GCC Bugzilla Bug 8691

The following program fails to build with 
`-O3 -fno-implicit-templates':

lidiaadm@lap200:~/gcc-3.2/bugs> cat string-O3.cc 
#include<string>

int main() {
    std::string ("Hello World!");

    return 0;
}

lidiaadm@lap200:~/gcc-3.2/bugs> g++ -Wall -O3 -fno-implicit-templates -o string-O3 string-O3.cc 
/tmp/ccQrTsAF.o: In function `main':
/tmp/ccQrTsAF.o(.text+0x40): undefined reference to `char* std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_S_construct<char const*>(char const*, char const*, std::allocator<char> const&, std::forward_iterator_tag)'
collect2: ld returned 1 exit status


Experimenting with `-fkeep-inline-functions' etc. didn't help.
If string-O3.cc is compiled with `-O2 -fno-implicit-templates' 
or simply `-O3', everything is ok.

It seems the optimizer interacts in some strange way with 
template instantiation. Since we observed the problem with
std::string, one may put this PR in the category libstdc++.
But in my opinion the user can't be supposed to anticipate
what the optimizer is going to do; therefore the set of 
required explicit template instantiations should not be 
depend on the optimization level and I think it's a bug in
the C++ front end.

If you don't consider gcc's behaviour a bug, then I request
that the documentation explicitly states that one can't use 
`-fno-implicit-templates' with `-O3' or higher.

Release:
gcc version 3.2

Environment:
linux 2.4.18 i686

How-To-Repeat:
g++ -Wall -O3 -fno-implicit-templates -o string-O3 string-O3.cc

Fix:
Workaround: Restrict optimization to `-O2'.

State-Changed-From-To: open->feedback
State-Changed-Why: I cannot reproduce the bug with 3.2 (or 3.1 or 3.2.1)
    on i686-pc-linux-gnu (using the original testcase and the
    preprocessed version).
    Are you really using the gcc 3.2 release or some prerelease
    or some modified version from a LINUX distributor instead?
    If the latter is true, you should try to install gcc 3.2.1
    from source.
    Please contact us, whether the problems still persist
    or vanished. In the latter case, please add "-v" to the
    command line options and provide us with the output so that we
    we can investigate the problem.

State-Changed-From-To: feedback->analyzed
State-Changed-Why: Confirmed.
    
    May be a binutils issue.

From: Paolo Carlini <pcarlini@unitus.it>
To: gcc-gnats@gcc.gnu.org,  gcc-prs@gcc.gnu.org, 
 cludwig@cdc.informatik.tu-darmstadt.de,  gcc-bugs@gcc.gnu.org, 
 nobody@gcc.gnu.org, Reichelt <reichelt@igpm.rwth-aachen.de>
Cc:  
Subject: Re: c++/8691: -O3 and -fno-implicit-templates are incompatible
Date: Mon, 25 Nov 2002 00:50:50 +0100

 Hi,
 
 I can reproduce the problem.
 
 Ciao, Paolo.
 
 ////////
 
 paolo:~> g++ -v -O3 -fno-implicit-templates 8691.cc
 Reading specs from /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/specs
 Configured with: ../gcc-3.2.1/configure --prefix=/usr
 Thread model: posix
 gcc version 3.2.1
   /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/cc1plus -v -D__GNUC__=3 -D__GNUC_MINOR__=2 -D__GNUC_PATCHLEVEL__=1 -D__GXX_ABI_VERSION=102 -D__ELF__ -Dunix 
 -D__gnu_linux__ -Dlinux -D__ELF__ -D__unix__ -D__gnu_linux__ -D__linux__ -D__unix -D__linux -Asystem=posix -D__OPTIMIZE__ -D__STDC_HOSTED__=1 
 -D_GNU_SOURCE -Acpu=i386 -Amachine=i386 -Di386 -D__i386 -D__i386__ -D__tune_i686__ -D__tune_pentiumpro__ 8691.cc -D__GNUG__=3 -D__DEPRECATED 
 -D__EXCEPTIONS -quiet -dumpbase 8691.cc -O3 -version -fno-implicit-templates -o /tmp/cc1au9LA.s
 GNU CPP version 3.2.1 (cpplib) (i386 Linux/ELF)
 GNU C++ version 3.2.1 (i686-pc-linux-gnu)
          compiled by GNU C version 3.2.1.
 #include "..." search starts here:
 #include <...> search starts here:
   /usr/include/c++/3.2.1
   /usr/include/c++/3.2.1/i686-pc-linux-gnu
   /usr/include/c++/3.2.1/backward
   /usr/local/include
   /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/include
   /usr/i686-pc-linux-gnu/include
   /usr/include
 End of search list.
   /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/../../../../i686-pc-linux-gnu/bin/as -V -Qy -o /tmp/cc457xQo.o /tmp/cc1au9LA.s
 GNU assembler version 2.13.1 (i686-pc-linux-gnu) using BFD version 2.13.1
   /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/collect2 --eh-frame-hdr -m elf_i386 -dynamic-linker /lib/ld-linux.so.2 
 /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/../../../crt1.o /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/../../../crti.o 
 /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/crtbegin.o -L/usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1 
 -L/usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/../../../../i686-pc-linux-gnu/lib -L/usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/../../.. /tmp/cc457xQo.o 
 -lstdc++ -lm -lgcc_s -lgcc -lc -lgcc_s -lgcc /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/crtend.o /usr/lib/gcc-lib/i686-pc-linux-gnu/3.2.1/../../../crtn.o
 /tmp/cc457xQo.o: In function `main':
 /tmp/cc457xQo.o(.text+0x40): undefined reference to `char* std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_S_construct<char 
 const*>(char const*, char const*, std::allocator<char> const&, std::forward_iterator_tag)'
 collect2: ld returned 1 exit status
 
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8691
 

From: Paolo Carlini <pcarlini@unitus.it>
To: gcc-gnats@gcc.gnu.org,  gcc-prs@gcc.gnu.org, 
 cludwig@cdc.informatik.tu-darmstadt.de,  gcc-bugs@gcc.gnu.org, 
 nobody@gcc.gnu.org, Reichelt <reichelt@igpm.rwth-aachen.de>
Cc:  
Subject: Re: c++/8691: -O3 and -fno-implicit-templates are incompatible
Date: Mon, 25 Nov 2002 01:05:36 +0100

 ... however, I *cannot* reproduce it with current mainline.
 
 I think we should mark it as a 3.2 regression.
 
 Volker?
 
 Ciao, Paolo.
 
 //////
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8691
 

From: Volker Reichelt <reichelt@igpm.rwth-aachen.de>
To: pcarlini@unitus.it
Cc: gcc-gnats@gcc.gnu.org, cludwig@cdc.informatik.tu-darmstadt.de,
        gcc-bugs@gcc.gnu.org, nobody@gcc.gnu.org
Subject: Re: c++/8691: -O3 and -fno-implicit-templates are incompatible
Date: Mon, 25 Nov 2002 02:45:18 +0100

 On 25 Nov, Paolo Carlini wrote:
 > ... however, I *cannot* reproduce it with current mainline.
 > 
 > I think we should mark it as a 3.2 regression.
 
 Ok, here are the latest news: This may be a binutils issue.
 I'm running binutils 2.10.91:                      no bug
 You are running 2.13.1:                            bug
 My assembler file on a machine with binutils 2.12: bug
 
 I'll append to assembler file to the PR.
 
 I cannot reproduce the problem with gcc 3.0.4 and mainline.
 Thus, it's a 3.2 regression.
 
 Ciao,
 Volker
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8691
 
 

From: Christoph Ludwig <cludwig@cdc.informatik.tu-darmstadt.de>
To: reichelt@igpm.rwth-aachen.de, gcc-bugs@gcc.gnu.org,
	gcc-prs@gcc.gnu.org, nobody@gcc.gnu.org, gcc-gnats@gcc.gnu.org
Cc:  
Subject: Re: c++/8691: -O3 and -fno-implicit-templates are incompatible
Date: Mon, 25 Nov 2002 09:20:59 +0100

 Others already confirmed they can reproduce the bug with specific 
 binutils versions. Just for completeness, here is the data you asked for:
 
 lidiaadm@lap200:~> g++ -v
 Reading specs from /opt/gcc/gcc-3.2/lib/gcc-lib/i686-pc-linux-gnu/3.2/specs
 Configured with: /home/lidiaadm/tmp/gcc-3.2-src/gcc-3.2/configure --prefix=/opt/gcc/gcc-3.2 --enable-threads=posix --enable-version-specific-runtime-libs --enable-__cxa_atexit
 Thread model: posix
 gcc version 3.2
 
 I am running SuSE Linux with the binutils that shipped with the
 distribution: 
 
 lidiaadm@lap200:~> ld -v
 GNU ld version 2.11.92.0.10 20011021 (SuSE)
 
 Just in case it helps you tracking down the problem's cause: On my
 system, gcc 3.1 is the first compiler release that fails to build the
 test program. 3.0.4 builds it fine.
 
 Regards
 
 Christoph
 
 
 On Sun, Nov 24, 2002 at 09:30:50PM -0000, reichelt@igpm.rwth-aachen.de wrote:
 > Synopsis: -O3 and -fno-implicit-templates are incompatible
 > 
 > State-Changed-From-To: open->feedback
 > State-Changed-By: reichelt
 > State-Changed-When: Sun Nov 24 13:30:49 2002
 > State-Changed-Why:
 >     I cannot reproduce the bug with 3.2 (or 3.1 or 3.2.1)
 >     on i686-pc-linux-gnu (using the original testcase and the
 >     preprocessed version).
 >     Are you really using the gcc 3.2 release or some prerelease
 >     or some modified version from a LINUX distributor instead?
 >     If the latter is true, you should try to install gcc 3.2.1
 >     from source.
 >     Please contact us, whether the problems still persist
 >     or vanished. In the latter case, please add "-v" to the
 >     command line options and provide us with the output so that we
 >     we can investigate the problem.
 > 
 > http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8691
 
 -- 
 http://www.informatik.tu-darmstadt.de/TI/Mitarbeiter/cludwig.html
 LiDIA-CA: http://www.informatik.tu-darmstadt.de/TI/Forschung/LiDIA-CA/
 

State-Changed-From-To: analyzed->closed
State-Changed-Why: Fixed for 3.3.