[Bug bootstrap/55388] New: [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[Bug bootstrap/55388] New: [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

             Bug #: 55388
           Summary: [4.8 regression] ICE in int_mode_for_mode at
                    stor-layout.c:423 breaks sparc64-linux bootstrap
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: mikpe@it.uu.se


Attempting to bootstrap gcc-4.8-20121118 on sparc64-linux fails with:

make[5]: Entering directory
`/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/libsupc++'
/bin/sh ../libtool --tag CXX --tag disable-shared   --mode=compile
/mnt/scratch/objdir48/./gcc/xgcc -shared-libgcc -B/mnt/scratch/objdir48/./gcc
-nostdinc++ -L/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/src
-L/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/src/.libs
-B/mnt/scratch/install48/sparc-unknown-linux/bin/
-B/mnt/scratch/install48/sparc-unknown-linux/lib/ -isystem
/mnt/scratch/install48/sparc-unknown-linux/include -isystem
/mnt/scratch/install48/sparc-unknown-linux/sys-include   
-I/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/../libgcc
-I/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/include/sparc-unknown-linux
-I/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/include
-I/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/libsupc++  -prefer-pic
-D_GLIBCXX_SHARED  -Wall -Wextra -Wwrite-strings -Wcast-qual -Wabi 
-fdiagnostics-show-location=once    -ffunction-sections -fdata-sections 
-frandom-seed=class_type_info.lo -g -O2 -D_GNU_SOURCE  -c -o class_type_info.lo
/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/libsupc++/class_type_info.cc
libtool: compile:  /mnt/scratch/objdir48/./gcc/xgcc -shared-libgcc
-B/mnt/scratch/objdir48/./gcc -nostdinc++
-L/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/src
-L/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/src/.libs
-B/mnt/scratch/install48/sparc-unknown-linux/bin/
-B/mnt/scratch/install48/sparc-unknown-linux/lib/ -isystem
/mnt/scratch/install48/sparc-unknown-linux/include -isystem
/mnt/scratch/install48/sparc-unknown-linux/sys-include
-I/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/../libgcc
-I/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/include/sparc-unknown-linux
-I/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/include
-I/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/libsupc++ -D_GLIBCXX_SHARED -Wall
-Wextra -Wwrite-strings -Wcast-qual -Wabi -fdiagnostics-show-location=once
-ffunction-sections -fdata-sections -frandom-seed=class_type_info.lo -g -O2
-D_GNU_SOURCE -c
/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/libsupc++/class_type_info.cc  -fPIC
-DPIC -D_GLIBCXX_SHARED -o class_type_info.o
/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/libsupc++/class_type_info.cc: In
member function 'virtual bool __cxxabiv1::__class_type_info::__do_upcast(const
__cxxabiv1::__class_type_info*, void**) const':
/mnt/scratch/gcc-4.8-20121118/libstdc++-v3/libsupc++/class_type_info.cc:46:6:
internal compiler error: in int_mode_for_mode, at stor-layout.c:423
 bool __class_type_info::
      ^
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make[5]: *** [class_type_info.lo] Error 1
make[5]: Leaving directory
`/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3/libsupc++'
make[4]: *** [all-recursive] Error 1
make[4]: Leaving directory
`/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3'
make[3]: *** [all] Error 2
make[3]: Leaving directory
`/mnt/scratch/objdir48/sparc-unknown-linux/libstdc++-v3'
make[2]: *** [all-stage2-target-libstdc++-v3] Error 2
make[2]: Leaving directory `/mnt/scratch/objdir48'
make[1]: *** [stage2-bubble] Error 2
make[1]: Leaving directory `/mnt/scratch/objdir48'
make: *** [bootstrap] Error 2

This problem was not present in 4.8-20121111.

I'll try to derive a minimized test case later today (-ENOTIME right now).

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #1 from Mikael Pettersson <mikpe at it dot uu.se> 2012-11-19 19:32:37 UTC ---
Fails at r193600 so not the same reason as PR55391.  Bisecting...

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #2 from Mikael Pettersson <mikpe at it dot uu.se> 2012-11-20 22:21:45 UTC ---
It's caused by r193599:
http://gcc.gnu.org/ml/gcc-cvs/2012-11/msg00546.html

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

Mikael Pettersson <mikpe at it dot uu.se> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rsandifo at gcc dot gnu.org

--- Comment #3 from Mikael Pettersson <mikpe at it dot uu.se> 2012-11-21 07:52:19 UTC ---
Still fails the same way at r193687.  Adding author to CC:.

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[rsandifo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #4 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> 2012-11-21 18:56:23 UTC ---
Please provide preprocessed source

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #5 from Mikael Pettersson <mikpe at it dot uu.se> 2012-11-21 21:00:00 UTC ---
Created attachment 28757
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28757
preprocessed libstdc++-v3/libsupc++/dyncast.cc

Full configuration options:
/mnt/scratch/gcc-4.8-r193687/configure --prefix=/mnt/scratch/install48
--with-gmp=/home/mikpe/pkgs/linux-sparc64/gmp-5.0.5
--with-mpfr=/home/mikpe/pkgs/linux-sparc64/mpfr-3.1.1
--with-mpc=/home/mikpe/pkgs/linux-sparc64/mpc-1.0.1 --enable-multilib
--disable-plugin --disable-lto --disable-nls --enable-threads=posix
--enable-checking=release --disable-libmudflap --enable-languages=c,c++
--build=sparc-unknown-linux --target=sparc-unknown-linux --with-cpu=ultrasparc
--enable-targets=all --disable-libsanitizer

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #6 from Mikael Pettersson <mikpe at it dot uu.se> 2012-11-21 21:09:34 UTC ---
Oh, I uploaded dyncast.i because that was the most recent ICE on my screen just
now.  It ICEs with the same message as class_type_info.cc so it should be Ok as
a test case.  Sorry about the confusion.

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[rsandifo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #7 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> 2012-11-21 23:28:47 UTC ---
(In reply to comment #5)
> Created attachment 28757 [details]
> preprocessed libstdc++-v3/libsupc++/dyncast.cc
> 
> Full configuration options:
> /mnt/scratch/gcc-4.8-r193687/configure --prefix=/mnt/scratch/install48
> --with-gmp=/home/mikpe/pkgs/linux-sparc64/gmp-5.0.5
> --with-mpfr=/home/mikpe/pkgs/linux-sparc64/mpfr-3.1.1
> --with-mpc=/home/mikpe/pkgs/linux-sparc64/mpc-1.0.1 --enable-multilib
> --disable-plugin --disable-lto --disable-nls --enable-threads=posix
> --enable-checking=release --disable-libmudflap --enable-languages=c,c++
> --build=sparc-unknown-linux --target=sparc-unknown-linux --with-cpu=ultrasparc
> --enable-targets=all --disable-libsanitizer

Thanks.  After trying and failing to convince it to fail in cross environment,
it finally dawned on me that the failure was during stage 2, so presumably
this is a miscompile of cc1plus.  Sorry for the runaround.

I'll try to reproduce the failure on gcc62, but I'm guessing it will take
a while to reach the failure point.

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-11-22
                 CC|                            |ebotcazou at gcc dot
                   |                            |gnu.org
     Ever Confirmed|0                           |1

--- Comment #8 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-11-22 08:32:51 UTC ---
> Thanks.  After trying and failing to convince it to fail in cross environment,
> it finally dawned on me that the failure was during stage 2, so presumably
> this is a miscompile of cc1plus.  Sorry for the runaround.

Yes, this looks like a miscompilation.

> I'll try to reproduce the failure on gcc62, but I'm guessing it will take
> a while to reach the failure point.

This reproduces only with Mikael's configure options, i.e. in release mode, so
it isn't that long.

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #9 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-11-22 11:02:43 UTC ---
Richard, did you already start to investigate seriously?  If no, I can take it
over if you want, it's SPARC and I have a big share in the problematic patch.

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |ebotcazou at gcc dot
                   |gnu.org                     |gnu.org
   Target Milestone|---                         |4.8.0

--- Comment #10 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-11-22 17:47:38 UTC ---
> It's caused by r193599:
> http://gcc.gnu.org/ml/gcc-cvs/2012-11/msg00546.html

Yep, the problem is that alias.c:nonoverlapping_component_refs_p considers that
accesses to bitfields never overlap with accesses to other components...

I also think that the size handling in set_mem_attributes_minus_bitpos might
need to be further refined, now that bitfields are handled:

       /* Respect mode size.  */
      attrs.size_known_p = defattrs->size_known_p;
      attrs.size = defattrs->size;
      /* ??? Is this really necessary?  We probably should always get
     the size from the type below.  */

I don't see how the final size can be smaller than the mode size.

[Bug bootstrap/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423 breaks sparc64-linux bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #11 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-11-22 18:26:05 UTC ---
In fact the miscompilation is already eliminated by replacing the call to
adjust_address with adjust_bitfield_address in store_bit_field.  Not clear what
I was thinking when I left this one out...

[Bug rtl-optimization/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423

[rsandifo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #12 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> 2012-11-23 07:55:56 UTC ---
(In reply to comment #9)
> Richard, did you already start to investigate seriously?  If no, I can take it
> over if you want, it's SPARC and I have a big share in the problematic patch.

Sorry, was away yesterday, so it would have been "not yet".  But it looks like
you've already pinned it, thanks!

[Bug rtl-optimization/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

--- Comment #13 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-11-23 16:00:35 UTC ---
Author: ebotcazou
Date: Fri Nov 23 16:00:26 2012
New Revision: 193760

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=193760
Log:
    PR rtl-optimization/55388
    * alias.c (nonoverlapping_component_refs_p): Handle bitfields.
    * emit-rtl.c (adjust_address_1): Deal with VOIDmode early.
    * expmed.c (store_bit_field): Turn the call to adjust_address
    into a call to adjust_bitfield_address_size.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/alias.c
    trunk/gcc/emit-rtl.c
    trunk/gcc/expmed.c

[Bug rtl-optimization/55388] [4.8 regression] ICE in int_mode_for_mode at stor-layout.c:423

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55388

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED

--- Comment #14 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-11-23 16:02:20 UTC ---
Fix applied.