[Bug target/51726] New: LTO and attribute 'selectany'

[Bug target/51726] New: LTO and attribute 'selectany'

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

             Bug #: 51726
           Summary: LTO and attribute 'selectany'
    Classification: Unclassified
           Product: gcc
           Version: 4.6.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: d.g.gorbachev@gmail.com
            Target: *-*-mingw32


Created attachment 26213
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26213
Testcase

foobar.c: In function 'main':
foobar.c:1:32: error: 'bar':'selectany' attribute applies only to initialized
objects

[Bug target/51726] LTO and attribute 'selectany'

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-01-02
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-01-02 11:19:41 UTC ---
It's a target bug - the target should ignore the attribute on extern
declarations.

    case VAR_DECL:
      if (lookup_attribute ("selectany", DECL_ATTRIBUTES (decl)))
        {
          if (DECL_INITIAL (decl)
              /* If an object is initialized with a ctor, the static
                 initialization and destruction code for it is present in
                 each unit defining the object.  The code that calls the
                 ctor is protected by a link-once guard variable, so that
                 the object still has link-once semantics,  */
              || TYPE_NEEDS_CONSTRUCTING (TREE_TYPE (decl)))
            make_decl_one_only (decl, DECL_ASSEMBLER_NAME (decl));
          else
            error ("%q+D:'selectany' attribute applies only to "
                   "initialized objects", decl);
        }

thus, restrict processing to !DECL_EXTERNAL

[Bug target/51726] LTO and attribute 'selectany'

[d.g.gorbachev at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

--- Comment #2 from Dmitry Gorbachev <d.g.gorbachev at gmail dot com> 2012-01-23 20:22:00 UTC ---
Created attachment 26437
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26437
Patch (not tested)

[Bug lto/51726] LTO and attribute 'selectany'

[d.g.gorbachev at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

Dmitry Gorbachev <d.g.gorbachev at gmail dot com> changed:

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

--- Comment #3 from Dmitry Gorbachev <d.g.gorbachev at gmail dot com> ---
The attached patch does not work in GCC 4.10.0 - stream_read_tree() returns
NULL in lto_read_tree_1() (gcc/lto-streamer-in.c:1184):

>        DECL_INITIAL (expr) = stream_read_tree (ib, data_in);

`bar' is treated as a common var.

[Bug target/51726] LTO and attribute 'selectany'

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

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

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

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
On trunk the backend has to use varpool_node::get (decl)->get_constructor ().

But seriously the backend should process the attribute once at compile-time
and then discard it rather than trying to re-do the job at WPA/LTRANS time.

That is, _not_ do this in i386_pe_encode_section_info which is way too late
but instead directly in ix86_handle_selectany_attribute so LTO can see
the decl is one-only (and also the linker is informed correctly).

Still completely a backend bug.  IMHO as soon as it's not necessary to
retain the attribute on the decl it's a sign of a good implementation.

[Bug target/51726] LTO and attribute 'selectany'

[jacek at codeweavers dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

Jacek Caban <jacek at codeweavers dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jacek at codeweavers dot com

--- Comment #5 from Jacek Caban <jacek at codeweavers dot com> ---
It seems like selectany is not supported at all in LTO:

$ cat selectany.c 
__attribute__((selectany)) int i = 0;

$ i686-w64-mingw32-gcc selectany.c selectany.c -flto -shared
/tmp/cchTGZVt.o (symbol from plugin):(.text+0x0): multiple definition of `i'
/tmp/ccAy2OzL.o (symbol from plugin):(.text+0x0): first defined here
collect2: error: ld returned 1 exit status

$ i686-w64-mingw32-gcc --version
i686-w64-mingw32-gcc (GCC) 5.1.0
Copyright (C) 2015 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Should I file separated bug?

[Bug target/51726] LTO and attribute 'selectany'

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

--- Comment #6 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Created attachment 36427
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=36427&action=edit
Suggested patch

I will do some further testing with that patch, and later on send patch
upstream with a testcase for the common-case with selectany-attribute.

It would be appreciated  if you could test patch too.

[Bug target/51726] LTO and attribute 'selectany'

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|4.10.0                      |5.0

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Kai Tietz from comment #6)
> Created attachment 36427 [details]
> Suggested patch
> 
> I will do some further testing with that patch, and later on send patch
> upstream with a testcase for the common-case with selectany-attribute.
> 
> It would be appreciated  if you could test patch too.

I believe you can skip adding the attribute after you handled it this way
(unless other code literally looks for the attribute rather than its effect)

[Bug target/51726] LTO and attribute 'selectany'

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

Kai Tietz <ktietz at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |thiago at kde dot org

--- Comment #8 from Kai Tietz <ktietz at gcc dot gnu.org> ---
*** Bug 67688 has been marked as a duplicate of this bug. ***

[Bug target/51726] LTO and attribute 'selectany'

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

--- Comment #9 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Author: ktietz
Date: Fri Oct  2 08:06:52 2015
New Revision: 228370

URL: https://gcc.gnu.org/viewcvs?rev=228370&root=gcc&view=rev
Log:
        PR target/51726
        * config/i386/winnt.c (ix86_handle_selectany_attribute): Handle
        selectany within this function without need to keep attribute.
        (i386_pe_encode_section_info): Remove selectany-code.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/i386/winnt.c

[Bug target/51726] LTO and attribute 'selectany'

[ktietz at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51726

--- Comment #10 from Kai Tietz <ktietz at gcc dot gnu.org> ---
Author: ktietz
Date: Fri Oct  2 08:08:38 2015
New Revision: 228371

URL: https://gcc.gnu.org/viewcvs?rev=228371&root=gcc&view=rev
Log:
        PR target/51726
        * g++.dg/ext/selectany2.C: Allow uninitialized variable case.

Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/g++.dg/ext/selectany2.C