[Bug c++/50255] New: Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[Bug c++/50255] New: Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[stephan.bergmann.secondary at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

             Bug #: 50255
           Summary: Linker stumbles over non-grouped text/rodata for a
                    non-virtual thunk
    Classification: Unclassified
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: stephan.bergmann.secondary@googlemail.com


The below error has first been discussed at
<http://gcc.gnu.org/ml/gcc/2011-08/msg00455.html>.  It can at least be observed
when building recent versions of LibreOffice (e.g., revision
bf3ff35d8c96315c35cf8dc8495be4b488b55cb6 of
<git://anongit.freedesktop.org/libreoffice/core>) on Fedora 15 i386 (i.e., with
GCC 4.6.0 20110603 (Red Hat 4.6.0-10)).  The problem is that linking together
objects via

g++ -shared vbarange.o vbasheetobjects.o

fails with errors like

`.L109' referenced in section
`.rodata._ZN19ScVbaCollectionBaseIN4cppu15WeakImplHelper1IN3ooo3vba11XCollectionEEEE4ItemERKN3com3sun4star3uno3AnyESD_[ScVbaCollectionBase<cppu::WeakImplHelper1<ooo::vba::XCollection>
>::Item(com::sun::star::uno::Any const&, com::sun::star::uno::Any const&)]' of
vbasheetobjects.o: defined in discarded section
`.text._ZN19ScVbaCollectionBaseIN4cppu15WeakImplHelper1IN3ooo3vba11XCollectionEEEE4ItemERKN3com3sun4star3uno3AnyESD_[non-virtual
thunk to ScVbaCollectionBase<cppu::WeakImplHelper1<ooo::vba::XCollection>
>::Item(com::sun::star::uno::Any const&, com::sun::star::uno::Any const&)]' of
vbasheetobjects.o

The two C++ files vbarange.cxx and vbasheetobjects.cxx are compiled with
identical g++ command lines, including switches -fPIC -fno-common -pipe
-fvisibility=hidden -fvisibility-inlines-hidden -std=c++0x -fexceptions
-fno-enforce-eh-specs -Os besides more mundane -D, -I, -M, and -W switches. 
Both objects include code for the
_ZThn20_N19ScVbaCollectionBaseIN4cppu15WeakImplHelper1IN3ooo3vba11XCollectionEEEE4ItemERKN3com3sun4star3uno3AnyESD_
function, but there are two things that are odd:

First, the code emitted for the two functions is rather different in the two
object files, even though both are compiled with identical command line
switches.  One difference is that In vbasheetobjects.o, the code contains a
rodata section (see below) with a number of labels into code that follows,
whereas the code in vbarange.o does not contain such a rodata section.  Could
it be that -Os can cause these differences?  (If necessary, I can easily make
available the -S output for the relevant code from the two different files.)

Second, the directive for the rodata section mentioned above is

.section
.rodata._ZN19ScVbaCollectionBaseIN4cppu15WeakImplHelper1IN3ooo3vba11XCollectionEEEE4ItemERKN3com3sun4star3uno3AnyESD_,"aG",@progbits,_ZN19ScVbaCollectionBaseIN4cppu15WeakImplHelper1IN3ooo3vba11XCollectionEEEE4ItemERKN3com3sun4star3uno3AnyESD_,comdat

while the directive for the corresponding text section (split in multiple
parts) is

.section
.text._ZN19ScVbaCollectionBaseIN4cppu15WeakImplHelper1IN3ooo3vba11XCollectionEEEE4ItemERKN3com3sun4star3uno3AnyESD_,"axG",@progbits,_ZThn20_N19ScVbaCollectionBaseIN4cppu15WeakImplHelper1IN3ooo3vba11XCollectionEEEE4ItemERKN3com3sun4star3uno3AnyESD_,comdat

Note the difference in the GroupName part of the .section directives, the
rodata section using the symbol denoting the plain function instead of the
non-virtual thunk.  The theory is that the rodata and text sections should
belong to a single group, so that the linker would either keep or remove them
together.  What apparently happens instead is that the linker uses the text
section from vbarange.o, drops the corresponding text section from
vbasheetobjects.o, and then stumbles over the left-over rodata section. 
(Manually adapting the GroupName of the rodata section directive to match the
text section makes the link succeed.)

The problem has been worked around for now in LibreOffice by compiling
vbasheetobjects.o without -Os (see
<http://cgit.freedesktop.org/libreoffice/core/commit/?id=148e0ccb50ce419e18e452eb7ccfe03cb4881634>,
i.e., that change must be reverted before the problem can be observed in the
code revision mentioned at the beginning), which makes the problem go aways
rather by chance.

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-08-31 21:03:36 UTC ---
>GCC 4.6.0 20110603 (Red Hat 4.6.0-10)).

Does it happen with non RedHat version of the compiler meaning non modified
version of GCC?  Really you should have reported this first to redhat as they
have some modifications to their compiler that might have caused this issue.

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-08-31 21:06:43 UTC ---
This sounds like http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49538 but without
a testcase it is hard to say if it was really on the 4.6 branch or only with
RedHat's compiler.

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[caolanm at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

Caolan McNamara <caolanm at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |caolanm at redhat dot com,
                   |                            |jakub at redhat dot com

--- Comment #3 from Caolan McNamara <caolanm at redhat dot com> 2011-09-01 08:29:54 UTC ---
caolanm->jakub: this is the one I mentioned as FOSDEM 11.

caolanm->sberg: able to attach something standalone that can reproduce this,
i.e. .ii

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[stephan.bergmann.secondary at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #4 from Stephan Bergmann <stephan.bergmann.secondary at googlemail dot com> 2011-09-01 08:39:59 UTC ---
caolanm: will work on it, yes  (wanted to clarify first whether looking at
those group name mismatches is actually looking in the right direction)

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[stephan.bergmann.secondary at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #5 from Stephan Bergmann <stephan.bergmann.secondary at googlemail dot com> 2011-09-01 12:10:54 UTC ---
By the way, the problem of the rodata section having a different GroupName than
the text section is also present when you compile vbasheetobjects.cxx with
plain (non-RedHat-modified) GCC 4.6.0 and GCC 4.6.1, respectively.

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[stephan.bergmann.secondary at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #6 from Stephan Bergmann <stephan.bergmann.secondary at googlemail dot com> 2011-09-02 13:15:42 UTC ---
While I still don't have a stripped down test case, I at least know now what is
going wrong (on recent gcc-4_6-branch rev 178396, at least):

When compiling vbasheetobjects.cxx, the text section for the non-virtual thunk
(with section name ".text._ZN19..." and group name "_ZThn20_N19...") is first
obtained at

get_section()
hot_function_section()
function_section_1()
function_section_1()
assemble_start_function()
assemble_thunk()
cgraph_expand_function()
cgraph_expand_all_functions()
cgraph_optimize()
cgraph_finalize_compilation_unit()
cp_write_global_declarations()
compile_file()
do_compile()
toplev_main()
main()

There, get_section() is called with decl pointing to the thunk, as
assemble_thunk() sets global current_function_decl = thunk_fndecl;

Later on, the corresponding rodata section for the non-virtual thunk (with
section name ".rodata._ZN19..." and erroneous group name "_ZN19..." instead of
"_ZThn20_N19...") is first obtained at

get_section()
default_function_rodata_section()
final_scan_insn()
final()
rest_of_handle_final()
execute_one_pass()
execute_pass_list()
execute_pass_list()
execute_pass_list()
tree_rest_of_compilation()
cgraph_expand_function()
...

a few lines further down in the same invocation of cgraph_expand_function()
(and in final_scan_insn() we are at
switch_to_section(targetm.asm_out.function_rodata_section(current_function_decl));).
 But this time, current_function_decl points to the function itself, not the
non-virtual thunk, as tree_rest_of_compilation() sets current_function_decl =
fndecl;

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[stephan.bergmann.secondary at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #7 from Stephan Bergmann <stephan.bergmann.secondary at googlemail dot com> 2011-09-07 12:12:04 UTC ---
Created attachment 25220
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25220
reduced test case

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[stephan.bergmann.secondary at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #8 from Stephan Bergmann <stephan.bergmann.secondary at googlemail dot com> 2011-09-07 12:14:56 UTC ---
Created attachment 25221
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25221
script compiling test case and showing broken result

I finally have a stripped down test case now that shows that compiling the
given code leads to mismatched group names for the .text and .rodata sections
of the non-virtual thunk (the first three .sections in the output).

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-09-07
                 CC|                            |jakub at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-09-07 14:33:52 UTC ---
This changed on the trunk with Honza's Cgraph thunk reorg (which isn't really
backportable):
http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=173517
in r173516 we have
        .section       
.text._ZN1TI1WI1XEE1hEP1A,"axG",@progbits,_ZThn8_N1TI1WI1XEE1hEP1A,comdat
        .section       
.rodata._ZN1TI1WI1XEE1hEP1A,"aG",@progbits,_ZN1TI1WI1XEE1hEP1A,comdat
        .section       
.text._ZN1TI1WI1XEE1hEP1A,"axG",@progbits,_ZThn8_N1TI1WI1XEE1hEP1A,comdat
while in r173517
        .section       
.text._ZN1TI1WI1XEE1hEP1A,"axG",@progbits,_ZN1TI1WI1XEE1hEP1A,comdat
        .section       
.rodata._ZN1TI1WI1XEE1hEP1A,"aG",@progbits,_ZN1TI1WI1XEE1hEP1A,comdat
        .section       
.text._ZN1TI1WI1XEE1hEP1A,"axG",@progbits,_ZN1TI1WI1XEE1hEP1A,comdat

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |jakub at gcc dot gnu.org
                   |gnu.org                     |

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-09-07 15:27:16 UTC ---
Created attachment 25223
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25223
gcc46-pr50255.patch

Untested fix.  If we decide to emit the thunk in the same section as the
function, it'd better use the right comdat group as well.

[Bug c++/50255] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-09-08 06:03:11 UTC ---
Author: jakub
Date: Thu Sep  8 06:03:01 2011
New Revision: 178675

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=178675
Log:
    PR c++/50255
    * method.c (use_thunk): If emitting thunk into the same section
    as function, use the same DECL_COMDAT_GROUP as well.

Modified:
    branches/gcc-4_6-branch/gcc/cp/ChangeLog
    branches/gcc-4_6-branch/gcc/cp/method.c

[Bug c++/50255] [4.5 Regression] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.5.4
            Summary|Linker stumbles over        |[4.5 Regression] Linker
                   |non-grouped text/rodata for |stumbles over non-grouped
                   |a non-virtual thunk         |text/rodata for a
                   |                            |non-virtual thunk

--- Comment #12 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-09-08 06:06:40 UTC ---
Fixed for 4.6, 4.7 doesn't have this bug, 4.4 neither, so 4.5 is the only one
left unfixed right now.

[Bug c++/50255] [4.5 Regression] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
           Priority|P3                          |P2

[Bug c++/50255] [4.5 Regression] Linker stumbles over non-grouped text/rodata for a non-virtual thunk

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50255

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|4.5.4                       |4.6.2

--- Comment #13 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-07-02 10:14:35 UTC ---
Fixed in 4.6.2.