[Bug debug/48333] New: -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[Bug debug/48333] New: -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48333

           Summary: -fcompare-debug failure (length) - "memmove" x
                    "__builtin_memmove"
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: debug
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: zsojka@seznam.cz
                CC: aoliva@gcc.gnu.org
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu


Created attachment 23798
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=23798
partially reduced testcase

Compiler output:
$ gcc -O -fkeep-inline-functions -fcompare-debug testcase.ii 
testcase.ii: In function ‘std::_Resetiosflags
std::resetiosflags(std::ios_base::fmtflags)’:
testcase.ii:20948:5: warning: extended initializer lists only available with
-std=c++0x or -std=gnu++0x [enabled by default]
testcase.ii: In function ‘std::_Setiosflags
std::setiosflags(std::ios_base::fmtflags)’:
testcase.ii:20966:5: warning: extended initializer lists only available with
-std=c++0x or -std=gnu++0x [enabled by default]
testcase.ii: In function ‘std::_Setbase std::setbase(int)’:
testcase.ii:20984:5: warning: extended initializer lists only available with
-std=c++0x or -std=gnu++0x [enabled by default]
testcase.ii: In function ‘std::_Setfill<_CharT> std::setfill(_CharT)’:
testcase.ii:21010:7: warning: extended initializer lists only available with
-std=c++0x or -std=gnu++0x [enabled by default]
testcase.ii: In function ‘std::_Setprecision std::setprecision(int)’:
testcase.ii:21028:5: warning: extended initializer lists only available with
-std=c++0x or -std=gnu++0x [enabled by default]
testcase.ii: In function ‘std::_Setw std::setw(int)’:
testcase.ii:21046:5: warning: extended initializer lists only available with
-std=c++0x or -std=gnu++0x [enabled by default]
gcc: error: testcase.ii: -fcompare-debug failure (length)

$ diff testcase.*gkd
24071c24071
<         (call (mem:QI (symbol_ref:DI ("memmove") [flags 0x41]  <function_decl
# memmove>) [ memmove S1 A8])
---
>         (call (mem:QI (symbol_ref:DI ("memmove") [flags 0x41]  <function_decl # memmove>) [ __builtin_memmove S1 A8])

Reducing the testcase is very slow. I met this problem several times, but never
with a small testcase.

[Bug debug/48333] -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48333

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-03-30 08:39:16 UTC ---
Interesting ;)

[Bug debug/48333] -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[aoliva at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48333

Alexandre Oliva <aoliva at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2011.03.30 21:45:57
         AssignedTo|unassigned at gcc dot       |aoliva at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #2 from Alexandre Oliva <aoliva at gcc dot gnu.org> 2011-03-30 21:45:57 UTC ---
Mine.  -fcompare-debug fails similarly (fwrite_unlocked vs builtin) on c-pch.c
during bootstrap, too.

[Bug debug/48333] -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[aoliva at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48333

--- Comment #3 from Alexandre Oliva <aoliva at gcc dot gnu.org> 2011-03-31 06:12:22 UTC ---
Created attachment 23832
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=23832
Patch that fixes the bug

It's a garbage collection issue: we're supposed to share mem_attrs involving
__builtin and non-__builtin functions, and we do, unless GC manages to discard
mem_attrs at precisely the worst possible moment, and does so at only one of
the compilations, and we're unlucky to end up inserting a different one next
time it is required.

A more general solution would amount to checking for builtins at dump time, and
discarding the __builtin prefix or noting it as optional or somesuch, but I'm
testing this simpler and hopefully effective solution instead.

[Bug debug/48333] -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48333

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-05-19 10:35:16 UTC ---
Can you please ping this patch?
It makes sense given:
    case tcc_declaration:
      /* Consider __builtin_sqrt equal to sqrt.  */
      return (TREE_CODE (arg0) == FUNCTION_DECL
              && DECL_BUILT_IN (arg0) && DECL_BUILT_IN (arg1)
              && DECL_BUILT_IN_CLASS (arg0) == DECL_BUILT_IN_CLASS (arg1)
              && DECL_FUNCTION_CODE (arg0) == DECL_FUNCTION_CODE (arg1));
in operand_equal_p...

[Bug debug/48333] -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[aoliva at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48333

--- Comment #5 from Alexandre Oliva <aoliva at gcc dot gnu.org> 2011-06-04 10:08:12 UTC ---
Author: aoliva
Date: Sat Jun  4 10:08:09 2011
New Revision: 174636

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=174636
Log:
PR debug/48333
* calls.c (emit_call_1): Prefer the __builtin declaration of
builtin functions.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/calls.c

[Bug debug/48333] -fcompare-debug failure (length) - "memmove" x "__builtin_memmove"

[aoliva at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48333

Alexandre Oliva <aoliva at gcc dot gnu.org> changed:

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

--- Comment #6 from Alexandre Oliva <aoliva at gcc dot gnu.org> 2011-06-04 10:18:16 UTC ---
Fixed