_GLOBAL__I_0__gfortrani_l8_to_l4_offset':
/farm/dje/src/src/libgfortran/runtime/main.c:119: internal compiler error: Segmentation fault

Re: AIX bootstrap failure

[David Edelsohn]

>>>>> Jerry Quinn writes:


Jerry> David Edelsohn wrote:
>>>>>>> jlquinn  writes:
>> 
>> 
jlquinn> I updated the tree (no local patches), and reconfigured and bootstrapped mainline.  The config is:
jlquinn> ../gcc/configure   --prefix=/home/jlquinn/gcc/dev --with-cpu=power4 --with-gcc-version-trigger=/home/jlquinn/gcc/dev/gcc/gcc/version.c --enable-languages=c,c++,java
>> 
jlquinn> This is on AIX 5.2 pl 3, using xlc.
>> 
jlquinn> Bootstrap comparison failure!
jlquinn> ./fold-const.o differs
>> 
>> I bootstrap with GCC, not XLC, and I have been bootstrapping
>> successfully with a few local patches, but nothing that should affect
>> this type of failure.
>> 
>> My local patches include a declaration of atoll(), which exists on
>> AIX 5.2 but not AIX 5.1, and the header cache from GCC built on AIX 5.1
>> causes problems.  The other patch adds -Wl,-bbigtoc to work around a
>> recently re-introduced TOC overflow, which I am surprised that you did not
>> encounter.
>> 
>> The stage1 compiler has been known to cause bootstrap miscompares,
>> but it is very rare.

Jerry> I refreshed today and got the same bootstrap comparison failure again 
Jerry> with xlc.  Should I file a bug or is bootstrapping from xlc not going to 
Jerry> be supported in the future.

	You probably should file a bug.

	Something seems very strange about your bootstrap because you
should have received a TOC overflow in stage2 and stage3.  If you were
able to bootstrap GCC without any change to address that, something more
fundamental is wrong.

Jerry> Since I can't use xlc, I tried pulling down a gcc 3.3.3 binary for aix 
Jerry> 5.1.  That built the stage 1 compiler but xgcc immediately crashed.  Can 
Jerry> you point me to a binary that I can use to bootstrap with?

	I use GCC to bootstrap mainline.  As I mentioned in an earlier
message, if you use an AIX 5.1 GCC, you need to add a declaration of
atoll().  I add it to aix52.h

#ifndef _AIX52
extern long long int    atoll(const char *);  
#endif

I also add -bbigtoc to the linker flags in t-aix52:

BOOT_LDFLAGS = -Wl,-bbigtoc

David

Re: AIX bootstrap failure

[Jerry Quinn]

David Edelsohn wrote:
>>>>>>jlquinn  writes:
> 
> 
> jlquinn> I updated the tree (no local patches), and reconfigured and bootstrapped mainline.  The config is:
> jlquinn> ../gcc/configure   --prefix=/home/jlquinn/gcc/dev --with-cpu=power4 --with-gcc-version-trigger=/home/jlquinn/gcc/dev/gcc/gcc/version.c --enable-languages=c,c++,java
> 
> jlquinn> This is on AIX 5.2 pl 3, using xlc.
> 
> jlquinn> Bootstrap comparison failure!
> jlquinn> ./fold-const.o differs
> 
> 	I bootstrap with GCC, not XLC, and I have been bootstrapping
> successfully with a few local patches, but nothing that should affect
> this type of failure.
> 
> 	My local patches include a declaration of atoll(), which exists on
> AIX 5.2 but not AIX 5.1, and the header cache from GCC built on AIX 5.1
> causes problems.  The other patch adds -Wl,-bbigtoc to work around a
> recently re-introduced TOC overflow, which I am surprised that you did not
> encounter.
> 
> 	The stage1 compiler has been known to cause bootstrap miscompares,
> but it is very rare.

I refreshed today and got the same bootstrap comparison failure again 
with xlc.  Should I file a bug or is bootstrapping from xlc not going to 
be supported in the future.

Since I can't use xlc, I tried pulling down a gcc 3.3.3 binary for aix 
5.1.  That built the stage 1 compiler but xgcc immediately crashed.  Can 
you point me to a binary that I can use to bootstrap with?

Thanks,
Jerry

Re: AIX bootstrap failure

[David Edelsohn]

>>>>> jlquinn  writes:

jlquinn> I updated the tree (no local patches), and reconfigured and bootstrapped mainline.  The config is:
jlquinn> ../gcc/configure   --prefix=/home/jlquinn/gcc/dev --with-cpu=power4 --with-gcc-version-trigger=/home/jlquinn/gcc/dev/gcc/gcc/version.c --enable-languages=c,c++,java

jlquinn> This is on AIX 5.2 pl 3, using xlc.

jlquinn> Bootstrap comparison failure!
jlquinn> ./fold-const.o differs

	I bootstrap with GCC, not XLC, and I have been bootstrapping
successfully with a few local patches, but nothing that should affect
this type of failure.

	My local patches include a declaration of atoll(), which exists on
AIX 5.2 but not AIX 5.1, and the header cache from GCC built on AIX 5.1
causes problems.  The other patch adds -Wl,-bbigtoc to work around a
recently re-introduced TOC overflow, which I am surprised that you did not
encounter.

	The stage1 compiler has been known to cause bootstrap miscompares,
but it is very rare.

David

Re: AIX bootstrap failure

[jlquinn]

I updated the tree (no local patches), and reconfigured and bootstrapped mainline.  The config is:

../gcc/configure   --prefix=/home/jlquinn/gcc/dev --with-cpu=power4 --with-gcc-version-trigger=/home/jlquinn/gcc/dev/gcc/gcc/version.c --enable-languages=c,c++,java

This is on AIX 5.2 pl 3, using xlc.

Jerry Quinn

Comparing stage2 and stage3 of the compiler
make[1]: Entering directory `/home/jlquinn/gcc/dev/build/gcc'
rm -f .bad_compare
case "slowcompare" in *compare | *compare-lean ) stage=2 ;; * ) stage=`echo slowcompare | sed -e 's,^[a-z]*compare\([0-9][0-9]*\).*,\1,'` ;; esac; \
for dir in . cp java; do \
  if [ "`echo $dir/*.o`" != "$dir/*.o" ] ; then \
    for file in $dir/*.o; do \
      case "slowcompare" in \
        slowcompare* ) \
          tail +16c ./$file > tmp-foo1; \
          tail +16c stage$stage/$file > tmp-foo2 \
            && (cmp tmp-foo1 tmp-foo2 > /dev/null 2>&1 || echo $file differs >>
.bad_compare) || true; \
          ;; \
        fastcompare* ) \
          cmp $file stage$stage/$file 16 16 > /dev/null 2>&1; \
          test $? -eq 1 && echo $file differs >> .bad_compare || true; \
          ;; \
        gnucompare* ) \
          cmp --ignore-initial=16 $file stage$stage/$file > /dev/null 2>&1; \
          test $? -eq 1 && echo $file differs >> .bad_compare || true; \
          ;; \
      esac ; \
    done; \
  else true; fi; \
done
rm -f tmp-foo*
case "slowcompare" in *compare | *compare-lean ) stage=2 ;; * ) stage=`echo slowcompare | sed -e 's,^[a-z]*compare\([0-9][0-9]*\).*,\1,'` ;; esac; \
if [ -f .bad_compare ]; then \
  echo "Bootstrap comparison failure!"; \
  cat .bad_compare; \
  exit 1; \
else \
  case "slowcompare" in \
    *-lean ) rm -rf stage$stage ;; \
    *) ;; \
  esac; true; \
fi
Bootstrap comparison failure!
./fold-const.o differs
make[1]: *** [slowcompare] Error 1
make[1]: Leaving directory `/home/jlquinn/gcc/dev/build/gcc'
make: *** [bootstrap] Error 2

AIX bootstrap failure

[David Edelsohn]

	AIX has not been able to bootstrap for the past three days due to
various RTL check failures and now a segmentation fault.  The number of
simultaneous problems is getting overwhelming.

David

Re: AIX bootstrap failure

[Mark Mitchell]

--On Tuesday, October 01, 2002 07:03:19 PM -0400 David Edelsohn 
<dje@watson.ibm.com> wrote:

> 	I don't need any optimization enabled, just -maix64:

Good; I can reproduce it now.

Thanks,

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

Re: AIX bootstrap failure

[David Edelsohn]

	I don't need any optimization enabled, just -maix64:

$ ./xgcc -B./ -O0 -maix64 -S localename.ii
/gcc/dje/src/libstdc++-v3/src/localename.cc: In member function `void 
   std::locale::_Impl::_M_replace_categories(const std::locale::_Impl*, 
   unsigned int)':
/gcc/dje/src/libstdc++-v3/src/localename.cc:219: internal compiler error: in 
   int_mode_for_mode, at stor-layout.c:295

David

AIX bootstrap failure

[Mark Mitchell]

Your test case does not crash my powerpc-ibm-aix4.3.3.0 cross compiler
built on i686-pc-linux-gnu:

base-2.05$ ./cc1plus -quiet ~/tmp/test.C
bash-2.05$ ./cc1plus -version
GNU C++ version 3.3 20020930 (experimental) (powerpc-ibm-aix4.3.3.0)
bash-2.05$ wc -l ~/tmp/test.C
  20771 /home/mitchell/tmp/test.C

Unfortunately, I deleted the mail with the options in it accidentally
after you said that no options were required.  If options are required,
would you please resend that information to me?

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

Re: AIX bootstrap failure

[Mark Mitchell]

> 	I already did binary search.  The first try, removing your patch,
> fixes the bootstrap failure.

That settles that.

I'll look at your preprocessed file.

THaks,

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

Re: AIX bootstrap failure

[David Edelsohn]

>>>>> Mark Mitchell writes:

Mark> Based on the nature of the failure and the nature of my changes I do not
Mark> see any particular connection.  But, it is certainly possible.

Mark> If you send me preprocessed source, and a target triplet, I will build
Mark> a cross compiler and try to replicate the problem.

Mark> It would be nice if you would play the CVS binary search game to figure
Mark> out whose bug it is; that way we won't waste time asking the wrong
Mark> engineer to investigate.

	I already did binary search.  The first try, removing your patch,
fixes the bootstrap failure.

David

Re: AIX bootstrap failure

[Mark Mitchell]

--On Tuesday, October 01, 2002 01:44:30 PM -0400 David Edelsohn 
<dje@watson.ibm.com> wrote:

>>>>>> Graham Stott writes:
>
> Graham> This also fails on nearly all the cross compilers I'm building do
> you have any idea Graham> when it got introduced? I've been alway for 3
> weeks and updated my tree yesterday. Graham> so I don't know when it
> first appeared other than sometime in the last 3 weeks.
>
> 	It was introduced sometime yesterday or this morning.  I suspect
> it is due to Mark's patches to cp and stor-layout.c

Based on the nature of the failure and the nature of my changes I do not
see any particular connection.  But, it is certainly possible.

If you send me preprocessed source, and a target triplet, I will build
a cross compiler and try to replicate the problem.

It would be nice if you would play the CVS binary search game to figure
out whose bug it is; that way we won't waste time asking the wrong
engineer to investigate.

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

Re: AIX bootstrap failure

[David Edelsohn]

>>>>> Graham Stott writes:

Graham> This also fails on nearly all the cross compilers I'm building do you have any idea
Graham> when it got introduced? I've been alway for 3 weeks and updated my tree yesterday.
Graham> so I don't know when it first appeared other than sometime in the last 3 weeks.

	It was introduced sometime yesterday or this morning.  I suspect
it is due to Mark's patches to cp and stor-layout.c

David

Re: AIX bootstrap failure

[Graham Stott]

David,

David Edelsohn wrote:
> 	Bootstrap failed on AIX with GCC 3.3 20021001 experimental updated:
> 
> Tue Oct  1 10:45:47 EDT 2002
> Tue Oct  1 14:45:47 UTC 2002
> 
> /gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/gcc/xgcc -shared-libgcc -B/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/gcc/ -nostdinc++ -L/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/src -L/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/src/.libs -B/gcc/dje/install/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/bin/ -B/gcc/dje/install/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/lib/ -isystem /gcc/dje/install/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/include -maix64 -nostdinc++ -I/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/include/powerpc-ibm-aix4.3.3.0 -I/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/include -I/gcc/dje/src/libstdc++-v3/libsupc++ -I/gcc/dje/src/libstdc++-v3/libmath -g -O2 -maix64 -fno-implicit-templates -Wall -Wno-format -W -Wwrite-strings -Winline -fdiagno
s!
> tics-show-location=once -ffunction-sections -fdata-sections -g -c /gcc/dje/src/libstdc++-v3/src/localename.cc   -DPIC -o .libs/localename.o
> 
> /gcc/dje/src/libstdc++-v3/src/localename.cc: In member function `void 
>    std::locale::_Impl::_M_replace_categories(const std::locale::_Impl*, 
>    unsigned int)':
> /gcc/dje/src/libstdc++-v3/src/localename.cc:219: internal compiler error: in 
>    int_mode_for_mode, at stor-layout.c:292
> 
> int_mode_for_mode is being invoked with the argument of GET_MODE (op0)
> 
> op0 is (mem/s/u (reg/f:DI 131) [68 _S_facet_categories A8])
> 
> 

This also fails on nearly all the cross compilers I'm building do you have any idea
when it got introduced? I've been alway for 3 weeks and updated my tree yesterday.
so I don't know when it first appeared other than sometime in the last 3 weeks.

Cheers
Graham

AIX bootstrap failure

[David Edelsohn]

	Bootstrap failed on AIX with GCC 3.3 20021001 experimental updated:

Tue Oct  1 10:45:47 EDT 2002
Tue Oct  1 14:45:47 UTC 2002

/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/gcc/xgcc -shared-libgcc -B/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/gcc/ -nostdinc++ -L/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/src -L/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/src/.libs -B/gcc/dje/install/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/bin/ -B/gcc/dje/install/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/lib/ -isystem /gcc/dje/install/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/include -maix64 -nostdinc++ -I/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/include/powerpc-ibm-aix4.3.3.0 -I/gcc/dje/build/powerpc-ibm-aix4.3.3.0-20021001/powerpc-ibm-aix4.3.3.0/ppc64/libstdc++-v3/include -I/gcc/dje/src/libstdc++-v3/libsupc++ -I/gcc/dje/src/libstdc++-v3/libmath -g -O2 -maix64 -fno-implicit-templates -Wall -Wno-format -W -Wwrite-strings -Winline -fdiagnos!
tics-show-location=once -ffunction-sections -fdata-sections -g -c /gcc/dje/src/libstdc++-v3/src/localename.cc   -DPIC -o .libs/localename.o

/gcc/dje/src/libstdc++-v3/src/localename.cc: In member function `void 
   std::locale::_Impl::_M_replace_categories(const std::locale::_Impl*, 
   unsigned int)':
/gcc/dje/src/libstdc++-v3/src/localename.cc:219: internal compiler error: in 
   int_mode_for_mode, at stor-layout.c:292

int_mode_for_mode is being invoked with the argument of GET_MODE (op0)

op0 is (mem/s/u (reg/f:DI 131) [68 _S_facet_categories A8])

Re: AIX bootstrap failure

[Aldy Hernandez]

>>>>> "Richard" == Richard Kenner <kenner@vlsi1.ultra.nyu.edu> writes:

Newbie question...

 > I would say that needs to be the next step.  Use the normal methods of
 > debugging bootstrap failures.

And what are these methods?

Aldy

Re: AIX bootstrap failure

[Dave Love]

>>>>> "David" == David Edelsohn <dje@watson.ibm.com> writes:

 David> You invoked the bootstrap in some strange way 

I did `./configure; make bootstrap' in a clean top-level egcs
directory, which worked last time I tried it on this system and I
thought was approved.  I admit to not following the instructions
there...  (shouldn't they be ammended?)

 David> because it should not be mixing "cc" with "stage1/xgcc".  I
 David> and others do not have this problem.  AIX "cc" is invoked
 David> without builtin alloca support while GCC provides builting
 David> alloca support.

Indeed.

[I report it as I won't be able to look at what it's up to until next
week.]

Re: AIX bootstrap failure

[Jeffrey A Law]

  In message < rzqen4u5ntj.fsf@djlvig.dl.ac.uk >you write:
  > On rs6000-ibm-aix4.1.4.0 the 311097 snapshot fails to bootstrap with
  > native tools:
  > 
  > 	stage1/xgcc -Bstage1/ -c  -DIN_GCC    -O2 -g   -DHAVE_CONFIG_H     -I. 
  > -I. -I./config rtl.c
  > 	cc  -DIN_GCC    -O2 -g   -DHAVE_CONFIG_H   -o genattr  genattr.o rtl.o 
  > ` case "obstack.o" in ?*) echo obstack.o ;; esac ` ` case "cc"@"alloca.o" i
  > n "cc"@?*) echo alloca.o ;; esac ` ` case "" in ?*) echo  ;; esac `   -lm -
  > lld
  > ld: 0711-317 ERROR: Undefined symbol: .alloca
  > ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more informati
  > on.
  > make: The error code from the last command is 8.
Hmmm, I'm not sure how this is happening.


HOST_CC should have been set to CC in your makefile which should
have made HOST_CC=stage1/xgcc -Bstage1/.

Instead HOST_CC=cc for some unknown reason.

jeff

Re: AIX bootstrap failure

[David Edelsohn]

>>>>> Dave Love writes:

Dave> On rs6000-ibm-aix4.1.4.0 the 311097 snapshot fails to bootstrap with
Dave> native tools:

Dave> stage1/xgcc -Bstage1/ -c  -DIN_GCC    -O2 -g   -DHAVE_CONFIG_H     -I. -I. -I./config rtl.c
Dave> cc  -DIN_GCC    -O2 -g   -DHAVE_CONFIG_H   -o genattr  genattr.o rtl.o ` case "obstack.o" in ?*) echo obstack.o ;; esac ` ` case "cc"@"alloca.o" in "cc"@?*) echo alloca.o ;; esac ` ` case "" in ?*) echo  ;; esac `   -lm -lld

Dave> ld: 0711-317 ERROR: Undefined symbol: .alloca
Dave> ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more information.
Dave> make: The error code from the last command is 8.

	You invoked the bootstrap in some strange way because it should
not be mixing "cc" with "stage1/xgcc".  I and others do not have this
problem.  AIX "cc" is invoked without builtin alloca support while GCC
provides builting alloca support.

David

AIX bootstrap failure

[Dave Love]

On rs6000-ibm-aix4.1.4.0 the 311097 snapshot fails to bootstrap with
native tools:

	stage1/xgcc -Bstage1/ -c  -DIN_GCC    -O2 -g   -DHAVE_CONFIG_H     -I. -I. -I./config rtl.c
	cc  -DIN_GCC    -O2 -g   -DHAVE_CONFIG_H   -o genattr  genattr.o rtl.o ` case "obstack.o" in ?*) echo obstack.o ;; esac ` ` case "cc"@"alloca.o" in "cc"@?*) echo alloca.o ;; esac ` ` case "" in ?*) echo  ;; esac `   -lm -lld
ld: 0711-317 ERROR: Undefined symbol: .alloca
ld: 0711-345 Use the -bloadmap or -bnoquiet option to obtain more information.
make: The error code from the last command is 8.