[Bug bootstrap/45612] New: [4.6 regression] Reference to undefined label building libada no Solaris 2/SPARC

[Bug bootstrap/45612] New: [4.6 regression] Reference to undefined label building libada no Solaris 2/SPARC

[ro at gcc dot gnu dot org]

Between 20100903 and 20100908, mainline bootstrap with Ada started failing when
compiling the 32-bit g-debpoo.adb:

% /var/gcc/gcc-4.6.0-20100908/11-gcc-gas/./gcc/xgcc
-B/var/gcc/gcc-4.6.0-20100908/11-gcc-gas/./gcc/
-B/usr/local/sparc-sun-solaris2.11/bin/ -B/usr/local/sparc-sun-solaris2.11/lib/
-isystem /usr/local/sparc-sun-solaris2.11/include -isystem
/usr/local/sparc-sun-solaris2.11/sys-include    -c -g -O2  -fPIC  -W -Wall
-gnatpg   g-debpoo.adb -o g-debpoo.o -save-temps
g-debpoo.s: Assembler messages:
g-debpoo.s:5362: Error: undefined symbol `.LL363' in operation
g-debpoo.s:5363: Error: undefined symbol `.LL363' in operation


-- 
           Summary: [4.6 regression] Reference to undefined label building
                    libada no Solaris 2/SPARC
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ro at gcc dot gnu dot org
 GCC build triplet: sparc-sun-solaris2*
  GCC host triplet: sparc-sun-solaris2*
GCC target triplet: sparc-sun-solaris2*


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada no Solaris 2/SPARC

[ro at gcc dot gnu dot org]

------- Comment #1 from ro at gcc dot gnu dot org  2010-09-09 10:29 -------
Created an attachment (id=21749)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=21749&action=view)
assembler output


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada no Solaris 2/SPARC

[ebotcazou at gcc dot gnu dot org]

------- Comment #2 from ebotcazou at gcc dot gnu dot org  2010-09-09 11:09 -------
This compiled fine on 20100907 for me.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada no Solaris 2/SPARC

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.6.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada no Solaris 2/SPARC

[ro at CeBiTec dot Uni-Bielefeld dot DE]

------- Comment #3 from ro at CeBiTec dot Uni-Bielefeld dot DE  2010-09-10 15:15 -------
Subject: Re:  [4.6 regression] Reference to undefined label building libada no
Solaris 2/SPARC

A reghunt identified that the regression was caused by this patch:

2010-09-07  Jan Hubicka  <jh@suse.cz>

        * tree-inline.c (tree_inlinable_function_p): Do not test
DECL_REPLACEABL
E_P.
        * ipa-inline.c (cgraph_default_inline_p, update_caller_keys,
update_call
ee_keys,
        cgraph_decide_inlining): Test function availability.
        * cif-code.def (OVERWRITABLE): New code.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[ro at gcc dot gnu dot org]

------- Comment #4 from ro at gcc dot gnu dot org  2010-09-10 15:19 -------
Jan, could you please have a look.


-- 

ro at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hubicka at gcc dot gnu dot
                   |                            |org
            Summary|[4.6 regression] Reference  |[4.6 regression] Reference
                   |to undefined label building |to undefined label building
                   |libada no Solaris 2/SPARC   |libada on Solaris 2/SPARC


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[danglin at gcc dot gnu dot org]

------- Comment #5 from danglin at gcc dot gnu dot org  2010-09-12 13:06 -------
*** Bug 45639 has been marked as a duplicate of this bug. ***


-- 

danglin at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |danglin at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[ro at CeBiTec dot Uni-Bielefeld dot DE]

------- Comment #7 from ro at CeBiTec dot Uni-Bielefeld dot DE  2010-09-17 08:55 -------
Subject: Re:  [4.6 regression] Reference to undefined label building libada on
Solaris 2/SPARC

> ------- Comment #6 from hubicka at gcc dot gnu dot org  2010-09-16 20:57 -------
> This is really strange case. The patch should at most introduce extra inlining
> that naturally should not introduce undefined symbols.
> It is used as:
>         sethi   %hi(_GLOBAL_OFFSET_TABLE_-(.LL363-.)), %o3
>         or      %o3, %lo(_GLOBAL_OFFSET_TABLE_-(.LL363-.)), %o3
>         call    gnat__debug_pools__put_line, 0
>
> so it is passed to gnat__debug_pools__put_line.  Any idea what should be there?
> It does not seem like variable, rather like code label. Can I have

Unfortunately not; you'd have to ask one of the Ada maintainers.  Eric
is on the Cc:, he might know.

> -fdump-tree-optimized (or perhaps better -fdump-tree-all) dumps?

Sure.  I couldn't attach it to the PR (1.8 MB even compressed with
bzip2), so I've put it at

http://www.cebitec.uni-bielefeld.de/~ro/files/g-debpoo-dump-tree-all.tar.bz2

Thanks.
        Rainer


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2010.10.06 14:51:02
     Ever Confirmed|0                           |1

--- Comment #22 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2010-10-06 14:51:02 UTC ---
Investigating.

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[hainque at adacore dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #21 from hainque at adacore dot com <hainque at adacore dot com> 2010-10-01 22:11:52 UTC ---
hubicka at gcc dot gnu.org wrote:
> We might be missing some code in inliner that passes the fact that
> the label is user label...

 I don't yet have a clear track here, but can provide a pretty easy
 way to reproduce with a cross and get starters for further analysis:

 $ configure --target=sparc-sun-solaris2.10 --enable-languages=c,ada
             --disable-libada --disable-bootstrap

 $ make CFLAGS="-O0 -g"
 ...
 let it fail when it gets to C parts of runtime libraries.
 you should have gnat1 at this stage. then ...

 $ cd gcc
 $ make -C ada ../stamp-gnatlib1-rts

 this prepares the runtime library source symlinks in ada/rts. now:

 $ ./gnat1 -Iada/rts ada/rts/g-debpoo.adb -O2 -fdump-tree-all -da

 the anomaly is visible in the output, in particular in one call 
 to debug_pools__put_line from the code for 'Dereference'

  gnat__debug_pools__dereference__2:
     ...
     ld   [%g1+%lo(gnat__debug_pools__code_address_for_dereference_end)], %o4
     ...
     sethi   %hi(.LL544), %o3
     ...
     call    gnat__debug_pools__put_line, 0
     or     %o3, %lo(.LL544), %o3

 LL544 is nowhere for me.

 From a quick glance, this one of the calls to Put_Line in

     <<Dereference_Label>>

      if not Valid then
         if Pool.Raise_Exceptions then
            raise Accessing_Not_Allocated_Storage;
         else
            Put (Output_File (Pool),
                 "error: Accessing not allocated storage, at ");
            Put_Line (Output_File (Pool), Pool.Stack_Trace_Depth, null,
                      Dereference_Label'Address,
                      Code_Address_For_Dereference_End);
         end if;

      else
         Header := Header_Of (Storage_Address);

         if Header.Block_Size < 0 then
            if Pool.Raise_Exceptions then
               raise Accessing_Deallocated_Storage;
            else
               Put (Output_File (Pool),
                    "error: Accessing deallocated storage, at ");
               Put_Line
                 (Output_File (Pool), Pool.Stack_Trace_Depth, null,
                  Dereference_Label'Address,
                  Code_Address_For_Dereference_End);


 which indeed passe the address of the <<Dereference_Label>> code
 label.

 Compiling -da then grep suggests that the label disappears as of

  ada/rts/g-debpoo.adb.153r.vregs:
   (code_label/s 12 11 13 4 538 ("dereference_label") [4 uses])

  ada/rts/g-debpoo.adb.154r.into_cfglayout:
   (note/s 12 11 14 2 ("dereference_label") NOTE_INSN_DELETED_LABEL 538)

 This is still very rough and obviously needs further investigation, but
 yields starters.

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[dave at hiauly1 dot hia.nrc.ca]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #20 from dave at hiauly1 dot hia.nrc.ca 2010-10-01 18:34:45 UTC ---
> --- Comment #18 from Jan Hubicka <hubicka at gcc dot gnu.org> 2010-10-01 17:24:27 UTC ---
> Hmm, I tried to use one of compile farm sparc machines, but no luck. Either
> with ssh on them I end on gcc12 that is amd64 or they are down.
> I am looking into the dumps if I can make sense of it, so far I don't seem to
> be able to find the code taking address of local label in .optimized dump at
> all.

I had trouble too.  I kept all the dumps just in case we needed something
earlier.

Dave

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #19 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2010-10-01 17:35:47 UTC ---
> Hmm, I tried to use one of compile farm sparc machines, but no luck. Either
> with ssh on them I end on gcc12 that is amd64 or they are down.

Try grobluk (ssh -p 9202 login@gcc12.fsffrance.org), it's a 6x machine.  Ada
compilers are available in /opt/cfarm/release.

> I am looking into the dumps if I can make sense of it, so far I don't seem to
> be able to find the code taking address of local label in .optimized dump at
> all.
> 
> We might be missing some code in inliner that passes the fact that the label is
> user label...

I'll look into this next week.

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[hubicka at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #18 from Jan Hubicka <hubicka at gcc dot gnu.org> 2010-10-01 17:24:27 UTC ---
Hmm, I tried to use one of compile farm sparc machines, but no luck. Either
with ssh on them I end on gcc12 that is amd64 or they are down.
I am looking into the dumps if I can make sense of it, so far I don't seem to
be able to find the code taking address of local label in .optimized dump at
all.

We might be missing some code in inliner that passes the fact that the label is
user label...

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[hainque at adacore dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #17 from hainque at adacore dot com <hainque at adacore dot com> 2010-09-30 09:23:25 UTC ---
ro at gcc dot gnu.org wrote:
> Eric, Olivier,

> could you please have a look at Jan's question in Comment #6?  This
> bug currently breaks Ada bootstrap on Solaris 2/SPARC and hampers my
> Solaris testing.

 I'll see if I can get at something with a cross (faster to build), but
 please don't give up other tracks as I have very limited bandwidth for
 this right now.

 Olivier

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[dave at hiauly1 dot hia.nrc.ca]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #15 from dave at hiauly1 dot hia.nrc.ca 2010-09-30 01:26:27 UTC ---
On Thu, 30 Sep 2010, dave at hiauly1 dot hia.nrc.ca wrote:

For reference, I have also attached the hppa2.0w-hp-hpux11.11 .s file.
The problem is the reference to L$0371 in this insn:

        ldo L$0371-L$0375(%r1),%r1

Dave

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[dave at hiauly1 dot hia.nrc.ca]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #13 from dave at hiauly1 dot hia.nrc.ca 2010-09-30 01:18:27 UTC ---
On Tue, 28 Sep 2010, hubicka at ucw dot cz wrote:

> Hi,
> I would really apprechiate the answer to comment #6.  Since the patch should
> only
> introduce more (valid) inlining this must be latent bug somewhere....

Attached .optimized and .nothrow from hppa2.0w-hp-hpux11.11.  It appears
to me that the call is attempting to pass a code location.

Dave

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[dave at hiauly1 dot hia.nrc.ca]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #12 from dave at hiauly1 dot hia.nrc.ca 2010-09-28 15:32:01 UTC ---
> Hi,
> I would really apprechiate the answer to comment #6.  Since the patch should
> only
> introduce more (valid) inlining this must be latent bug somewhere....

In the hppa-hpux case, the label would have had to be a code label.
The linker and dynamic loader don't support label differences between
code and data symbols.  This suggests that the change exposed an
inlining bug that removed the symbol.

I'll take a look on my next build.

Dave

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #11 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2010-09-28 12:34:28 UTC ---
> Eric, Olivier,
> 
> could you please have a look at Jan's question in Comment #6?  This bug
> currently breaks Ada bootstrap on Solaris 2/SPARC and hampers my Solaris
> testing.

Sorry, I won't really have time to look into this before a couple of weeks. 
The GCC Compile Farm has several SPARC/Linux boxes with pre-built Ada
compilers.

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

Rainer Orth <ro at gcc dot gnu.org> changed:

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

--- Comment #10 from Rainer Orth <ro at gcc dot gnu.org> 2010-09-28 12:11:35 UTC ---
Eric, Olivier,

could you please have a look at Jan's question in Comment #6?  This bug
currently
breaks Ada bootstrap on Solaris 2/SPARC and hampers my Solaris testing.

Thanks,
  Rainer

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[hubicka at ucw dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #9 from Jan Hubicka <hubicka at ucw dot cz> 2010-09-28 12:07:54 UTC ---
Hi,
I would really apprechiate the answer to comment #6.  Since the patch should
only
introduce more (valid) inlining this must be latent bug somewhere....

Honza

[Bug bootstrap/45612] [4.6 regression] Reference to undefined label building libada on Solaris 2/SPARC

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45612

--- Comment #8 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2010-09-27 17:48:14 UTC ---
Jan,

again, no progress at all for 2 1/2 weeks on a bootstrap failure on a
primary platform.  Please fix or revert.

    Rainer