[Bug c++/49378] New: [4.7 Regression] C++ is broken

[Bug c++/49378] New: [4.7 Regression] C++ is broken

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

           Summary: [4.7 Regression] C++ is broken
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: hjl.tools@gmail.com


On Fedora 15/x86-64, revision 174953 failed most of C++ testcase
for 32bit:

http://gcc.gnu.org/ml/gcc-testresults/2011-06/msg01336.html

(gdb) r
The program being debugged has been started already.
Start it from the beginning? (y or n) y

Starting program: /export/home/hjl/tmp/bad 

Program received signal SIGSEGV, Segmentation fault.
0x0018d7b8 in (anonymous
namespace)::future_error_category::~future_error_category (this=0x18d7a0,
__in_chrg=<optimized out>)
    at /export/gnu/import/git/gcc/libstdc++-v3/src/future.cc:29
29      struct future_error_category : public std::error_category
(gdb) bt
#0  0x0018d7b8 in (anonymous
namespace)::future_error_category::~future_error_category (this=0x18d7a0,
__in_chrg=<optimized out>)
    at /export/gnu/import/git/gcc/libstdc++-v3/src/future.cc:29
#1  0x00295ec7 in __cxa_finalize () from /lib/libc.so.6
#2  0x001767c4 in __do_global_dtors_aux () from /tmp/libstdc++.so.6
#3  0x001ea860 in _fini () from /tmp/libstdc++.so.6
#4  0x0011f2fd in _dl_fini () from /lib/ld-linux.so.2
#5  0x00295b31 in __run_exit_handlers () from /lib/libc.so.6
#6  0x00295bbd in exit () from /lib/libc.so.6
#7  0x0027d41b in __libc_start_main () from /lib/libc.so.6
#8  0x08048ce1 in _start ()
(gdb) disass
Dump of assembler code for function (anonymous
namespace)::future_error_category::~future_error_category():
   0x0018d7a0 <+0>:    push   %ebx
   0x0018d7a1 <+1>:    call   0x176847 <__i686.get_pc_thunk.bx>
   0x0018d7a6 <+6>:    add    $0x8784e,%ebx
   0x0018d7ac <+12>:    sub    $0x18,%esp
   0x0018d7af <+15>:    mov    %eax,(%esp)
   0x0018d7b2 <+18>:    lea    -0x2bcc(%ebx),%edx
=> 0x0018d7b8 <+24>:    mov    %edx,(%eax)
   0x0018d7ba <+26>:    call   0x173b4c <_ZNSt14error_categoryD2Ev@plt>
   0x0018d7bf <+31>:    add    $0x18,%esp
   0x0018d7c2 <+34>:    pop    %ebx
   0x0018d7c3 <+35>:    ret    
End of assembler dump.
(gdb) 

EAX is never initialized.

[Bug c++/49378] [4.7 Regression] C++ is broken

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

H.J. Lu <hjl.tools at gmail dot com> changed:

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

--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> 2011-06-12 01:15:03 UTC ---
Revision 174951 is OK.  Revision 174952:

http://gcc.gnu.org/ml/gcc-cvs/2011-06/msg00441.html

is the cause.

[Bug middle-end/49378] [4.7 Regression] C++ is broken

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c++                         |middle-end
   Target Milestone|---                         |4.7.0

[Bug middle-end/49378] [4.7 Regression] C++ is broken

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> 2011-06-12 13:36:39 UTC ---
On x86-64, I got

000000000006d270 <_ZN12_GLOBAL__N_121system_error_categoryD0Ev>:
   6d270:       48 8d 05 79 d4 27 00    lea    0x27d479(%rip),%rax        #
2ea6f0 <_ZTVN12_GLOBAL__N_121system_error_categoryE+0x10>
   6d277:       53                      push   %rbx
   6d278:       48 89 fb                mov    %rdi,%rbx
   6d27b:       48 89 07                mov    %rax,(%rdi)
   6d27e:       e8 55 a0 fe ff          callq  572d8
<_ZNSt14error_categoryD2Ev@plt>
   6d283:       48 89 df                mov    %rbx,%rdi
   6d286:       5b                      pop    %rbx
   6d287:       e9 2c 9d fe ff          jmpq   56fb8 <_ZdlPv@plt>
   6d28c:       90                      nop
   6d28d:       90                      nop
   6d28e:       90                      nop
   6d28f:       90                      nop

RAX is also uninitialized.

[Bug middle-end/49378] [4.7 Regression] C++ is broken

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

--- Comment #3 from Jan Hubicka <hubicka at gcc dot gnu.org> 2011-06-12 14:41:18 UTC ---
Hi,
looks like I have the lucky kernel. 

I think the problem is that we now mark as local thunks that are not having
address taken that makes i386 BE to use register passing conventions.

Does the following patch help?

Index: ipa.c
===================================================================
--- ipa.c       (revision 174958)
+++ ipa.c       (working copy)
@@ -120,6 +120,7 @@ static bool
 cgraph_non_local_node_p_1 (struct cgraph_node *node, void *data
ATTRIBUTE_UNUSED)
 {
    return !(cgraph_only_called_directly_or_aliased_p (node)
+           && !ipa_ref_has_aliases_p (&node->ref_list)
            && node->analyzed
            && !DECL_EXTERNAL (node->decl)
            && !node->local.externally_visible
@@ -132,7 +133,11 @@ cgraph_non_local_node_p_1 (struct cgraph
 static bool
 cgraph_local_node_p (struct cgraph_node *node)
 {
-   return !cgraph_for_node_and_aliases (cgraph_function_or_thunk_node (node,
NULL),
+   struct cgraph_node *n = cgraph_function_or_thunk_node (node, NULL);
+   
+   if (n->thunk.thunk_p)
+     return false;
+   return !cgraph_for_node_and_aliases (n,
                                        cgraph_non_local_node_p_1, NULL, true);

 }

[Bug middle-end/49378] [4.7 Regression] C++ is broken

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

--- Comment #4 from H.J. Lu <hjl.tools at gmail dot com> 2011-06-12 14:51:12 UTC ---
(In reply to comment #3)
> Hi,
> looks like I have the lucky kernel. 
> 
> I think the problem is that we now mark as local thunks that are not having
> address taken that makes i386 BE to use register passing conventions.
> 
> Does the following patch help?
> 

Can you upload the patch as an attachment? It is hard to apply
it as a comment. Thanks.

[Bug middle-end/49378] [4.7 Regression] C++ is broken

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

--- Comment #5 from H.J. Lu <hjl.tools at gmail dot com> 2011-06-12 15:15:09 UTC ---
Created attachment 24502
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24502
Preprocessed 32bit future.cc

[Bug middle-end/49378] [4.7 Regression] C++ is broken

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> 2011-06-12 16:39:41 UTC ---
Author: hubicka
Date: Sun Jun 12 16:39:38 2011
New Revision: 174969

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=174969
Log:
    PR middle-end/49378
    * ipa.c (cgraph_non_local_node_p_1, cgraph_local_node_p): Rule out
    aliases and thunks.

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

[Bug middle-end/49378] [4.7 Regression] C++ is broken

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49378

H.J. Lu <hjl.tools at gmail dot com> changed:

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

--- Comment #7 from H.J. Lu <hjl.tools at gmail dot com> 2011-06-23 14:10:02 UTC ---
Fixed.