[Bug debug/60060] New: [4.9 Rgression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[Bug debug/60060] New: [4.9 Rgression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

            Bug ID: 60060
           Summary: [4.9 Rgression] lto1: internal compiler error: in
                    add_AT_specification, at dwarf2out.c:4096
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: debug
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dominiq at lps dot ens.fr
                CC: aldyh at gcc dot gnu.org, burnus at gcc dot gnu.org,
                    rguenth at gcc dot gnu.org

At r207473 compiling gfortran.dg/namelist_51.f90 with '-g -flto' gives the
following ICE

lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

Compiling the tests with '-g' or '-g -flto -O3' gives on x86_64-apple-darwin13
the warning

warning: invalid DWARF generated by the compiler: DIE 0x000001f0 has multiple 
AT_calling_convention attributes in
'/var/folders/8q/sh_swgz96r7f5vnn08f7fxr00000gn/T//ccLHCwpM.ltrans0.ltrans.o'.

Note that before r207471 (and after r205679) the ICE was

/opt/gcc/work/gcc/testsuite/gfortran.dg/namelist_51.f90: In function
'readnamelist.1900':
/opt/gcc/work/gcc/testsuite/gfortran.dg/namelist_51.f90:8:0: internal compiler
error: in lto_output_tree, at lto-streamer-out.c:1390
     call readNamelist()

[Bug fortran/60060] [4.9 Rgression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |lto
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-02-05
          Component|debug                       |fortran
   Target Milestone|---                         |4.9.0
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.

[Bug fortran/60060] [4.9 Rgression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
The issue is that we create the DIE for the namelist before we've emitted
DIEs for global vars.  Thus we run into

static dw_die_ref
gen_namelist_decl (tree name, dw_die_ref scope_die, tree item_decls)
{
...
  FOR_EACH_CONSTRUCTOR_VALUE (CONSTRUCTOR_ELTS (item_decls), i, value)
    {
      nml_item_ref_die = lookup_decl_die (value);
      if (!nml_item_ref_die)
        nml_item_ref_die = force_decl_die (value);

which forces the creation of a DIE for mem_nml as context of nxc and
creates a declaration DIE of nxc in that context.

The actual AT_specification is added when we output debuginfo for mem_nml
but then again via emit_debug_global_declarations.

Without LTO we don't call dwarf2out_global_decl on mem_nml::NXC.

I believe that somehow we have to honor TREE_ASM_WRITTEN here to avoid
outputting the decl twice ...?

That said, frontends disconnect on what are "globals" is obviously an
area to improve upon ... as is removing the various langhooks around
this area (write_global_declarations) in the light of cgraph/varpool/symtab.

[Bug fortran/60060] [4.9 Rgression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

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

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

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Another possibility is to simply stream the FEs idea of what is a global ...

[Bug fortran/60060] [4.9 Regression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
wrapup_global_declarations looks like it's a no-op in LTO (DECL_DEFER_OUTPUT
is always false).  I am testing the following which avoids (re-)outputting
debuginfo for local statics.

Index: gcc/lto/lto-lang.c
===================================================================
--- gcc/lto/lto-lang.c  (revision 207658)
+++ gcc/lto/lto-lang.c  (working copy)
@@ -1082,17 +1082,11 @@ lto_write_globals (void)
   if (flag_wpa)
     return;

-  /* Record the global variables.  */
-  vec<tree> lto_global_var_decls = vNULL;
+  /* Output debug info for global variables.  */
   varpool_node *vnode;
   FOR_EACH_DEFINED_VARIABLE (vnode)
-    lto_global_var_decls.safe_push (vnode->decl);
-
-  tree *vec = lto_global_var_decls.address ();
-  int len = lto_global_var_decls.length ();
-  wrapup_global_declarations (vec, len);
-  emit_debug_global_declarations (vec, len);
-  lto_global_var_decls.release ();
+    if (!decl_function_context (vnode->decl))
+      debug_hooks->global_decl (vnode->decl);
 }

 static tree

[Bug fortran/60060] [4.9 Regression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

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

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

--- Comment #5 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
The ICE is gone with the patch in comment 4 without regression. However on
darwin I still get the warnings reported in comment 0. Although this could be
hidden with the suitable dg-prune-output, I'ld prefer to see this fixed also.

[Bug fortran/60060] [4.9 Regression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

--- Comment #6 from Tobias Burnus <burnus at gcc dot gnu.org> ---
(In reply to Dominique d'Humieres from comment #0)
> warning: invalid DWARF generated by the compiler: DIE 0x000001f0 has
> multiple  AT_calling_convention attributes in
> '/var/folders/8q/sh_swgz96r7f5vnn08f7fxr00000gn/T//ccLHCwpM.ltrans0.ltrans.
> o'.

(In reply to Dominique d'Humieres from comment #5)
> The ICE is gone with the patch in comment 4 without regression. However on
> darwin I still get the warnings reported in comment 0.

The calling convention should be (except on sh, cf. sh/sh.c's
TARGET_DWARF_CALLING_CONVENTION) be always DW_CC_normal - which is not emitted
as DWARF symbol. The only exception is Fortran's MAIN__ function, which is set
to DW_CC_program.

I wonder whether you see twice DW_CC_program or some other DW_CC_*. You could
try readelf to find out which AT_calling_convention are generated.

[Bug fortran/60060] [4.9 Regression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Tobias Burnus from comment #6)
> (In reply to Dominique d'Humieres from comment #0)
> > warning: invalid DWARF generated by the compiler: DIE 0x000001f0 has
> > multiple  AT_calling_convention attributes in
> > '/var/folders/8q/sh_swgz96r7f5vnn08f7fxr00000gn/T//ccLHCwpM.ltrans0.ltrans.
> > o'.
> 
> (In reply to Dominique d'Humieres from comment #5)
> > The ICE is gone with the patch in comment 4 without regression. However on
> > darwin I still get the warnings reported in comment 0.
> 
> The calling convention should be (except on sh, cf. sh/sh.c's
> TARGET_DWARF_CALLING_CONVENTION) be always DW_CC_normal - which is not
> emitted as DWARF symbol. The only exception is Fortran's MAIN__ function,
> which is set to DW_CC_program.
> 
> I wonder whether you see twice DW_CC_program or some other DW_CC_*. You
> could try readelf to find out which AT_calling_convention are generated.

I see

 <1><18c>: Abbrev Number: 3 (DW_TAG_subprogram)
    <18d>   DW_AT_name        : (indirect string, offset: 0x175): mem_nml       
    <191>   DW_AT_decl_file   : 1       
    <192>   DW_AT_decl_line   : 4       
    <193>   DW_AT_main_subprogram: 1    
    <193>   DW_AT_calling_convention: 2 (program)
    <194>   DW_AT_inline      : 1       (inlined)
    <195>   DW_AT_main_subprogram: 1    
    <195>   DW_AT_calling_convention: 2 (program)
    <196>   DW_AT_sibling     : <0x50b> 

with -O3 -g -flto and

 <1><185>: Abbrev Number: 2 (DW_TAG_subprogram)
    <186>   DW_AT_name        : (indirect string, offset: 0x1e0): mem_nml       
    <18a>   DW_AT_decl_file   : 1       
    <18b>   DW_AT_decl_line   : 4       
    <18c>   DW_AT_main_subprogram: 1    
    <18c>   DW_AT_calling_convention: 2 (program)
    <18d>   DW_AT_low_pc      : 0x400f45        
    <195>   DW_AT_high_pc     : 0x15    
    <19d>   DW_AT_frame_base  : 1 byte block: 9c        (DW_OP_call_frame_cfa)
    <19f>   DW_AT_GNU_all_tail_call_sites: 1    
    <19f>   DW_AT_main_subprogram: 1    
    <19f>   DW_AT_calling_convention: 2 (program)
    <1a0>   DW_AT_sibling     : <0x2a3> 

with -g

Looks like a different bug - can you please file it separately as this bug
assigned to me is about LTO issues?

add_calling_convention_attribute is called twice.  First via

Breakpoint 5, add_calling_convention_attribute (subr_die=0x7ffff6da1730, decl=
    <function_decl 0x7ffff6d7e700 mem_nml>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:16859
16859     enum dwarf_calling_convention value = DW_CC_normal;
Missing separate debuginfos, use: zypper install
glibc-debuginfo-2.18-4.11.1.x86_64 libgmp10-debuginfo-5.1.3-58.2.x86_64
(gdb) bt
#0  add_calling_convention_attribute (subr_die=0x7ffff6da1730, 
    decl=<function_decl 0x7ffff6d7e700 mem_nml>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:16859
#1  0x00000000008bc477 in gen_subprogram_die (
    decl=<function_decl 0x7ffff6d7e700 mem_nml>, context_die=0x7ffff6d71000)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:18535
#2  0x00000000008c1668 in force_decl_die (
    decl=<function_decl 0x7ffff6d7e700 mem_nml>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20119
#3  0x00000000008c1581 in get_context_die (
    context=<function_decl 0x7ffff6d7e700 mem_nml>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20089
#4  0x00000000008c15e6 in force_decl_die (decl=<var_decl 0x7ffff6d77e40 nxc>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20105
#5  0x00000000008c2fb0 in gen_namelist_decl (
    name=<identifier_node 0x7ffff6d7cac8 input>, scope_die=0x7ffff6da1640, 
    item_decls=<constructor 0x7ffff6d735d0>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20672
#6  0x00000000008c27d0 in gen_decl_die (
    decl=<namelist_decl 0x7ffff6c507e8 input>, origin=<tree 0x0>, 
    context_die=0x7ffff6da1640)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20475
#7  0x00000000008c0d75 in process_scope_var (stmt=<block 0x7ffff6d713c0>, 
    decl=<namelist_decl 0x7ffff6c507e8 input>, origin=<tree 0x0>, 
    context_die=0x7ffff6da1640)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:19984
#8  0x00000000008c0df1 in decls_for_scope (stmt=<block 0x7ffff6d713c0>, 
    context_die=0x7ffff6da1640, depth=0)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20009
#9  0x00000000008bbd94 in gen_subprogram_die (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>, context_die=0x0)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:18383
#10 0x00000000008c220a in gen_decl_die (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>, origin=<tree 0x0>, 
    context_die=0x0) at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20376
#11 0x00000000008c345f in dwarf2out_decl (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20808
#12 0x00000000008c3481 in dwarf2out_function_decl (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20816
#13 0x000000000092e677 in rest_of_handle_final ()

and once via

Breakpoint 5, add_calling_convention_attribute (subr_die=0x7ffff6da1640, 
    decl=<function_decl 0x7ffff6d7e800 readnamelist>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:16859
16859     enum dwarf_calling_convention value = DW_CC_normal;
#0  add_calling_convention_attribute (subr_die=0x7ffff6da1640, 
    decl=<function_decl 0x7ffff6d7e800 readnamelist>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:16859
#1  0x00000000008bc477 in gen_subprogram_die (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>, context_die=0x0)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:18535
#2  0x00000000008c220a in gen_decl_die (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>, origin=<tree 0x0>, 
    context_die=0x0) at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20376
#3  0x00000000008c345f in dwarf2out_decl (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20808
#4  0x00000000008c3481 in dwarf2out_function_decl (
    decl=<function_decl 0x7ffff6d7e800 readnamelist>)
    at /space/rguenther/src/svn/trunk/gcc/dwarf2out.c:20816
#5  0x000000000092e677 in rest_of_handle_final ()

[Bug fortran/60060] [4.9 Regression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

--- Comment #8 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> Looks like a different bug - can you please file it separately as this bug
> assigned to me is about LTO issues?

PR60152.

[Bug fortran/60060] [4.9 Regression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Author: rguenth
Date: Wed Feb 12 09:01:30 2014
New Revision: 207715

URL: http://gcc.gnu.org/viewcvs?rev=207715&root=gcc&view=rev
Log:
2014-02-12  Richard Biener  <rguenther@suse.de>

    PR lto/60060
    * lto-lang.c (lto_write_globals): Do not call
    wrapup_global_declarations or emit_debug_global_declarations
    but emit debug info for non-function scope variables
    directly.

Modified:
    trunk/gcc/lto/ChangeLog
    trunk/gcc/lto/lto-lang.c

[Bug fortran/60060] [4.9 Regression] lto1: internal compiler error: in add_AT_specification, at dwarf2out.c:4096

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60060

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

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

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.