[Bug middle-end/50598] New: [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[Bug middle-end/50598] New: [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

             Bug #: 50598
           Summary: [4.7 Regression] Undefined symbols: "___emutls_v.*",
                    ... on *-apple-darwin*
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: dominiq@lps.ens.fr
                CC: hubicka@gcc.gnu.org
              Host: *-apple-darwin*
            Target: *-apple-darwin*
             Build: *-apple-darwin*


Since revision 179430 (see
http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg00193.html r179421 is OK)
libgomp.c++/pr24455.C and libgomp.fortran/threadprivate4.f90 fail with

FAIL: libgomp.c++/pr24455.C  -O0  (test for excess errors)
Excess errors:
Undefined symbols:
  "___emutls_v.i", referenced from:
      ___emutls_v.i$non_lazy_ptr in ccB25Iow.o
     (maybe you meant: ___emutls_v.i$non_lazy_ptr)
ld: symbol(s) not found

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jh at suse dot cz

--- Comment #1 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-10-08 20:33:53 UTC ---
Revision 179428 is OK, the failures appear at revision 179429.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #2 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-10-08 21:50:37 UTC ---
Reverting the change for gcc/cgraphunit.c in  revision 179429 fixes the pr.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ro at CeBiTec dot
                   |                            |Uni-Bielefeld.DE

--- Comment #3 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-10-10 13:00:54 UTC ---
Same failures on sparc-sun-solaris2.(8|9), i386-pc-solaris2.9, and
alpha-dec-osf5.1b (see
http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg01107.html and
http://gcc.gnu.org/ml/gcc-testresults/2011-10/msg01105.html ).

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #4 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-10-15 18:36:00 UTC ---
The difference for the test libgomp.c++/pr24455.C comes from the companion file
libgomp.c++/pr24455-1.C. With gcc/cgraphunit.c in revision 179429 reverted,
pr24455-1.s is

[macbook] f90/bug% cat pr24455-1_rev.s
    .globl ___emutls_v.i
    .data
    .align 5
___emutls_v.i:
    .quad    4
    .quad    4
    .quad    0
    .quad    0
    .constructor
    .destructor
    .align 1
    .subsections_via_symbols

while for trunk it is

[macbook] f90/bug% cat pr24455-1_orig.s
    .constructor
    .destructor
    .align 1
    .subsections_via_symbols

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

Iain Sandoe <iains at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-10-24
                 CC|                            |iains at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #5 from Iain Sandoe <iains at gcc dot gnu.org> 2011-10-24 09:12:14 UTC ---
also responsible for a bunch of fails in libstdc++-v3 30_threads/...

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

John David Anglin <danglin at gcc dot gnu.org> changed:

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

--- Comment #6 from John David Anglin <danglin at gcc dot gnu.org> 2011-10-30 17:17:57 UTC ---
Same failure on hppa*-*-hpux*.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #7 from Iain Sandoe <iains at gcc dot gnu.org> 2011-11-03 18:22:05 UTC ---
$ more ../gcc-live-trunk/libgomp/testsuite/libgomp.c++/pr24455-1.C 
// { dg-do compile }
// { dg-require-effective-target tls }
extern int i;
#pragma omp threadprivate (i)

int i;

===
if this is compiled with  "-x c "  we get:

$ more pr24455-1.s         .text
__GLOBAL__sub_I_65535_0_pr24455_1.C:
        pushl   %ebx
        subl    $24, %esp
        call    ___x86.get_pc_thunk.bx
L00000000001$pb:
        movl    $0, 12(%esp)
        movl    $4, 8(%esp)
        movl    $4, 4(%esp)
        movl    L___emutls_v.i$non_lazy_ptr-L00000000001$pb(%ebx), %eax
        movl    %eax, (%esp)
        call    ___emutls_register_common
        addl    $24, %esp
        popl    %ebx
        ret
        .mod_init_func
        .align 2
        .long   __GLOBAL__sub_I_65535_0_pr24455_1.C
        .comm   ___emutls_v.i,16,2
        .section __TEXT,__textcoal_nt,coalesced,pure_instructions
        .weak_definition        ___x86.get_pc_thunk.bx
        .private_extern ___x86.get_pc_thunk.bx
___x86.get_pc_thunk.bx:
        movl    (%esp), %ebx
        ret
        .section __IMPORT,__pointers,non_lazy_symbol_pointers
L___emutls_v.i$non_lazy_ptr:
        .indirect_symbol ___emutls_v.i
        .long   0
        .subsections_via_symbols

====

if compiled with g++ we get an empty body (as per comment #4).

the tree-dumps are not very interesting - because the tu shows as empty ...

I'm guessing that somehow g++ is dropping the extern .. which allows the
revised version of cgraphunit to prune the variable .. but I'm not sure how to
go about debugging this at the moment ...

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #8 from Iain Sandoe <iains at gcc dot gnu.org> 2011-11-04 08:56:36 UTC ---
the following also fails (for both c and c++):

extern __thread int i;

__thread int i;

===
 - producing an empty asm file
===

gcc-4.6 produces:

        .globl ___emutls_v.i
        .data
        .align 2
___emutls_v.i:
        .long   4
        .long   4
        .long   0
        .long   0
       .constructor
        .destructor 
        .align 1
        .subsections_via_symbols

====

I'm wondering if, when the variables are pruned - if they are emutls proxies,
they need to check whether the parent var is prune-able. 

 I looked at tree-emutls.c and it seems that the proxy variable should (in
principle) be inheriting the attributes (including external and public) of the
thread var.   Perhaps they are missed in some update.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #9 from Iain Sandoe <iains at gcc dot gnu.org> 2011-11-04 10:50:36 UTC ---
in tree-emults:
(gdb) call debug_tree(var->decl)
 <var_decl 0x41da24e0 i
    type <integer_type 0x41d17360 int public SI
        size <integer_cst 0x41d0455c constant 32>
        unit size <integer_cst 0x41d04578 constant 4>
        align 32 symtab 0 alias set -1 canonical type 0x41d17360 precision 32
min <integer_cst 0x41d04850 -2147483648> max <integer_cst 0x41d0486c
2147483647>
        pointer_to_this <pointer_type 0x41d17a80>>
    public static tls-global-dynamic SI file ../tests/common-var.C line 4 col
14 size <integer_cst 0x41d0455c 32> unit size <integer_cst 0x41d04578 4>
    align 32 context <translation_unit_decl 0x41d1fec4 D.1589>
    (mem/c/i:SI (symbol_ref:SI ("i") [flags 0x1412] <var_decl 0x41da24e0 i>) [0
i+0 S4 A32])>

===
(gdb) call debug_tree(cdecl)
 <var_decl 0x41da2660 __emutls_v.i
    type <record_type 0x41da2600 __emutls_object BLK
        size <integer_cst 0x41d04a48 constant 128>
        unit size <integer_cst 0x41d04a64 constant 16>
        align 32 symtab 0 alias set -1 canonical type 0x41da2600
        fields <field_decl 0x41d1233c __size type <integer_type 0x41d173c0
unsigned int>
            unsigned SI file (null) line 0 col 0
            size <integer_cst 0x41d0455c constant 32>
            unit size <integer_cst 0x41d04578 constant 4>
            align 32 offset_align 128
            offset <integer_cst 0x41d04594 constant 0>
            bit offset <integer_cst 0x41d04620 constant 0> context <record_type
0x41da2600 __emutls_object> chain <field_decl 0x41d122e0 __align>>>
    public static ignored BLK file ../tests/common-var.C line 4 col 14 size
<integer_cst 0x41d04a48 128> unit size <integer_cst 0x41d04a64 16>
    align 32 context <translation_unit_decl 0x41d1fec4 D.1589> initial
<constructor 0x41d3ac90>
    (mem/s/c:BLK (symbol_ref:SI ("__emutls_v.i") [flags 0x1402] <var_decl
0x41da2660 __emutls_v.i>) [0 __emutls_v.i+0 S16 A32])>

===

there are a lot of calls to cgraph_remove_unreachable_nodes () ... and it seems
that there are two vnodes:

(gdb) call debug_tree(vnode->decl)
 <var_decl 0x41da2660 __emutls_v.i
    type <record_type 0x41da2600 __emutls_object BLK
        size <integer_cst 0x41d04a48 constant 128>
        unit size <integer_cst 0x41d04a64 constant 16>
        align 32 symtab 0 alias set -1 canonical type 0x41da2600
        fields <field_decl 0x41d1233c __size type <integer_type 0x41d173c0
unsigned int>
            unsigned SI file (null) line 0 col 0
            size <integer_cst 0x41d0455c constant 32>
            unit size <integer_cst 0x41d04578 constant 4>
            align 32 offset_align 128
            offset <integer_cst 0x41d04594 constant 0>
            bit offset <integer_cst 0x41d04620 constant 0> context <record_type
0x41da2600 __emutls_object> chain <field_decl 0x41d122e0 __align>>>
    public static ignored BLK file ../tests/common-var.C line 4 col 14 size
<integer_cst 0x41d04a48 128> unit size <integer_cst 0x41d04a64 16>
    align 32 context <translation_unit_decl 0x41d1fec4 D.1589> initial
<constructor 0x41d3ac90>
    (mem/s/c:BLK (symbol_ref:SI ("__emutls_v.i") [flags 0x1402] <var_decl
0x41da2660 __emutls_v.i>) [0 __emutls_v.i+0 S16 A32])>

(gdb) call debug_tree(vnode->decl)
 <var_decl 0x41da24e0 i
    type <integer_type 0x41d17360 int public SI
        size <integer_cst 0x41d0455c constant 32>
        unit size <integer_cst 0x41d04578 constant 4>
        align 32 symtab 0 alias set -1 canonical type 0x41d17360 precision 32
min <integer_cst 0x41d04850 -2147483648> max <integer_cst 0x41d0486c
2147483647>
        pointer_to_this <pointer_type 0x41d17a80>>
    public static tls-global-dynamic SI file ../tests/common-var.C line 4 col
14 size <integer_cst 0x41d0455c 32> unit size <integer_cst 0x41d04578 4>
    align 32 context <translation_unit_decl 0x41d1fec4 D.1589>
    (mem/c/i:SI (symbol_ref:SI ("i") [flags 0x1412] <var_decl 0x41da24e0 i>) [0
i+0 S4 A32])
    value-expr <var_decl 0x41da2660 __emutls_v.i
        type <record_type 0x41da2600 __emutls_object BLK
            size <integer_cst 0x41d04a48 constant 128>
            unit size <integer_cst 0x41d04a64 constant 16>
            align 32 symtab 0 alias set -1 canonical type 0x41da2600 fields
<field_decl 0x41d1233c __size>>
        public static ignored BLK file ../tests/common-var.C line 4 col 14 size
<integer_cst 0x41d04a48 128> unit size <integer_cst 0x41d04a64 16>
        align 32 context <translation_unit_decl 0x41d1fec4 D.1589> initial
<constructor 0x41d3ac90>
        (mem/s/c:BLK (symbol_ref:SI ("__emutls_v.i") [flags 0x1402] <var_decl
0x41da2660 __emutls_v.i>) [0 __emutls_v.i+0 S16 A32])>>

the second seems to be marked as needed and the first is not.

.. not sure where I'm going with the debugging here .. not familiar with this
part of the system.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

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

--- Comment #10 from Jan Hubicka <hubicka at gcc dot gnu.org> 2011-11-09 12:36:03 UTC ---
trying to reproduce this on x86_64... the confusing thing is that emultls don't
seem to fiddle with aliases in any special way.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #11 from Iain Sandoe <iains at gcc dot gnu.org> 2011-11-10 20:52:56 UTC ---
(In reply to comment #10)
> trying to reproduce this on x86_64... the confusing thing is that emultls don't
> seem to fiddle with aliases in any special way.

Well, if there's any input I can make .. specific question to answer .. 

The only other thing I could suggest is a binutils emutls target - e.g.
cris-elf (I think that is now building again).  I suspect that this is broken
for all such targets.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #12 from Iain Sandoe <iains at gcc dot gnu.org> 2011-11-11 11:21:11 UTC ---
BTW, this now blocks bootstrap on x86 darwin because it causes libitm build to
fail.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #13 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-11-11 12:00:29 UTC ---
join the --disable-libitm club, needed for a few platforms

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[hubicka at ucw dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #14 from Jan Hubicka <hubicka at ucw dot cz> 2011-11-13 12:41:03 UTC ---
Hi,
I finally debugged this.
It is an ordering issue - emutls adds new variables but
remove_unreachable_functions expects that visibility pass has been run and
externally_visible flag computed.  I will add add_new_variable for backend use
just like cgraph API have.

Honza

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #15 from Iain Sandoe <iains at gcc dot gnu.org> 2011-11-13 12:51:00 UTC ---
(In reply to comment #14)
> Hi,
> I finally debugged this.

Great! :-)
> It is an ordering issue - emutls adds new variables but
> remove_unreachable_functions expects that visibility pass has been run and
> externally_visible flag computed.  I will add add_new_variable for backend use
> just like cgraph API have.

Just in case I misunderstand your reference to "backend".

I think these vars will only be added by tree-emutls pass.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #16 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-11-13 12:54:11 UTC ---
I have found that reverting the change for gcc/cgraphunit.c in  revision 179429
and doing a simple update is not enough to fix the libstdc++-v3 failures. In
top of that the libstdc++-v3 libraries have to be rebuilt too. The differences
are at least

libstdc++-v3/src/libstdc++-symbols.explist

which does not contain ___emutls_v._ZSt15__once_callable without the revert,
and

libstdc++-v3-mvd/src/mutex.o

[macbook] gcc/build_w% nm x86_64-apple-darwin10.8.0/libstdc++-v3/src/mutex.o
0000000000000080 s EH_frame1
                 U ___emutls_get_address
0000000000000020 D ___emutls_v._ZSt11__once_call
0000000000000000 T ___once_proxy

without the revert and

[macbook] gcc/build_w% nm
x86_64-apple-darwin10.8.0/libstdc++-v3-mvd/src/mutex.o
00000000000000a0 s EH_frame1
                 U ___emutls_get_address
0000000000000040 D ___emutls_v._ZSt11__once_call
0000000000000020 D ___emutls_v._ZSt15__once_callable
0000000000000000 T ___once_proxy

with it.

Is this a missing dependency in the make?

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[jh at suse dot de]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #17 from jh at suse dot de 2011-11-13 13:08:12 UTC ---
> Is this a missing dependency in the make?
Makefiles generally do not track changes in your compiler.  You need  
to fully rebuild then.
I am travelling today, but should have patch ready tomorrow.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[hubicka at ucw dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #18 from Jan Hubicka <hubicka at ucw dot cz> 2011-11-14 12:18:43 UTC ---
Hi,
coud you please verify that the attached patch solves the problem? 

Thanks,
Honza
Index: cgraph.h
===================================================================
--- cgraph.h    (revision 181310)
+++ cgraph.h    (working copy)
@@ -641,6 +641,7 @@ void debug_varpool_node_set (varpool_nod
 void free_varpool_node_set (varpool_node_set);
 void ipa_discover_readonly_nonaddressable_vars (void);
 bool cgraph_comdat_can_be_unshared_p (struct cgraph_node *);
+bool varpool_externally_visible_p (struct varpool_node *, bool);

 /* In predict.c  */
 bool cgraph_maybe_hot_edge_p (struct cgraph_edge *e);
@@ -681,6 +682,7 @@ bool const_value_known_p (tree);
 bool varpool_for_node_and_aliases (struct varpool_node *,
                            bool (*) (struct varpool_node *, void *),
                        void *, bool);
+void varpool_add_new_variable (tree);

 /* Walk all reachable static variables.  */
 #define FOR_EACH_STATIC_VARIABLE(node) \
Index: tree-emutls.c
===================================================================
--- tree-emutls.c    (revision 181310)
+++ tree-emutls.c    (working copy)
@@ -262,7 +262,7 @@ get_emutls_init_templ_addr (tree decl)
   if (DECL_EXTERNAL (to))
     varpool_node (to);
   else
-    varpool_finalize_decl (to);
+    varpool_add_new_variable (to);
   return build_fold_addr_expr (to);
 }

@@ -334,7 +334,7 @@ new_emutls_decl (tree decl, tree alias_o
   if (DECL_EXTERNAL (to))
     varpool_node (to);
   else if (!alias_of)
-    varpool_finalize_decl (to);
+    varpool_add_new_variable (to);
   else 
     varpool_create_variable_alias (to,
                    varpool_node_for_asm
Index: ipa.c
===================================================================
--- ipa.c    (revision 181310)
+++ ipa.c    (working copy)
@@ -647,7 +647,7 @@ cgraph_externally_visible_p (struct cgra

 /* Return true when variable VNODE should be considered externally visible. 
*/

-static bool
+bool
 varpool_externally_visible_p (struct varpool_node *vnode, bool aliased)
 {
   if (!DECL_COMDAT (vnode->decl) && !TREE_PUBLIC (vnode->decl))
Index: varpool.c
===================================================================
--- varpool.c    (revision 181310)
+++ varpool.c    (working copy)
@@ -414,6 +414,20 @@ varpool_finalize_decl (tree decl)
     varpool_assemble_pending_decls ();
 }

+/* Add the variable DECL to the varpool.
+   Unlike varpool_finalize_decl function is intended to be used
+   by middle end and allows insertion of new variable at arbitrary point
+   of compilation.  */
+void
+varpool_add_new_variable (tree decl)
+{
+  struct varpool_node *node;
+  varpool_finalize_decl (decl);
+  node = varpool_node (decl);
+  if (varpool_externally_visible_p (node, false))
+    node->externally_visible = true;
+}
+
 /* Return variable availability.  See cgraph.h for description of individual
    return values.  */
 enum availability

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #20 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-11-14 16:14:36 UTC ---
(In reply to comment #19)
> FAIL: 30_threads/thread/native_handle/typesizes.cc execution test

This test should be disabled on darwin (and probably everywhere except
GNU/Linux)

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #19 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-11-14 16:04:57 UTC ---
> coud you please verify that the attached patch solves the problem?

On x86_64-apple-darwin10, I have only done an update with the patch at revision
181350 (but I have rebuilt libstdc++-v3 and i386/libstdc++-v3 at stage 3;-). I
have tested libgomp and libstdc++-v3 without failures except

FAIL: libgomp.graphite/force-parallel-7.c execution test

and

FAIL: 30_threads/thread/native_handle/typesizes.cc execution test

which are likely unrelated. Full regstrapping scheduled for tonight. Thanks for
the patch.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #21 from Iain Sandoe <iains at gcc dot gnu.org> 2011-11-14 16:30:09 UTC ---
(In reply to comment #18)

> coud you please verify that the attached patch solves the problem? 

(applied, together with c#15 PR51094)

full regstrap incl. Ada and Java, i686-darwin9 results look good.
whilst I don't see the libgomp fail, we still have some regressions/new fails,
but a quick review of the log files suggests that they are unrelated to emuTLS.

full regstrap incl. Ada and Java, powerpc-darwin9 bootstrap (also together with
c#15 PR51094) is complete, testing underway; looking at the terminal output,
good so far.

thanks.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[howarth at nitro dot med.uc.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

Jack Howarth <howarth at nitro dot med.uc.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |howarth at nitro dot
                   |                            |med.uc.edu

--- Comment #22 from Jack Howarth <howarth at nitro dot med.uc.edu> 2011-11-14 16:46:21 UTC ---
(In reply to comment #18)

Bootstraps on x86_64-apple-darwin11 under Xcode 4.2. Still fails libitm
testsuite due to weakref issues.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[hubicka at ucw dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #23 from Jan Hubicka <hubicka at ucw dot cz> 2011-11-14 17:02:22 UTC ---
thank you! What kind of weakref issues you see?

Honza

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #24 from Jan Hubicka <hubicka at gcc dot gnu.org> 2011-11-14 17:32:52 UTC ---
Author: hubicka
Date: Mon Nov 14 17:32:47 2011
New Revision: 181358

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=181358
Log:

    PR middle-end/50598
    * cgraph.h (varpool_externally_visible_p): Declare.
    (varpool_add_new_variable): Declare.
    * tree-emultls.c (get_emutls_init_templ_addr,
    new_emutls_decl): Use varpool_add_new_variable.
    * ipa.c (varpool_externally_visible_p): Export.
    * varpool.c (varpool_add_new_variable): New function.

Modified:
    trunk/gcc/ChangeLog

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[vincenzo.innocente at cern dot ch]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

vincenzo Innocente <vincenzo.innocente at cern dot ch> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |vincenzo.innocente at cern
                   |                            |dot ch

--- Comment #25 from vincenzo Innocente <vincenzo.innocente at cern dot ch> 2011-11-17 11:11:34 UTC ---
confirm solved
cat once.cpp 
#include<mutex>
#include<iostream>

int main() {

  std::once_flag flag;
  std::call_once(flag,[](){std::cout << "hi" << std::endl;});

  return 0;
}
Vincenzos-MacBook-Pro:ExercisesBertinoro09 innocent$ c++ -v
Using built-in specs.
COLLECT_GCC=c++
COLLECT_LTO_WRAPPER=/usr/local/libexec/gcc/x86_64-apple-darwin11.2.0/4.7.0/lto-wrapper
Target: x86_64-apple-darwin11.2.0
Configured with: ./configure --enable-languages=c,c++,fortran
--disable-multilib --disable-bootstrap --enable-lto -disable-libitm
Thread model: posix
gcc version 4.7.0 20111112 (experimental) (GCC) 
Vincenzos-MacBook-Pro:ExercisesBertinoro09 innocent$ c++ -Ofast -pthread
-std=gnu++11 once.cpp 
Undefined symbols for architecture x86_64:
  "___emutls_v._ZSt15__once_callable", referenced from:
      _main in ccCeliaX.o
ld: symbol(s) not found for architecture x86_64
collect2: error: ld returned 1 exit status
Vincenzos-MacBook-Pro:ExercisesBertinoro09 innocent$ c++ -v
Using built-in specs.
COLLECT_GCC=c++
COLLECT_LTO_WRAPPER=/usr/local/libexec/gcc/x86_64-apple-darwin11.2.0/4.7.0/lto-wrapper
Target: x86_64-apple-darwin11.2.0
Configured with: ./configure --enable-languages=c,c++,fortran
--disable-multilib --disable-bootstrap --enable-lto -disable-libitm
Thread model: posix
gcc version 4.7.0 20111117 (experimental) (GCC) 
Vincenzos-MacBook-Pro:ExercisesBertinoro09 innocent$ c++ -Ofast -pthread
-std=gnu++11 once.cpp

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #26 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-11-18 00:16:48 UTC ---
(In reply to comment #20)
> (In reply to comment #19)
> > FAIL: 30_threads/thread/native_handle/typesizes.cc execution test
> 
> This test should be disabled on darwin (and probably everywhere except
> GNU/Linux)

I've disabled this on darwin

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #27 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-11-20 12:43:05 UTC ---
(In reply to comment #26)
> > > FAIL: 30_threads/thread/native_handle/typesizes.cc execution test
> > 
> > This test should be disabled on darwin (and probably everywhere except
> > GNU/Linux)
>
> I've disabled this on darwin

Thanks.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #28 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-11-20 12:50:39 UTC ---
(In reply to comment #25)
> confirm solved
> cat once.cpp 
> #include<iostream>
>
> int main() {
>
>   std::once_flag flag;
>   std::call_once(flag,[](){std::cout << "hi" << std::endl;});
>
>   return 0;
> }
> ...
> Undefined symbols for architecture x86_64:
>   "___emutls_v._ZSt15__once_callable", referenced from:
>       _main in ccCeliaX.o
> ld: symbol(s) not found for architecture x86_64
> ...

I am not sure to understand comment #25: is the pr fixed or is it still failing
for the above test on x86_64-apple-darwin11 (it passes on
x86_64-apple-darwin10)?

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #29 from Jonathan Wakely <redi at gcc dot gnu.org> 2011-11-20 13:46:58 UTC ---
Vincenzo's comment shows the test fail for this version

> gcc version 4.7.0 20111112 (experimental) (GCC) 

But pass for this one:

> gcc version 4.7.0 20111117 (experimental) (GCC)

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[vincenzo.innocente at cern dot ch]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #30 from vincenzo Innocente <vincenzo.innocente at cern dot ch> 2011-11-20 14:20:14 UTC ---
sorry Dominique not to have been clear, Jonathan answer is correct.
For what I'm concerned this specific PR can be closed.

vincenzo

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

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

--- Comment #31 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-11-20 15:10:51 UTC ---
> For what I'm concerned this specific PR can be closed.

So closing as fixed.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[whatmannerofburgeristhis at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

Matt Arsenault <whatmannerofburgeristhis at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |whatmannerofburgeristhis at
                   |                            |gmail dot com

--- Comment #33 from Matt Arsenault <whatmannerofburgeristhis at gmail dot com> 2012-10-04 00:58:37 UTC ---
I'm seeing this problem on 4.7.2 when using c++11 packaged_task. The same code
worked yesterday with 4.7.1 before I updated.

[Bug middle-end/50598] [4.7 Regression] Undefined symbols: "___emutls_v.*", ... on *-apple-darwin*

[whatmannerofburgeristhis at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50598

--- Comment #32 from Matt Arsenault <whatmannerofburgeristhis at gmail dot com> 2012-10-04 00:57:42 UTC ---
Created attachment 28350
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28350
Test case using packaged_task