[Bug ada/52362] New: gnat.dg/lto8.adb FAILs with gas/gld

[Bug ada/52362] New: gnat.dg/lto8.adb FAILs with gas/gld

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

             Bug #: 52362
           Summary: gnat.dg/lto8.adb FAILs with gas/gld
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ada
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ro@gcc.gnu.org
              Host: *-*-solaris2*, x86_64-unknown-linux-gnu
            Target: *-*-solaris2*, x86_64-unknown-linux-gnu
             Build: *-*-solaris2*, x86_64-unknown-linux-gnu


The gnat.dg/lto8.adb test FAILs for me when using gas and gld 2.22 on Solaris
(both SPARC and x86): e.g. on i386-pc-solaris2.10:

FAIL: gnat.dg/lto8.adb (test for excess errors)
Excess errors:
/var/gcc/regression/trunk/10-gcc-gas-gld/build/i386-pc-solaris2.10/amd64/libada/adainclude/ada.o:(.data+0x0):
multiple definition of `ada_E'
/var/gcc/regression/trunk/10-gcc-gas-gld/build/i386-pc-solaris2.10/amd64/libada/adainclude/ada.o:(.data+0x0):
first defined here
/var/gcc/regression/trunk/10-gcc-gas-gld/build/i386-pc-solaris2.10/amd64/libada/adainclude/interfac.o:(.data+0x0):
multiple definition of `interfaces_E'
/var/gcc/regression/trunk/10-gcc-gas-gld/build/i386-pc-solaris2.10/amd64/libada/adainclude/interfac.o:(.data+0x0):
first defined here

and many more, but on Solaris only the 64-bit test fails.  The failure doesn't
occur with as/ld, gas/ld, or as/gld.

I see the same failure on Linux/amd64, but here both the 64-bit and 32-bit test
are affected.

  Rainer

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

vries at gcc dot gnu.org changed:

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

--- Comment #1 from vries at gcc dot gnu.org 2012-03-13 16:26:41 UTC ---
Ran into this with x86_64 --disable-bootstrap with r185326:
...
spawn /home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/gcc/gnatmake
--GCC=/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/gcc/xgcc
--GNATBIND=/home/vries/gcc_versions/ref-12-03-13/nobootstrap\
/build/gcc/gnatbind
--GNATLINK=/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/gcc/gnatlink
-cargs -B/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/gcc -largs
--GCC=/home/vries/gcc_versions\
/ref-12-03-13/nobootstrap/build/gcc/xgcc
-B/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/gcc  -margs
--RTS=/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/x86_64-unknown-linux-gnu/./libada\
 -q -f /home/vries/gcc_versions/ref-12-03-13/src/gcc/testsuite/gnat.dg/lto8.adb
-gnatws -flto -lm -o ./lto8.exe^M
/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/x86_64-unknown-linux-gnu/./libada/adainclude/ada.o:(.data+0x0):
multiple definition of `ada_E'^M
/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/x86_64-unknown-linux-gnu/./libada/adainclude/ada.o:(.data+0x0):
first defined here^M
/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/x86_64-unknown-linux-gnu/./libada/adainclude/interfac.o:(.data+0x0):
multiple definition of `interfaces_E'^M
/home/vries/gcc_versions/ref-12-03-13/nobootstrap/build/x86_64-unknown-linux-gnu/./libada/adainclude/interfac.o:(.data+0x0):
first defined here^M
...

configure line:
...
Configured with: /home/vries/gcc_versions/ref-12-03-13/src/configure
--prefix=/home/vries/gcc_versions/ref-12-03-13/nobootstrap/install
--with-cloog=/home/vries/gcc_versions/infra
--with-ppl=/home/vries/gcc_versions/infra
--with-gmp=/home/vries/gcc_versions/infra
--with-mpfr=/home/vries/gcc_versions/infra
--with-mpc=/home/vries/gcc_versions/infra --disable-bootstrap
--enable-checking=yes,rtl --enable-languages=c,fortran,ada,java,objc,c++
...

ld version:
...
GNU ld (GNU Binutils for Ubuntu) 2.21.53.20110810
...

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

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

--- Comment #2 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-03-14 20:43:16 UTC ---
What's different with --disable-bootstrap?  Missing linker plugin support?

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

--- Comment #3 from vries at gcc dot gnu.org 2012-03-14 21:02:51 UTC ---
(In reply to comment #2)
> What's different with --disable-bootstrap?  Missing linker plugin support?

I did a reference build, bootstrap and nobootstrap, and the test passed for
bootstrap and failed for nobootstrap:
...
ref-last/FAIL:./gcc/testsuite/gnat/gnat.sum:PASS: gnat.dg/lto8.adb (test for
excess errors)
ref-last/FAIL:./gcc/testsuite/gnat/gnat.sum:PASS: gnat.dg/lto8.adb execution
test
ref-last/FAIL.nobootstrap:./gcc/testsuite/gnat/gnat.sum:FAIL: gnat.dg/lto8.adb
(test for excess errors)
ref-last/FAIL.nobootstrap:./gcc/testsuite/gnat/gnat.sum:UNRESOLVED:
gnat.dg/lto8.adb compilation failed to produce executable
...

After that, I tested an tail-merge (AFAIK unrelated) patch on top of the same
sources, and now the test fails for both:
...
test-pr51879.8.4.patch/with/FAIL:./gcc/testsuite/gnat/gnat.sum:FAIL:
gnat.dg/lto8.adb (test for excess errors)
test-pr51879.8.4.patch/with/FAIL:./gcc/testsuite/gnat/gnat.sum:UNRESOLVED:
gnat.dg/lto8.adb compilation failed to produce executable
test-pr51879.8.4.patch/with/FAIL.nobootstrap:./gcc/testsuite/gnat/gnat.sum:FAIL:
gnat.dg/lto8.adb (test for excess errors)
test-pr51879.8.4.patch/with/FAIL.nobootstrap:./gcc/testsuite/gnat/gnat.sum:UNRESOLVED:
gnat.dg/lto8.adb compilation failed to produce executable
...

So I'm not sure that the failure is linked to --nobootstrap.

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

--- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-03-14 21:10:38 UTC ---
> After that, I tested an tail-merge (AFAIK unrelated) patch on top of the same
> sources, and now the test fails for both:

With exactly the same invocation of make -k check, in particular from the same
directory?

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[vries at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

--- Comment #5 from vries at gcc dot gnu.org 2012-03-14 21:28:40 UTC ---
(In reply to comment #4)
> > After that, I tested an tail-merge (AFAIK unrelated) patch on top of the same
> > sources, and now the test fails for both:
> 
> With exactly the same invocation of make -k check, in particular from the same
> directory?

Same invocation of make -k check:
...
make \
  -k \
  -j9 \
  check \
  > CHECKLOG 2>&1
...

Not the same dirs (4 different dirs {ref,patch}x{bootstrap,nobootstrap}).

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|*-*-solaris2*,              |
                   |x86_64-unknown-linux-gnu    |
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2012-05-13
               Host|*-*-solaris2*,              |
                   |x86_64-unknown-linux-gnu    |
         AssignedTo|unassigned at gcc dot       |ebotcazou at gcc dot
                   |gnu.org                     |gnu.org
     Ever Confirmed|0                           |1
              Build|*-*-solaris2*,              |
                   |x86_64-unknown-linux-gnu    |

--- Comment #6 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-05-13 07:09:33 UTC ---
Reproducible on PowerPC/Linux as well.

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

--- Comment #7 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-05-14 11:14:28 UTC ---
When there are many files to link, gnatlink passes a file containing the file
names, instead of the file names directly, on the GCC link line.  There is a
bad interaction with LTO:

eric@atlantis:~/build/gcc-4_7-branch/native32> cat t.c
int main (void) { return 0; }
eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -c t.c -flto
eric@atlantis:~/build/gcc-4_7-branch/native32> cat FILE
INPUT (./t.o)
eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -o t FILE
eric@atlantis:~/build/gcc-4_7-branch/native32> ./t
eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -o t FILE -flto
/tmp/ccTBohVP.ltrans0.ltrans.o: In function `main':
ccTBohVP.ltrans0.o:(.text+0x0): multiple definition of `main'
./t.o:t.c:(.text+0x0): first defined here
./t.o: In function `main':
t.c:(.text+0x0): multiple definition of `main'
./t.o:t.c:(.text+0x0): first defined here
collect2: error: ld returned 1 exit status

We should probably use a regular response file instead:
eric@atlantis:~/build/gcc-4_7-branch/native32> cat FILE2
./t.o
eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -o t -Wl,@FILE2
-flto
eric@atlantis:~/build/gcc-4_7-branch/native32> ./t

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

--- Comment #8 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-05-14 11:22:45 UTC ---
(In reply to comment #7)
> When there are many files to link, gnatlink passes a file containing the file
> names, instead of the file names directly, on the GCC link line.  There is a
> bad interaction with LTO:
> 
> eric@atlantis:~/build/gcc-4_7-branch/native32> cat t.c
> int main (void) { return 0; }
> eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -c t.c -flto
> eric@atlantis:~/build/gcc-4_7-branch/native32> cat FILE
> INPUT (./t.o)
> eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -o t FILE
> eric@atlantis:~/build/gcc-4_7-branch/native32> ./t
> eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -o t FILE -flto
> /tmp/ccTBohVP.ltrans0.ltrans.o: In function `main':
> ccTBohVP.ltrans0.o:(.text+0x0): multiple definition of `main'
> ./t.o:t.c:(.text+0x0): first defined here
> ./t.o: In function `main':
> t.c:(.text+0x0): multiple definition of `main'
> ./t.o:t.c:(.text+0x0): first defined here
> collect2: error: ld returned 1 exit status
> 
> We should probably use a regular response file instead:
> eric@atlantis:~/build/gcc-4_7-branch/native32> cat FILE2
> ./t.o
> eric@atlantis:~/build/gcc-4_7-branch/native32> gcc/xgcc -Bgcc -o t -Wl,@FILE2
> -flto
> eric@atlantis:~/build/gcc-4_7-branch/native32> ./t

Yes, we don't "understand" linker scripts.  The linker plugin would need to
handle feeding additional inputs towards the link stage - and we'd need to
drop the original FILE input from the final link (which is where the multiple
references come from).

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

--- Comment #9 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-05-25 09:24:12 UTC ---
Author: ebotcazou
Date: Fri May 25 09:24:08 2012
New Revision: 187874

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=187874
Log:
    PR ada/52362
    * config.gcc (i[34567]86-*-mingw* | x86_64-*-mingw*): Set gas and
    gnu_ld variables to yes.
    * configure.ac (HAVE_GNU_LD): Move to after config.gcc inclusion.
    (HAVE_GNU_AS): Likewise.
    * config.in: Regenerate.
    * configure: Likewise.
ada/
    * link.c (__gnat_object_file_option): Set to "-Wl,@" for GNU linker.
    (__gnat_using_gnu_linker): Delete.
    * gnatlink.adb (Gnatlink): Declare Object_File_Option_Ptr here...
    Declare Object_File_Option string constant and Using_GNU_response_file
    boolean constant.
    (Process_Binder_File): ...instead of here.  Delete Using_GNU_Linker,
    Opening and Closing local variables.  Do not handle the GNU linker
    specially.
    (Write_RF): New procedure to write into the response file.  Escape some
    characters if a GNU response file is used.  Keep track of error status.
    Invoke Write_RF to write into the response file.  Delete the file only
    if the link was successful.
    * mlib-utl.adb: Do not `with' package System.
    (Gcc): Likewise.  Declare Object_File_Option string constant and
    Using_GNU_response_file boolean constant.
    (Write_RF): Take a string instead of address and length.  Escape some
    characters if a GNU response file is used.
    Invoke Write_RF to write into the response file.  Delete the file only
    if the link was successful.  Do not warn if it cannot be deleted.    

Added:
    trunk/gcc/testsuite/gnat.dg/lto14.adb
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/ada/ChangeLog
    trunk/gcc/ada/gnatlink.adb
    trunk/gcc/ada/link.c
    trunk/gcc/ada/mlib-utl.adb
    trunk/gcc/config.gcc
    trunk/gcc/config.in
    trunk/gcc/configure
    trunk/gcc/configure.ac
    trunk/gcc/testsuite/ChangeLog

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

--- Comment #10 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-05-25 10:07:27 UTC ---
Author: ebotcazou
Date: Fri May 25 10:07:23 2012
New Revision: 187877

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=187877
Log:
    PR ada/52362
    * config.gcc (i[34567]86-*-mingw* | x86_64-*-mingw*): Set gas and
    gnu_ld variables to yes.
    * configure.ac (HAVE_GNU_LD): Move to after config.gcc inclusion.
    (HAVE_GNU_AS): Likewise.
    * config.in: Regenerate.
    * configure: Likewise.
ada/
    * link.c (__gnat_object_file_option): Set to "-Wl,@" for GNU linker.
    (__gnat_using_gnu_linker): Delete.
    * gnatlink.adb (Gnatlink): Declare Object_File_Option_Ptr here...
    Declare Object_File_Option string constant and Using_GNU_response_file
    boolean constant.
    (Process_Binder_File): ...instead of here.  Delete Using_GNU_Linker,
    Opening and Closing local variables.  Do not handle the GNU linker
    specially.
    (Write_RF): New procedure to write into the response file.  Escape some
    characters if a GNU response file is used.  Keep track of error status.
    Invoke Write_RF to write into the response file.  Delete the file only
    if the link was successful.
    * mlib-utl.adb: Do not `with' package System.
    (Gcc): Likewise.  Declare Object_File_Option string constant and
    Using_GNU_response_file boolean constant.
    (Write_RF): Take a string instead of address and length.  Escape some
    characters if a GNU response file is used.
    Invoke Write_RF to write into the response file.  Delete the file only
    if the link was successful.  Do not warn if it cannot be deleted.

Added:
    branches/gcc-4_7-branch/gcc/testsuite/gnat.dg/lto14.adb
      - copied unchanged from r187874, trunk/gcc/testsuite/gnat.dg/lto14.adb
Modified:
    branches/gcc-4_7-branch/gcc/ChangeLog
    branches/gcc-4_7-branch/gcc/ada/ChangeLog
    branches/gcc-4_7-branch/gcc/ada/gnatlink.adb
    branches/gcc-4_7-branch/gcc/ada/link.c
    branches/gcc-4_7-branch/gcc/ada/mlib-utl.adb
    branches/gcc-4_7-branch/gcc/config.gcc
    branches/gcc-4_7-branch/gcc/config.in
    branches/gcc-4_7-branch/gcc/configure
    branches/gcc-4_7-branch/gcc/configure.ac
    branches/gcc-4_7-branch/gcc/testsuite/ChangeLog

[Bug ada/52362] gnat.dg/lto8.adb FAILs with gas/gld

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52362

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.7.1

--- Comment #11 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-05-25 10:16:46 UTC ---
Thanks for reporting the problem.