[Bug bootstrap/65232] New: [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[Bug bootstrap/65232] New: [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[doko at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

            Bug ID: 65232
           Summary: [5 Regression] bootstrap failure (ICE in
                    change_symbol_block, at varasm.c:1230) on
                    arm-linux-gnueabihf, in libstdc++ stage1
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
          Assignee: unassigned at gcc dot gnu.org
          Reporter: doko at gcc dot gnu.org

seen with r221042

configured with --with-arch=armv7-a --with-fpu=vfpv3-d16 --with-float=hard
--with-mode=thumb

../../../../../src/libstdc++-v3/src/c++98/valarray.cc  -fPIC -DPIC
-D_GLIBCXX_SHARED -o valarray.o
../../../../../src/libstdc++-v3/src/c++98/strstream.cc:417:1: internal compiler
error: in change_symbol_block, at varasm.c:1230
 } // namespace
 ^
0x9dfd7d change_symbol_block
    ../../src/gcc/varasm.c:1230
0x9e0045 make_decl_rtl(tree_node*)
    ../../src/gcc/varasm.c:1346
0x9e79cf do_assemble_alias(tree_node*, tree_node*)
    ../../src/gcc/varasm.c:5608
0x9efc41 varpool_node::assemble_aliases()
    ../../src/gcc/varpool.c:557
0x9efd61 varpool_node::assemble_decl()
    ../../src/gcc/varpool.c:600
0x9f01e9 symbol_table::output_variables()
    ../../src/gcc/varpool.c:750
0x47d2eb symbol_table::compile()
    ../../src/gcc/cgraphunit.c:2360
0x47d3b7 symbol_table::finalize_compilation_unit()
    ../../src/gcc/cgraphunit.c:2436
0x2fb569 cp_write_global_declarations()
    ../../src/gcc/cp/decl2.c:4754
Please submit a full bug report,
with preprocessed source if appropriate.

complete build log at
https://launchpadlibrarian.net/198874381/buildlog_ubuntu-vivid-armhf.gcc-5_5-20150227-1ubuntu11_BUILDING.txt.gz

[Bug bootstrap/65232] [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
                 CC|                            |hubicka at gcc dot gnu.org
   Target Milestone|---                         |5.0

[Bug bootstrap/65232] [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

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

--- Comment #1 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
It looks like this is a failure while building the stage1 libraries, which
means we _should_ be able to reproduce on a cross build.  Would you mind
posting the preprocessed source code for the failed allocator-inst.cc file?

That is, re-run the following line but with -save-temps, and post the .ii file?

/build/buildd/gcc-5-5-20150227/build/./gcc/xgcc -shared-libgcc
-B/build/buildd/gcc-5-5-20150227/build/./gcc -nostdinc++
-L/build/buildd/gcc-5-5-20150227/build/arm-linux-gnueabihf/libstdc++-v3/src
-L/build/buildd/gcc-5-5-20150227/build/arm-linux-gnueabihf/libstdc++-v3/src/.libs
-L/build/buildd/gcc-5-5-20150227/build/arm-linux-gnueabihf/libstdc++-v3/libsupc++/.libs
-B/usr/arm-linux-gnueabihf/bin/ -B/usr/arm-linux-gnueabihf/lib/ -isystem
/usr/arm-linux-gnueabihf/include -isystem /usr/arm-linux-gnueabihf/sys-include
-isystem /build/buildd/gcc-5-5-20150227/build/sys-include
-I/build/buildd/gcc-5-5-20150227/src/libstdc++-v3/../libgcc
-I/build/buildd/gcc-5-5-20150227/build/arm-linux-gnueabihf/libstdc++-v3/include/arm-linux-gnueabihf
-I/build/buildd/gcc-5-5-20150227/build/arm-linux-gnueabihf/libstdc++-v3/include
-I/build/buildd/gcc-5-5-20150227/src/libstdc++-v3/libsupc++ -D_GLIBCXX_SHARED
-fno-implicit-templates -Wall -Wextra -Wwrite-strings -Wcast-qual -Wabi
-fdiagnostics-show-location=once -ffunction-sections -fdata-sections
-frandom-seed=allocator-inst.lo -g -O2 -D_GNU_SOURCE -c
../../../../../src/libstdc++-v3/src/c++98/allocator-inst.cc  -fPIC -DPIC
-D_GLIBCXX_SHARED -o allocator-inst.o

[Bug bootstrap/65232] [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[doko at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

--- Comment #2 from Matthias Klose <doko at gcc dot gnu.org> ---
Created attachment 34900
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34900&action=edit
preprocessed source

preprocessed source, produced with r221076

[Bug bootstrap/65232] [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[doko at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

--- Comment #3 from Matthias Klose <doko at gcc dot gnu.org> ---
reverting r221040 lets the bootstrap succeed.

[Bug bootstrap/65232] [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-02-28
     Ever confirmed|0                           |1

--- Comment #4 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
Confirmed with an x86-64 cross to arm-linux-gnueabihf with:

configure --enable-languages=c++ --disable-bootstrap
--target=arm-linux-gnueabihf

and then:

./cc1plus -quiet strstream.ii -O2 -I./     -fPIC  -fdata-sections

[Bug bootstrap/65232] [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

--- Comment #5 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
For the following symbol:

(symbol_ref:SI ("_ZTCSt9strstream8_So") [flags 0x82] <var_decl 0x7fffeecd4090
_ZTCSt9strstream8_So>)

We fail this assertion:

change_symbol_block (rtx symbol, struct object_block *block)
{
  if (block != SYMBOL_REF_BLOCK (symbol))
    {
      gcc_assert (SYMBOL_REF_BLOCK_OFFSET (symbol) < 0);
      SYMBOL_REF_BLOCK (symbol) = block;
    }
}

The symbol's block offset has been previously set (to 0) in place_block_symbol
here:

      if (snode->alias)
    {
      rtx target = DECL_RTL (snode->ultimate_alias_target ()->decl);

      gcc_assert (MEM_P (target)
              && GET_CODE (XEXP (target, 0)) == SYMBOL_REF
              && SYMBOL_REF_HAS_BLOCK_INFO_P (XEXP (target, 0)));
      target = XEXP (target, 0);
      place_block_symbol (target);
      SYMBOL_REF_BLOCK_OFFSET (symbol) = SYMBOL_REF_BLOCK_OFFSET (target);
      return;
    }

Interestingly, without the patch in r221040, snode->alias is not set, so a
different path is used to set the offset.

Be that as it may, the killer is that change_symbol_block() is being called on
mainline through do_assemble_alias -> make_decl_rtl, because the symbol is
considered an alias which it wasn't in r221040.

Honza?

[Bug bootstrap/65232] [5 Regression] bootstrap failure (ICE in change_symbol_block, at varasm.c:1230) on arm-linux-gnueabihf, in libstdc++ stage1

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65232

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Hmm, it seems we triggered another bug of aliases WRT anchors.  How does the
declarations in question look like? (basically all aliases introduced by
ipa-icf can also be coded by hand).

Is it the path
  /* For a duplicate declaration, we can be called twice on the                 
     same DECL node.  Don't discard the RTL already made.  */                   
  if (DECL_RTL_SET_P (decl))                                                    
...?

Why the declaration is considered duplicate to start with?