[Bug libstdc++/12175] New: Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[Bug libstdc++/12175] New: Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[gcc-bugzilla at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175

           Summary: Tru64 UNIX bootstrap failure: multiply defined symbols
                    linking libstdc++.so
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: critical
          Priority: P2
         Component: libstdc++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ro at techfak dot uni-bielefeld dot de
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: alpha-dec-osf4.0f
  GCC host triplet: alpha-dec-osf4.0f
GCC target triplet: alpha-dec-osf4.0f

Bootstrapping on alpha-dec-osf4.0f (and alpha-dec-osf5.1) started failing
between 20030804 and 20030819: libstdc++.so fails to link due to duplicate
symbols.

/bin/ksh ../libtool --tag CXX --mode=link /vol/gcc/obj/gcc-3.4-20030901/4.0f-gcc/gcc/xgcc -shared-libgcc -B/vol/gcc/obj/gcc-3.4-20030901/4.0f-gcc/gcc/ -nostdinc++  -L/vol/gcc/obj/gcc-3.4-20030901/4.0f-gcc/alpha-dec-osf4.0f/libstdc++-v3/src -L/vol/gcc/obj/gcc-3.4-20030901/4.0f-gcc/alpha-dec-osf4.0f/libstdc++-v3/src/.libs -B/vol/gcc/share/alpha-dec-osf4.0f/bin/ -B/vol/gcc/share/alpha-dec-osf4.0f/lib/ -isystem /vol/gcc/share/alpha-dec-osf4.0f/include -isystem /vol/gcc/share/alpha-dec-osf4.0f/sys-include   -fno-implicit-templates -Wall -Wno-format -W -Wwrite-strings  -fdiagnostics-show-location=once     -o libstdc++.la -rpath /vol/gcc/lib -version-info 6:0:0  -lm  allocator-inst.lo codecvt.lo complex_io.lo concept-inst.lo ctype.lo demangle.lo ext-inst.lo fstream-inst.lo functexcept.lo globals.lo io-inst.lo ios.lo istream-inst.lo limits.lo locale.lo locale-inst.lo localename.lo misc-inst.lo ostream-inst.lo sstream-inst.lo stdexcept.lo stl_tree.lo streambuf-inst.lo string-inst.lo !
 st!
rstream.lo valarray-inst.lo wstring-inst.lo codecvt_members.lo collate_members.lo ctype_members.lo messages_members.lo monetary_members.lo numeric_members.lo time_members.lo basic_file.lo c++locale.lo ../libmath/libmath.la ../libsupc++/libsupc++convenience.la -lm
/usr/bin/ld:
.libs/misc-inst.o: _GLOBAL__F__ZNKSt5ctypeIcE2isEjc: multiply defined
.libs/ostream-inst.o: _GLOBAL__F__ZNSt15basic_streambufIcSt11char_traitsIcEE5sputcEc: multiply defined
.libs/sstream-inst.o: _GLOBAL__F__ZNSt15basic_streambufIcSt11char_traitsIcEE5sputcEc: multiply defined
collect2: ld returned 1 exit status

E.g., the first of those three is defined here:

_GLOBAL__F__ZNKSt5ctypeIcE2isEjc | 0000000000103432 | D | 0000000000000000
.libs/locale-inst.o
_GLOBAL__F__ZNKSt5ctypeIcE2isEjc | 0000000000004928 | D | 0000000000000000
.libs/misc-inst.o

This symbol didn't occur in the 20030804 versions of those objects.

Environment:
System: OSF1 rimsky V4.0 1229 alpha
Machine: alpha
	
host: alpha-dec-osf4.0f
build: alpha-dec-osf4.0f
target: alpha-dec-osf4.0f
configured with: '/vol/gnu/src/gcc/gcc-dist/configure' '--prefix=/vol/gcc '--with-local-prefix=/vol/gcc '--disable-nls 'alpha-dec-osf4.0f

How-To-Repeat:
Bootstrap as above.

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hubicka at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
          Component|libstdc++                   |c++
     Ever Confirmed|                            |1
           Keywords|                            |wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2003-09-07 02:38:21
               date|                            |
            Summary|Tru64 UNIX bootstrap        |[3.4 Regression] Tru64 UNIX
                   |failure: multiply defined   |bootstrap failure: multiply
                   |symbols linking libstdc++.so|defined symbols linking
                   |                            |libstdc++.so
   Target Milestone|---                         |3.4


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-07 02:38 -------
I had the same problem on i686-unknown-openbsd3.1 on a testcase in the testsuite but it 
never got fixed, Jan Hubicka tried to fix it but the patches did not work.
The problem is that the code to pick the name for the global initializer in the file is picking 
up weak symbols for some reason.
I cannot test any more patches for this problem as I have a different bootstrap problem on 
i686-unknown-openbsd3.1 now for some reason.

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From hubicka at ucw dot cz  2003-09-07 11:15 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

> never got fixed, Jan Hubicka tried to fix it but the patches did not work.
> The problem is that the code to pick the name for the global initializer in the file is picking 
> up weak symbols for some reason.

I will look into deeper today.  I reorganized parts of this code
yesterday so I hope it is already fixed, but lets see what is going on.

Honza
> I cannot test any more patches for this problem as I have a different bootstrap problem on 
> i686-unknown-openbsd3.1 now for some reason.
> 
> 
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175


pinskia at gcc dot gnu dot org changed:

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


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-12 14:07 -------
*** Bug 12259 has been marked as a duplicate of this bug. ***

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[dave at hiauly1 dot hia dot nrc dot ca]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-12 14:37 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu

> ------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-12
> 14:07 -------
> *** Bug 12259 has been marked as a duplicate of this bug. ***

This may be the same bug but I had a successful build at
Tue Sep  9 21:24:42 UTC 2003.  The Tru64 UNIX build started
failing in August.

Dave

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[dave at hiauly1 dot hia dot nrc dot ca]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-12 18:23 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure

> This may be the same bug but I had a successful build at
> Tue Sep  9 21:24:42 UTC 2003.  The Tru64 UNIX build started
> failing in August.

The symbol wasn't in the library on Sep  9.  On hppa2.0-hp-hpux11.11,
it now appears as a global symbol in four objects.  There would appear
to be a problem with template instantiation.  It's a bit strange
that it's only this one symbol that has a problem.

On hppa-linux, the symbol occurs in four objects and isn't global:

00000000 t _GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_

The symbol is in the .ctors list of each object.  I don't think we
should be generating a global constructor symbol.  There aren't any
others.

Dave

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[ro at techfak dot uni-bielefeld dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2003-09-15 17:24 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

I just tried a new bootstrap on alpha-dec-osf4.0f.  Apart from an unrelated
problem (fold-const.c required more than the default 128 MB of data space,
a problem that bit me for libjava/verify.cc before), building libstdc++.so
still fails:

/usr/bin/ld:
.libs/codecvt.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/complex_io.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/ctype.o: _GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/ctype.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/ext-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/fstream-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/functexcept.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/io-inst.o: _GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/io-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/ios.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/istream-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/locale.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/locale-inst.o: _GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/locale-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/localename.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/misc-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/ostream-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/sstream-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/stdexcept.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/streambuf-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/string-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/strstream.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/valarray-inst.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/ctype_members.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/messages_members.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/monetary_members.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/numeric_members.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/time_members.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/basic_file.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
.libs/c++locale.o: _GLOBAL__F__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_: multiply defined
collect2: ld returned 1 exit status
make[4]: *** [libstdc++.la] Error 1

So this is the same symbol as on hppa-linux, and the only problematic one
as well.

	Rainer

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[ro at techfak dot uni-bielefeld dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2003-09-26 03:33 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

Jan,

any progress on this one?  Tru64 UNIX bootstrap is broken for more than a
month now, with no indication of a fix whatsoever.

	Rainer

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From hubicka at ucw dot cz  2003-09-26 08:44 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175
> 
> 
> 
> ------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2003-09-26 03:33 -------
> Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so
> 
> Jan,
> 
> any progress on this one?  Tru64 UNIX bootstrap is broken for more than a
> month now, with no indication of a fix whatsoever.

I tought it is fixed already.  I will try to look into this over
weekend. Sorry for that!

Honza
> 
> 	Rainer
> 
> 
> 
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From hubicka at ucw dot cz  2003-09-27 20:25 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175
> 
> 
> 
> ------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2003-09-26 03:33 -------
> Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so
> 
> Jan,
> 
> any progress on this one?  Tru64 UNIX bootstrap is broken for more than a
> month now, with no indication of a fix whatsoever.

OK, I was trying to reproduce it with Linux headers and I can't get that
symbol to be output by crosscompiler.  Can you please send me
preprocessed file of some of those files that do produce that symbol
(and are not supposed to do so)?

Honza
> 
> 	Rainer
> 
> 
> 
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[dave at hiauly1 dot hia dot nrc dot ca]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-28 16:51 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu

> Can you also send me the assembly file you get with latest GCC snapshot?
> You may also try to breakpoint on notice_global_symbol and try to figure
> out why the symbol passes the tests even tought it is not global.

notice_global_symbol isn't called with the symbol
"_GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_".  The closest
are "_ZSt23lexicographical_comparePKhS0_S0_S0_" and
"_ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_".

For the latter we have:

(gdb) p debug_tree (decl)
 <function_decl 7ab79150 lexicographical_compare
    type <function_type 7ab79070
	type <boolean_type 7aedd850 bool unsigned QI
	    size <integer_cst 7aedb198 constant 8>
	    unit size <integer_cst 7aedb1b0 constant 1>
	    align 8 symtab 4 alias set -1 precision 1 min <integer_cst 7aedb570 0> max <integer_cst 7aedb5b8 1>>
	DI
	size <integer_cst 7aedb750 constant 64>
	unit size <integer_cst 7aedbb88 constant 8>
	align 64 symtab 0 alias set -1
	arg-types <tree_list 7ab78d20 value <pointer_type 7ab79000>
	    chain <tree_list 7ab78d08 value <pointer_type 7ab79000>
		chain <tree_list 7ab78cf0 value <pointer_type 7ab79000>
		    chain <tree_list 7ab78cd8 value <pointer_type 7ab79000>
			chain <tree_list 7aece018 tree_2 value <void_type 7aec8cb0 void>>>>>>>
    used nothrow public static in_system_header external decl_1 decl_5 SI file /xxx/gnu/gcc-3.3/objdir/hppa2.0w-hp-hpux11.11/libstdc++-v3/include/bits/stl_algobase.h line 712 context <namespace_decl 7aeddd20 std>
    arguments <parm_decl 7ab797e0 __first1
	type <pointer_type 7ab79000 type <integer_type 7ab82850 signed char>
	    unsigned SI
	    size <integer_cst 7aedbbd0 constant 32>
	    unit size <integer_cst 7aedbc30 constant 4>
	    align 32 symtab 0 alias set -1>
	unsigned used in_system_header SI file /xxx/gnu/gcc-3.3/objdir/hppa2.0w-hp-hpux11.11/libstdc++-v3/include/bits/stl_algobase.h line 712 size <integer_cst 7aedbbd0 32> unit size <integer_cst 7aedbc30 4>
	align 32 context <function_decl 7ab79150 lexicographical_compare> initial <pointer_type 7ab79000> arg-type <pointer_type 7ab79000>
	chain <parm_decl 7ab79850 __last1 type <pointer_type 7ab79000>
	     unsigned used in_system_header SI file /xxx/gnu/gcc-3.3/objdir/hppa2.0w-hp-hpux11.11/libstdc++-v3/include/bits/stl_algobase.h line 712 size <integer_cst 7aedbbd0 32> unit size <integer_cst 7aedbc30 4>
	     align 32 context <function_decl 7ab79150 lexicographical_compare> initial <pointer_type 7ab79000> arg-type <pointer_type 7ab79000> chain <parm_decl 7ab798c0 __first2>>>
    result <result_decl 7ab799a0
	type <integer_type 7aedd5b0 int type_6 SI
	    size <integer_cst 7aedb2e8 constant 32>
	    unit size <integer_cst 7aedb390 constant 4>
	    align 32 symtab 3 alias set -1 precision 32 min <integer_cst 7aedb360 -2147483648> max <integer_cst 7aedb378 2147483647>
	    pointer_to_this <pointer_type 7aec9150>>
	in_system_header SI file /xxx/gnu/gcc-3.3/objdir/hppa2.0w-hp-hpux11.11/libstdc++-v3/include/bits/stl_algobase.h line 712 size <integer_cst 7aedb2e8 32> unit size <integer_cst 7aedb390 4>
	align 32 context <function_decl 7ab79150 lexicographical_compare>> initial <block 7ab99de8>
    pending-inline-info 7ab79bd0 template-info 7ab78d50
    (mem:SI (symbol_ref/v:SI ("@_ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_") <function_decl 7ab79150 lexicographical_compare>) [0 S4 A32])
    saved-insns 7ab7b000>

Dave

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[dave at hiauly1 dot hia dot nrc dot ca]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-28 17:40 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure

> > Can you also send me the assembly file you get with latest GCC snapshot?
> > You may also try to breakpoint on notice_global_symbol and try to figure
> > out why the symbol passes the tests even tought it is not global.
> 
> notice_global_symbol isn't called with the symbol

Sorry this is wrong.  We have the following decl:

(gdb) p debug_tree ($r26)
 <function_decl 7a45e700 _GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_
    type <function_type 7aecf690
	type <void_type 7aec8cb0 void type_6 VOID
	align 8 symtab 20 alias set -1
	pointer_to_this <pointer_type 7aec8d20>>
	type_6 DI
	size <integer_cst 7aedb750 constant 64>
	unit size <integer_cst 7aedbb88 constant 8>
	align 64 symtab 757 alias set -1
	arg-types <tree_list 7aece018 tree_2 value <void_type 7aec8cb0 void>>
	pointer_to_this <pointer_type 7add2a10>>
    used nothrow public static decl_5 SI file ../../../../gcc/libstdc++-v3/src/codecvt.cc line 152
    result <result_decl 7a45e770 type <void_type 7aec8cb0 void>
	VOID file ../../../../gcc/libstdc++-v3/src/codecvt.cc line 152
	align 8 context <function_decl 7a45e700 _GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_>> initial <block 7a528ac8>
    pending-inline-info 7a45e930
    saved-insns 7a71f200 chain <function_decl 7a47cee0 __static_initialization_and_destruction_0>>

(gdb) p first_global_object_name
$32 = 0x400c7900 "_ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_"
(gdb) p weak_global_object_name
$33 = 0x400c7b48 "_ZSt23lexicographical_comparePKhS0_S0_S0_"

Thus, we don't set first_global_object_name or weak_global_object_name
for _GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_.

Dave

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From hubicka at ucw dot cz  2003-09-28 20:17 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

> > Can you also send me the assembly file you get with latest GCC snapshot?
> > You may also try to breakpoint on notice_global_symbol and try to figure
> > out why the symbol passes the tests even tought it is not global.
> 
> Attached is the assembler file.  I will try a quick look at
> notice_global_symbol.
The symbol in question exists and is external symbol.
I am not quite sure what can call notice_global_symbol on external
symbol, but perhaps the attached patch will help.

Index: varasm.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/varasm.c,v
retrieving revision 1.386
diff -c -3 -p -r1.386 varasm.c
*** varasm.c	23 Sep 2003 19:47:00 -0000	1.386
--- varasm.c	28 Sep 2003 20:15:53 -0000
*************** notice_global_symbol (tree decl)
*** 1043,1048 ****
--- 1043,1049 ----
  {
    if ((!first_global_object_name || !weak_global_object_name)
        && TREE_PUBLIC (decl) && !DECL_COMMON (decl)
+       && !DECL_EXTERNAL (decl)
        && (TREE_CODE (decl) == FUNCTION_DECL
  	  || (TREE_CODE (decl) == VAR_DECL
  	      && (DECL_INITIAL (decl) != 0

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From hubicka at ucw dot cz  2003-09-28 20:19 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175
> 
> 
> 
> ------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-28 16:51 -------
> Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu
> 
> > Can you also send me the assembly file you get with latest GCC snapshot?
> > You may also try to breakpoint on notice_global_symbol and try to figure
> > out why the symbol passes the tests even tought it is not global.
> 
> notice_global_symbol isn't called with the symbol
> "_GLOBAL__I__ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_".  The closest
> are "_ZSt23lexicographical_comparePKhS0_S0_S0_" and
> "_ZSt23lexicographical_compareIPKaS1_EbT_S2_T0_S3_".

Yes, this is the decl I am looking for (the _GLOBAL symbol is added later).  It
really is external so probably the patch I sent you will help.
Can you send me backtrace of this call?
It seems wrong to me that someone calls notice_global_symbol on it.
Thank you for investigating this!

Honza

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[dave at hiauly1 dot hia dot nrc dot ca]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-28 20:39 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu

> Yes, this is the decl I am looking for (the _GLOBAL symbol is added later).
> It
> really is external so probably the patch I sent you will help.
> Can you send me backtrace of this call?
> It seems wrong to me that someone calls notice_global_symbol on it.

I think there is only one call that matters here, that's the one
in assemble_start_function.  I believe that there is a small bug
in notice_global_symbol.  It seems to me the function of this
routine is just to set appropriate values for first_global_object_name
and weak_global_object_name in a given translation unit.  The way
the code is written first_global_object_name or weak_global_object_name
can set several times until both get set.  I don't know if this matters
but we certainly don't get the "first" global object name saved in
first_global_object_name.s

Possibly, you are correct that we are accidently picking an external
symbol for the first global object name and this is causing the duplication
in constructor names.

Dave

Re: [Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[Jan Hubicka]

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175
> 
> 
> 
> ------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-28 20:39 -------
> Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu
> 
> > Yes, this is the decl I am looking for (the _GLOBAL symbol is added later).
> > It
> > really is external so probably the patch I sent you will help.
> > Can you send me backtrace of this call?
> > It seems wrong to me that someone calls notice_global_symbol on it.
> 
> I think there is only one call that matters here, that's the one
> in assemble_start_function.  I believe that there is a small bug
> in notice_global_symbol.  It seems to me the function of this
> routine is just to set appropriate values for first_global_object_name
> and weak_global_object_name in a given translation unit.  The way
> the code is written first_global_object_name or weak_global_object_name
> can set several times until both get set.  I don't know if this matters
> but we certainly don't get the "first" global object name saved in
> first_global_object_name.s

This does not matter much - in fact we need some global object name, but
you are right that we should avoid duplicated work.

So the call comes from assemble_start_function that calls it on external
decl?

Honza
> 
> Possibly, you are correct that we are accidently picking an external
> symbol for the first global object name and this is causing the duplication
> in constructor names.
> 
> Dave

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From hubicka at ucw dot cz  2003-09-28 20:49 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175
> 
> 
> 
> ------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-28 20:39 -------
> Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu
> 
> > Yes, this is the decl I am looking for (the _GLOBAL symbol is added later).
> > It
> > really is external so probably the patch I sent you will help.
> > Can you send me backtrace of this call?
> > It seems wrong to me that someone calls notice_global_symbol on it.
> 
> I think there is only one call that matters here, that's the one
> in assemble_start_function.  I believe that there is a small bug
> in notice_global_symbol.  It seems to me the function of this
> routine is just to set appropriate values for first_global_object_name
> and weak_global_object_name in a given translation unit.  The way
> the code is written first_global_object_name or weak_global_object_name
> can set several times until both get set.  I don't know if this matters
> but we certainly don't get the "first" global object name saved in
> first_global_object_name.s

This does not matter much - in fact we need some global object name, but
you are right that we should avoid duplicated work.

So the call comes from assemble_start_function that calls it on external
decl?

Honza
> 
> Possibly, you are correct that we are accidently picking an external
> symbol for the first global object name and this is causing the duplication
> in constructor names.
> 
> Dave

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[dave at hiauly1 dot hia dot nrc dot ca]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-29 11:57 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu

> The symbol in question exists and is external symbol.
> I am not quite sure what can call notice_global_symbol on external
> symbol, but perhaps the attached patch will help.

Yes, it does.  It gets us back to bootstrap land.

Thanks,
Dave

> Index: varasm.c
> ===================================================================
> RCS file: /cvs/gcc/gcc/gcc/varasm.c,v
> retrieving revision 1.386
> diff -c -3 -p -r1.386 varasm.c
> *** varasm.c	23 Sep 2003 19:47:00 -0000	1.386
> --- varasm.c	28 Sep 2003 20:15:53 -0000
> *************** notice_global_symbol (tree decl)
> *** 1043,1048 ****
> --- 1043,1049 ----
>   {
>     if ((!first_global_object_name || !weak_global_object_name)
>         && TREE_PUBLIC (decl) && !DECL_COMMON (decl)
> +       && !DECL_EXTERNAL (decl)
>         && (TREE_CODE (decl) == FUNCTION_DECL
>   	  || (TREE_CODE (decl) == VAR_DECL
>   	      && (DECL_INITIAL (decl) != 0

Dave

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[ro at techfak dot uni-bielefeld dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2003-09-29 13:47 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

Same on alpha-dec-osf4.0f.

Thanks.
	Rainer

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From hubicka at ucw dot cz  2003-09-29 14:17 -------
Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175
> 
> 
> 
> ------- Additional Comments From dave at hiauly1 dot hia dot nrc dot ca  2003-09-29 11:57 -------
> Subject: Re:  [3.4 Regression] Tru64 UNIX bootstrap failure: mu
> 
> > The symbol in question exists and is external symbol.
> > I am not quite sure what can call notice_global_symbol on external
> > symbol, but perhaps the attached patch will help.
> 
> Yes, it does.  It gets us back to bootstrap land.

OK,
I will install the patch as obvious then together with the change to 
set the variables just once.
Thanks for testing it!
Honza
> 
> Thanks,
> Dave
> 
> > Index: varasm.c
> > ===================================================================
> > RCS file: /cvs/gcc/gcc/gcc/varasm.c,v
> > retrieving revision 1.386
> > diff -c -3 -p -r1.386 varasm.c
> > *** varasm.c	23 Sep 2003 19:47:00 -0000	1.386
> > --- varasm.c	28 Sep 2003 20:15:53 -0000
> > *************** notice_global_symbol (tree decl)
> > *** 1043,1048 ****
> > --- 1043,1049 ----
> >   {
> >     if ((!first_global_object_name || !weak_global_object_name)
> >         && TREE_PUBLIC (decl) && !DECL_COMMON (decl)
> > +       && !DECL_EXTERNAL (decl)
> >         && (TREE_CODE (decl) == FUNCTION_DECL
> >   	  || (TREE_CODE (decl) == VAR_DECL
> >   	      && (DECL_INITIAL (decl) != 0
> 
> Dave
> 
> 
> 
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[cvs-commit at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175



------- Additional Comments From cvs-commit at gcc dot gnu dot org  2003-09-29 21:19 -------
Subject: Bug 12175

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	hubicka@gcc.gnu.org	2003-09-29 21:19:11

Modified files:
	gcc            : ChangeLog varasm.c gcov-io.c profile.c 
	gcc/cp         : ChangeLog except.c 

Log message:
	PR c++/12175
	* varasm.c (notice_global_symbol): Discard external symbols.
	
	PR optimization/12286
	* gcov-io.c (gcov_read_words): Fix memmove call.
	* profile.c (compute_branch_probabilities): Add extra sanity checks.
	
	PR C++/12047
	* except.c (build_eh_type_type): Call mark_used on the type.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.1216&r2=2.1217
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/varasm.c.diff?cvsroot=gcc&r1=1.386&r2=1.387
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/gcov-io.c.diff?cvsroot=gcc&r1=1.12&r2=1.13
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/profile.c.diff?cvsroot=gcc&r1=1.125&r2=1.126
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/ChangeLog.diff?cvsroot=gcc&r1=1.3709&r2=1.3710
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cp/except.c.diff?cvsroot=gcc&r1=1.159&r2=1.160

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-29 21:32 -------
Fixed by the patch.

[Bug c++/12175] [3.4 Regression] Tru64 UNIX bootstrap failure: multiply defined symbols linking libstdc++.so

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-25 03:23 -------
*** Bug 12291 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |phil dot el at wanadoo dot
                   |                            |fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12175