[Bug middle-end/54860] New: [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[Bug middle-end/54860] New: [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

             Bug #: 54860
           Summary: [4.8 Regression]: build fails on cris-elf configuring
                    libgfortran due to recent "attribute" changes in core
                    gcc
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Keywords: build, ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: hp@gcc.gnu.org
                CC: dodji@gcc.gnu.org
              Host: x86_64-unknown-linux-gnu
            Target: cris-axis-elf


Created attachment 28393
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28393
Repeat with gccobj/./gcc/f951 conftest.f -ffixed-form

A patch in the revision range (last_known_working:first_known_failing)
192197:192199 caused the build for cris-elf to fail as follows:

...
checking whether the /tmp/build20121008/gccobj/./gcc/gfortran
-B/tmp/build20121008/gccobj/./gcc/ -nostdinc
-B/tmp/build20121008/gccobj/cris-elf/newlib/ -isystem
/tmp/build20121008/gccobj/cris-elf/newlib/targ-include -isystem
/tmp/build20121008/gcc/newlib/libc/include
-B/tmp/build20121008/gccobj/cris-elf/libgloss/cris
-L/tmp/build20121008/gccobj/cris-elf/libgloss/libnosys
-L/tmp/build20121008/gcc/libgloss/cris -B/tmp/build20121008/pre/cris-elf/bin/
-B/tmp/build20121008/pre/cris-elf/lib/ -isystem
/tmp/build20121008/pre/cris-elf/include -isystem
/tmp/build20121008/pre/cris-elf/sys-include    linker
(/tmp/build20121008/gccobj/./gcc/collect-ld) supports shared libraries... yes
checking dynamic linker characteristics... no
checking how to hardcode library paths into programs... immediate
checking whether the GNU Fortran compiler is working... no
configure: error: GNU Fortran is not working; please report a bug in
http://gcc.gnu.org/bugzilla, attaching
/tmp/build20121008/gccobj/cris-elf/libgfortran/config.log
make[1]: *** [configure-target-libgfortran] Error 1
make[1]: Leaving directory `/tmp/build20121008/gccobj'

In that config.log:
configure:12882: /tmp/build20121008/gccobj/./gcc/gfortran
-B/tmp/build20121008/gccobj/./gcc/ -nostdinc
-B/tmp/build20121008/gccobj/cris-elf/newlib/ -isystem
/tmp/build20121008/gccobj/cris-elf/newlib/targ-include -isystem
/tmp/build20121008/gcc/newlib/libc/include
-B/tmp/build20121008/gccobj/cris-elf/libgloss/cris
-L/tmp/build20121008/gccobj/cris-elf/libgloss/libnosys
-L/tmp/build20121008/gcc/libgloss/cris -B/tmp/build20121008/pre/cris-elf/bin/
-B/tmp/build20121008/pre/cris-elf/lib/ -isystem
/tmp/build20121008/pre/cris-elf/include -isystem
/tmp/build20121008/pre/cris-elf/sys-include    -c -g  conftest.f >&5
f951: internal compiler error: Segmentation fault
0x89f135 crash_signal
        /tmp/build20121008/gcc/gcc/toplev.c:335
0xbcd90a htab_mod
        /tmp/build20121008/gcc/libiberty/hashtab.c:272
0xbcd90a htab_find_with_hash
        /tmp/build20121008/gcc/libiberty/hashtab.c:602
0xbd0068 lookup_scoped_attribute_spec(tree_node const*, tree_node const*)
        /tmp/build20121008/gcc/gcc/attribs.c:325
0xa69645 comp_type_attributes(tree_node const*, tree_node const*)
        /tmp/build20121008/gcc/gcc/tree.c:4353
0xa69a10 build_type_attribute_qual_variant(tree_node*, tree_node*, int)
        /tmp/build20121008/gcc/gcc/tree.c:4302
0x564203 build_library_function_decl_1
        /tmp/build20121008/gcc/gcc/fortran/trans-decl.c:2650
0x56437c gfc_build_library_function_decl_with_spec(tree_node*, char const*,
tree_node*, int, ...)
        /tmp/build20121008/gcc/gcc/fortran/trans-decl.c:2692
0x5644e3 gfc_build_builtin_function_decls()
        /tmp/build20121008/gcc/gcc/fortran/trans-decl.c:3067
0x545b7d gfc_create_decls
        /tmp/build20121008/gcc/gcc/fortran/f95-lang.c:175
0x545b7d gfc_be_parse_file
        /tmp/build20121008/gcc/gcc/fortran/f95-lang.c:190
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.
configure:12882: $? = 1
configure: failed program was:
|       program main
| 
|       end


Configure-test-code and command-line attached.

Author of only suspect patch in revision range CC:ed.
Just configure for --target=cris-elf and "make all-gcc" to produce f951.

[Bug middle-end/54860] [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[dodji at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

Dodji Seketeli <dodji at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2012-10-08
         AssignedTo|unassigned at gcc dot       |dodji at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1

--- Comment #1 from Dodji Seketeli <dodji at gcc dot gnu.org> 2012-10-08 21:58:09 UTC ---
I am looking at it.

[Bug middle-end/54860] [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[dodji at seketeli dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #2 from dodji at seketeli dot org <dodji at seketeli dot org> 2012-10-08 22:23:25 UTC ---
"hp at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org> a écrit:

> Just configure for --target=cris-elf and "make all-gcc" to produce
> f951.

So I did this on my system (a Fedora 17).

config.log says:

This file contains any messages produced by compilers while
running configure, to aid debugging if configure makes a mistake.

It was created by configure, which was
generated by GNU Autoconf 2.64.  Invocation command line was

  $ /home/dodji/devel/git/gcc/cur/configure --target=cris-elf
--enable-langages=c,c++,fortran


I am getting the f951 binary.

> Created attachment 28393
>
>   --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28393
>
> Repeat with gccobj/./gcc/f951 conftest.f -ffixed-form

But then when I do:

/home/dodji/devel/git/gcc/cur/build/gcc
$ cat ../../prtests/test.f
      program main
      end
$ ./f951 -quiet ../../prtests/test.f -ffixed-form
$

I am not seeing the crash.

What else am I missing to reproduce the issue?

[Bug middle-end/54860] [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #3 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-10-08 22:46:52 UTC ---
(In reply to comment #2)
> $ ./f951 -quiet ../../prtests/test.f -ffixed-form
> $
> 
> I am not seeing the crash.

Does running valgrind on that show anything suspicious?

> What else am I missing to reproduce the issue?

I don't know.  My x86_64 host is running a fairly vanilla Fedora 12 (save for
being behind "a few" updates), host g++  is gcc-4.4.3-4.fc12.x86_64.  In-tree
mpc=0.8 mpfr=2.4.1 gmp=4.3.0.  I'll try a few other hosts here.

[Bug middle-end/54860] [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #4 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-10-09 01:48:48 UTC ---
(In reply to comment #3)
> I'll try a few other hosts here.

Build also fails for x86_64 Fedora 8 (gcc-4.1.2-33) and Debian squeeze 6.0.4
"gcc version 4.4.5 (Debian 4.4.5-8)".

[Bug middle-end/54860] [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #5 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-10-09 03:25:53 UTC ---
(In reply to comment #4)
> (In reply to comment #3)
> > I'll try a few other hosts here.
> 
> Build also fails for x86_64 Fedora 8 (gcc-4.1.2-33) and Debian squeeze 6.0.4
> "gcc version 4.4.5 (Debian 4.4.5-8)".

Repeated on gcc47.  Feel free to use the stuff in ~hp there.  Should be obvious

(The file named "patch" is just to move the r192219 tarball to r192199.)

[Bug middle-end/54860] [4.8 Regression]: build fails on cris-elf configuring libgfortran due to recent "attribute" changes in core gcc

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #6 from Uros Bizjak <ubizjak at gmail dot com> 2012-10-09 06:37:12 UTC ---
I can confirm the failure on alpha:

(gdb) bt
#0  0x0000000120b4c288 in htab_find_with_hash ()
#1  0x0000000120b4fac0 in lookup_scoped_attribute_spec(tree_node const*,
tree_node const*) ()
    at ../../gcc-svn/trunk/gcc/attribs.c:325
#2  0x0000000120915950 in comp_type_attributes(tree_node const*, tree_node
const*) ()
    at ../../gcc-svn/trunk/gcc/tree.c:4353
#3  0x0000000120918118 in build_type_attribute_qual_variant(tree_node*,
tree_node*, int) ()
    at ../../gcc-svn/trunk/gcc/tree.c:4303
#4  0x00000001201ecf44 in build_library_function_decl_1(tree_node*, char
const*, tree_node*, int, __va_list_tag) ()
    at ../../gcc-svn/trunk/gcc/fortran/trans-decl.c:2650
#5  0x00000001201f3770 in gfc_build_library_function_decl_with_spec(tree_node*,
char const*, tree_node*, int, ...) ()
    at ../../gcc-svn/trunk/gcc/fortran/trans-decl.c:2692
#6  0x00000001201f38d0 in gfc_build_builtin_function_decls() () at
../../gcc-svn/trunk/gcc/fortran/trans-decl.c:3067
#7  0x00000001201c49f8 in gfc_be_parse_file() () at
../../gcc-svn/trunk/gcc/fortran/f95-lang.c:175
#8  0x000000012067516c in compile_file() () at
../../gcc-svn/trunk/gcc/toplev.c:546
#9  0x000000012067772c in toplev_main(int, char**) () at
../../gcc-svn/trunk/gcc/toplev.c:1866
#10 0x00000001200db798 in main () at ../../gcc-svn/trunk/gcc/main.c:36

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

Uros Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.8 Regression]: build     |[4.8 Regression]: build
                   |fails on cris-elf           |fails when configuring
                   |configuring libgfortran due |libgfortran due to recent
                   |to recent "attribute"       |"attribute" changes in core
                   |changes in core gcc         |gcc

--- Comment #7 from Uros Bizjak <ubizjak at gmail dot com> 2012-10-09 07:55:40 UTC ---
I can also trigger this ICE with a cross from x86_64-pc-linux-gnu to
alpha-linux-gnu:

#0  htab_find_with_hash (htab=0x0, element=0x7fffffffcb30, hash=484198) at
../../gcc-svn/trunk/libiberty/hashtab.c:600
#1  0x0000000000c2dbb6 in lookup_scoped_attribute_spec (ns=Unhandled dwarf
expression opcode 0xf3
) at ../../gcc-svn/trunk/gcc/attribs.c:325
#2  0x0000000000a91bc9 in comp_type_attributes (type1=0x2aaaae8199d8,
type2=0x2aaaae819930) at ../../gcc-svn/trunk/gcc/tree.c:4353
#3  0x0000000000a91f0b in build_type_attribute_qual_variant
(ttype=0x2aaaae819930, attribute=Unhandled dwarf expression opcode 0xf3
) at ../../gcc-svn/trunk/gcc/tree.c:4303

The problem is dereference of htab (=0x0) in htab_find_with_hash.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[dodji at seketeli dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #8 from dodji at seketeli dot org <dodji at seketeli dot org> 2012-10-09 14:08:20 UTC ---
I can reproduce it now.  I guess I shouldn't look at bugs around
midnight or something.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[dodji at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #9 from Dodji Seketeli <dodji at gcc dot gnu.org> 2012-10-09 14:52:24 UTC ---
Created attachment 28400
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28400
Candidate fix patch

Guys, could you please test this patch on your trees to see if it fixes the
issue for you?

I am testing it on my side.

Thank you in advance, and sorry for the inconvenience.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[davem at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

davem at gcc dot gnu.org changed:

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

--- Comment #10 from davem at gcc dot gnu.org 2012-10-09 16:40:45 UTC ---
This happens on sparc too, I'm doing a test bootstrap with your patch right
now.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[davem at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #11 from davem at gcc dot gnu.org 2012-10-09 17:12:25 UTC ---
I can confirm that the patch fixes the sparc bootstrap.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #12 from Uros Bizjak <ubizjak at gmail dot com> 2012-10-09 18:40:06 UTC ---
(In reply to comment #11)
> I can confirm that the patch fixes the sparc bootstrap.

Also alpha bootstrap.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #13 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-10-09 20:43:11 UTC ---
The patch fixes the problem for cris-elf with no regressions in test-suite
results.  Thanks!

But, the new tests introduce some new failures:

+g++.sum g++.dg/cpp0x/gen-attrs-17.2.C
+g++.sum g++.dg/cpp0x/gen-attrs-21.C
+g++.sum g++.dg/cpp0x/gen-attrs-39.C
+g++.sum g++.dg/cpp0x/gen-attrs-51.C
+g++.sum g++.dg/cpp0x/gen-attrs-52.C

(the rest of the new tests pass)

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[dodji at seketeli dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #14 from dodji at seketeli dot org <dodji at seketeli dot org> 2012-10-09 21:28:44 UTC ---
I have submitted the patch to
http://gcc.gnu.org/ml/gcc-patches/2012-10/msg00914.html.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[dodji at seketeli dot org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #15 from dodji at seketeli dot org <dodji at seketeli dot org> 2012-10-09 21:33:00 UTC ---
> --- Comment #13 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-10-09 20:43:11 UTC ---
>
> The patch fixes the problem for cris-elf with no regressions in test-suite
> results.  Thanks!

You are welcome.  Sorry for the breakage in the first place.

> But, the new tests introduce some new failures:

> +g++.sum g++.dg/cpp0x/gen-attrs-17.2.C
>
> +g++.sum g++.dg/cpp0x/gen-attrs-21.C
>
> +g++.sum g++.dg/cpp0x/gen-attrs-39.C
>
> +g++.sum g++.dg/cpp0x/gen-attrs-51.C
>
> +g++.sum g++.dg/cpp0x/gen-attrs-52.C

Hmmh.  I don't reach the point of running testsuites on my limited
cross-compilation setup here.  Would it be possible that I see the log
of these testsuite errors somehow?

Thank you.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #16 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-10-09 21:57:14 UTC ---
Created attachment 28407
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28407
Excerpt from g++.log with the mentioned new errors

I would have attached the whole of g++.log but at 44 MeB it seemed a little too
big.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #17 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-10-09 22:02:22 UTC ---
To be absolutely clear: I don't really care about the new errors, it was just
appropriate to mention them.  Don't let any investigation hold up your
break-fixing commit... ;)

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[dodji at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #18 from Dodji Seketeli <dodji at gcc dot gnu.org> 2012-10-10 10:25:13 UTC ---
Author: dodji
Date: Wed Oct 10 10:25:03 2012
New Revision: 192301

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=192301
Log:
PR middle-end/54860 - Make sure attributes hash table is created

On targets cris-elf, alpha and sparc (for instance) it can happen that
the attribute_tables variable is empty for fortran.  Thus
register_scoped_attributes (called by init_attributes) won't call
register_scoped_attributes, so the hash table member of
scoped_attributes is not created.

Later when we try to e.g, lookup an attribute by calling
lookup_scoped_attribute_spec, that NULL member hash table comes to
byte us as htab_find_with_hash crashes.

This patch fixes this by ensuring in register_scoped_attributes that
the hash table is created.

Tested on cris-elf, x86_64-unknown-linux-gnu against trunk and some
commenters on the bug bootstrapped it on alpha and sparc.

gcc/

    * attribs.c (register_scoped_attributes): Ensure the attribute
    hash table is created.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/attribs.c

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
                 CC|                            |jakub at gcc dot gnu.org
         Resolution|                            |FIXED
   Target Milestone|---                         |4.8.0

--- Comment #19 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-11-20 12:26:29 UTC ---
Assuming fixed.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #20 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-11-20 15:26:43 UTC ---
(In reply to comment #19)
> Assuming fixed.

Well yes!  Maybe we should clear the issue on who's supposed to close PR's
after a fix?  I'd be happy to do it for PR's I've opened, I've just left it to
whomever is "assigned" who may want to keep the PR open pending a backport.

[Bug middle-end/54860] [4.8 Regression]: build fails when configuring libgfortran due to recent "attribute" changes in core gcc

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54860

--- Comment #21 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-11-20 18:16:08 UTC ---
(In reply to comment #20)
> (In reply to comment #19)
> > Assuming fixed.
> 
> Well yes!  Maybe we should clear the issue on who's supposed to close PR's
> after a fix?  I'd be happy to do it for PR's I've opened, I've just left it to
> whomever is "assigned" who may want to keep the PR open pending a backport.

To wit, the original issue is resolved, but the new tests mentioned in comment
#13 still fail, so I'm going to clone this PR.