[Bug debug/49887] New: [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[Bug debug/49887] New: [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

           Summary: [4.7 regression] .debug_macro breaks many
                    Solaris/SPARC tests
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: debug
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ro@gcc.gnu.org
                CC: jakub@gcc.gnu.org
              Host: sparc-sun-solaris2.11
            Target: sparc-sun-solaris2.11
             Build: sparc-sun-solaris2.11


A recent Solaris 11/SPARC bootstrap with Sun as showed many testsuite
regressions:
a considerable number of them were gcc.dg/debug and g++.dg/debug tests at 
-gdwarf-2 -g3, but only with Sun as:

FAIL: gcc.dg/debug/20020220-1.c -gdwarf-2 -g3 (test for excess errors)
Excess errors:
Undefined            first referenced
 symbol                  in file
wm4.0.a88b226d1d8abb1951fbf09c0af98055 /var/tmp//ccV4aG6a.o

The failure vanishes with -save-temps, so I had to provide an unlink.so
that just returns success to capture the output files.  It turns out that
in a regular compile, I've got this section in 20020220-1.s:

    .section    ".debug_macro"
.LLdebug_macro0:
    .uahalf    0x4
    .byte    0x2
    .uaword    .LLdebug_line0
    .byte    0x7
    .uaword    .LLdebug_macro1
    .byte    0x3
    .byte    0
    .byte    0x1
    .byte    0x4
    .byte    0
    .section    ".debug_macro%wm4.0.a88b226d1d8abb1951fbf09c0af98055"
    .group   
wm4.0.a88b226d1d8abb1951fbf09c0af98055,".debug_macro%wm4.0.a88b226d1d8abb1951fbf09c0af98055",#comdat
.LLdebug_macro1:

but the group signature symbol isn't emitted, which leads to the undefined
symbol:

> elfdump -g 20020220-1.broken.o

Group Section:  .group%wm4.0.a88b226d1d8abb1951fbf09c0af98055
     index    flags / section         signature symbol
       [0]   [ COMDAT ]               wm4.0.a88b226d1d8abb1951fbf09c0af98055
       [1]   .debug_macro%wm4.0.a88b226d1d8abb1951fbf09c0af98055 [10]
$ nm 20020220-1.o


20020220-1.broken.o:

[Index]   Value      Size      Type  Bind  Other Shndx   Name

[4]     |         0|         0|SECT |LOCL |0    |6      |
[7]     |         0|         0|SECT |LOCL |0    |2      |
[8]     |         0|         0|SECT |LOCL |0    |4      |
[10]    |         0|         0|SECT |LOCL |0    |12     |
[11]    |         0|         0|SECT |LOCL |0    |7      |
[12]    |         0|         0|SECT |LOCL |0    |3      |
[2]     |         0|         0|SECT |LOCL |0    |10     |
[3]     |         0|         0|SECT |LOCL |0    |9      |
[9]     |         0|         0|SECT |LOCL |0    |11     |
[5]     |         0|         0|SECT |LOCL |0    |5      |
[6]     |         0|         0|SECT |LOCL |0    |8      |
[1]     |         0|         0|FILE |LOCL |0    |ABS    |20020220-1.c
[13]    |        16|       400|FUNC |GLOB |0    |2      |bar
[14]    |         0|        16|FUNC |GLOB |0    |2      |foo
[15]    |       416|        20|FUNC |GLOB |0    |2      |main
[16]    |         0|         0|NOTY |GLOB |0    |UNDEF 
|wm4.0.a88b226d1d8abb1951fbf09c0af98055

I still need to find out why this only happens on SPARC, but on x86 even with
Sun as all is fine.

  Rainer

[Bug debug/49887] [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-07-28 14:45:11 UTC ---
So, how are the group signature symbols supposed to be emitted on
SPARC/Solaris?  A dummy symbol of that name somewhere in the section?  Can it
be local, or must it be global symbol?
Would something like:
--- gcc/dwarf2out.c    2011-07-27 23:25:36.000000000 +0200
+++ gcc/dwarf2out.c    2011-07-28 16:43:29.181746188 +0200
@@ -20872,6 +20872,7 @@ output_macinfo (void)
                        DEBUG_MACRO_SECTION_LABEL,
                        ref->lineno);
       ASM_OUTPUT_LABEL (asm_out_file, label);
+      ASM_OUTPUT_LABEL (asm_out_file, IDENTIFIER_POINTER (comdat_key));
       ref->code = 0;
       free (CONST_CAST (char *, ref->info));
       ref->info = NULL;
fix it?  Of course I'd prefer if this was limited to targets where the
assembler isn't capable of handling it correctly...

[Bug debug/49887] [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

--- Comment #2 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-07-28 14:52:35 UTC ---
> --- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-07-28 14:45:11 UTC ---
> So, how are the group signature symbols supposed to be emitted on
> SPARC/Solaris?  A dummy symbol of that name somewhere in the section?  Can it
> be local, or must it be global symbol?

This is usually handled in gcc/config/sol2.c
(solaris_elf_asm_comdat_section, solaris_define_comdat_signature,
solaris_code_end).  I haven't yet checked why this doesn't trigger here.

> Would something like:
> --- gcc/dwarf2out.c    2011-07-27 23:25:36.000000000 +0200
> +++ gcc/dwarf2out.c    2011-07-28 16:43:29.181746188 +0200
> @@ -20872,6 +20872,7 @@ output_macinfo (void)
>                         DEBUG_MACRO_SECTION_LABEL,
>                         ref->lineno);
>        ASM_OUTPUT_LABEL (asm_out_file, label);
> +      ASM_OUTPUT_LABEL (asm_out_file, IDENTIFIER_POINTER (comdat_key));
>        ref->code = 0;
>        free (CONST_CAST (char *, ref->info));
>        ref->info = NULL;
> fix it?  Of course I'd prefer if this was limited to targets where the

Probably.

> assembler isn't capable of handling it correctly...

Me too: this should be local to the Solaris port if at all possible.  I
won't argue if the need to explicitly define the signature symbols is a
bug or just different from gas, though :-)

    Rainer

[Bug debug/49887] [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

--- Comment #3 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-07-29 12:58:25 UTC ---
> --- Comment #2 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-07-28 14:52:35 UTC ---
>> --- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-07-28 14:45:11 UTC ---
>> So, how are the group signature symbols supposed to be emitted on
>> SPARC/Solaris?  A dummy symbol of that name somewhere in the section?  Can it
>> be local, or must it be global symbol?
>
> This is usually handled in gcc/config/sol2.c
> (solaris_elf_asm_comdat_section, solaris_define_comdat_signature,
> solaris_code_end).  I haven't yet checked why this doesn't trigger here.

I see now what's going on: output_macinfo is run from toplev.c (compile_file):

    (*debug_hooks->finish) (main_input_filename);

which runs *after*

    targetm.asm_out.code_end ();

I've no idea what's a good way to resolve this.

    Rainer

[Bug debug/49887] [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-07-29 13:06:13 UTC ---
Emit that in targetm.asm_out.file_end () instead?

[Bug debug/49887] [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

--- Comment #5 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-07-29 16:10:04 UTC ---
> --- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-07-29 13:06:13 UTC ---
> Emit that in targetm.asm_out.file_end () instead?

Seems to work according to a quick test running gcc.dg/debug/debug.exp.
I'll run it through a full bootstrap cycle on both sparc and x86, as and
gas over the weekend.

I think I had some reason not to use TARGET_ASM_FILE_END for the
COMDAT-with-as patch, but cannot remember what it was.

Thanks.
        Rainer

[Bug debug/49887] [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

--- Comment #6 from Rainer Orth <ro at gcc dot gnu.org> 2011-08-01 12:14:28 UTC ---
Author: ro
Date: Mon Aug  1 12:14:21 2011
New Revision: 177020

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=177020
Log:
    PR debug/49887
    * config/sol2.c (solaris_code_end): Rename to solaris_file_end.
    * config/sol2-protos.h: Likewise.
    * config/i386/i386.c (ix86_code_end) [TARGET_SOLARIS]: Don't call
    solaris_code_end.
    * config/i386/sol2.h [!USE_GAS] (TARGET_ASM_FILE_END): Redefine.
    * config/sparc/sparc.c (sparc_file_end) [TARGET_SOLARIS]: Call
    solaris_file_end.
    * config/sparc/sol2.h (TARGET_ASM_CODE_END): Remove.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/i386/i386.c
    trunk/gcc/config/i386/sol2.h
    trunk/gcc/config/sol2-protos.h
    trunk/gcc/config/sol2.c
    trunk/gcc/config/sparc/sol2.h
    trunk/gcc/config/sparc/sparc.c

[Bug debug/49887] [4.7 regression] .debug_macro breaks many Solaris/SPARC tests

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49887

Rainer Orth <ro at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                URL|                            |http://gcc.gnu.org/ml/gcc-p
                   |                            |atches/2011-08/msg00030.htm
                   |                            |l
         Resolution|                            |FIXED
         AssignedTo|unassigned at gcc dot       |ro at gcc dot gnu.org
                   |gnu.org                     |
   Target Milestone|---                         |4.7.0

--- Comment #7 from Rainer Orth <ro at gcc dot gnu.org> 2011-08-01 12:15:58 UTC ---
Fixed for 4.7.0.