SIGSEGVs with g++3.2

SIGSEGVs with g++3.2

[Richard Guenther]

Hi!

I just tried todays gcc 3.2 branch and encountered SIGSEGVs in many
programs which look like

(gdb) run
Starting program: /net/bellatrix/home/rguenth/src/pooma-work/rhalk/Shock1d

Program received signal SIGSEGV, Segmentation fault.
0x080d4fa4 in std::basic_ios<char, std::char_traits<char>
>::_M_cache_facets(std::locale const&) (this=0x81402e4, __loc=@0x1) at
basic_ios.tcc:171
171           if (has_facet<__ctype_type>(__loc))
(gdb) bt
#0  0x080d4fa4 in std::basic_ios<char, std::char_traits<char>
>::_M_cache_facets(std::locale const&) (this=0x81402e4, __loc=@0x1) at
basic_ios.tcc:171
#1  0x080d4a73 in std::basic_ios<char, std::char_traits<char>
>::init(std::basic_streambuf<char, std::char_traits<char> >*)
(this=0x81402e4, __sb=0x1)
    at basic_ios.tcc:144
#2  0x08099a51 in std::ios_base::Init::_S_ios_create(bool) (__sync=72)
    at ostream:72
#3  0x08099e73 in Init (this=0x8140b24)
    at /home/rguenth/src/gcc/gcc3.2/libstdc++-v3/src/ios.cc:205
#4  0x080a4067 in __static_initialization_and_destruction_0
(__initialize_p=1,
    __priority=65535) at iostream:62
#5  0x080a408a in
_GLOBAL__I__ZThn8_NSt14basic_iostreamIwSt11char_traitsIwEED0Ev_home_rguenth_src_gcc_gcc3.2_libstdc___v3_src_io_inst.ccACBnhb
()
    at locale_facets.h:107
#6  0x08099845 in __do_global_ctors_aux ()
    at
/net/bellatrix/home/rguenth/src/pooma-work/r2/src/Utilities/Unique.cmpl.cpp:118
#7  0x0804a7fe in _init ()
#8  0x40154c24 in __libc_start_main () from /lib/libc.so.6
(gdb)

Does this look familiar to anybody?

bellatrix:~/src/pooma-work/rhalk$ ldd ./Shock1d
        libhdf5.so.0 => /home/rguenth/ix86/lib/libhdf5.so.0 (0x40017000)
        libm.so.6 => /lib/libm.so.6 (0x40119000)
        libc.so.6 => /lib/libc.so.6 (0x40138000)
        libz.so.1 => /lib/libz.so.1 (0x40254000)
        /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)

btw - this is on a SuSE 7.2 system. Oh - gcc 3.0.4 works without a
problem.

bellatrix:~/src/pooma-work$ g++ -v
Reading specs from
/home/rguenth/ix86/gcc3.2/lib/gcc-lib/i686-pc-linux-gnu/3.2/specs
Configured with: /home/rguenth/src/gcc/gcc3.2/configure
--prefix=/home/rguenth/ix86/gcc3.2 --enable-languages=c,c++,f77
--disable-shared --enable-threads
Thread model: posix
gcc version 3.2 20020802 (prerelease)

bellatrix:~/src/pooma-work$ ~/ix86/gcc3.0/bin/g++ -v
Reading specs from
/home/rguenth/ix86/gcc3.0/lib/gcc-lib/i686-pc-linux-gnu/3.0.5/specs
Configured with: /home/rguenth/src/gnu/gcc/configure
--prefix=/home/rguenth/ix86/gcc3.0 --enable-languages=c,c++,f77
Thread model: single
gcc version 3.0.5 20020403 (prerelease)

Thanks, Richard.

--
Richard Guenther <richard.guenther@uni-tuebingen.de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/
The GLAME Project: http://www.glame.de/

Re: SIGSEGVs with g++3.2

[Jakub Jelinek]

On Fri, Aug 02, 2002 at 03:44:38PM +0200, Richard Guenther wrote:
> Hi!
> 
> I just tried todays gcc 3.2 branch and encountered SIGSEGVs in many
> programs which look like
> 
> (gdb) run
> Starting program: /net/bellatrix/home/rguenth/src/pooma-work/rhalk/Shock1d
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x080d4fa4 in std::basic_ios<char, std::char_traits<char>
> >::_M_cache_facets(std::locale const&) (this=0x81402e4, __loc=@0x1) at
> basic_ios.tcc:171
> 171           if (has_facet<__ctype_type>(__loc))
> (gdb) bt
> #0  0x080d4fa4 in std::basic_ios<char, std::char_traits<char>
> >::_M_cache_facets(std::locale const&) (this=0x81402e4, __loc=@0x1) at
> basic_ios.tcc:171
> #1  0x080d4a73 in std::basic_ios<char, std::char_traits<char>
> >::init(std::basic_streambuf<char, std::char_traits<char> >*)
> (this=0x81402e4, __sb=0x1)
>     at basic_ios.tcc:144
> #2  0x08099a51 in std::ios_base::Init::_S_ios_create(bool) (__sync=72)
>     at ostream:72
> #3  0x08099e73 in Init (this=0x8140b24)
>     at /home/rguenth/src/gcc/gcc3.2/libstdc++-v3/src/ios.cc:205
> #4  0x080a4067 in __static_initialization_and_destruction_0
> (__initialize_p=1,
>     __priority=65535) at iostream:62
> #5  0x080a408a in
> _GLOBAL__I__ZThn8_NSt14basic_iostreamIwSt11char_traitsIwEED0Ev_home_rguenth_src_gcc_gcc3.2_libstdc___v3_src_io_inst.ccACBnhb
> ()
>     at locale_facets.h:107
> #6  0x08099845 in __do_global_ctors_aux ()
>     at
> /net/bellatrix/home/rguenth/src/pooma-work/r2/src/Utilities/Unique.cmpl.cpp:118
> #7  0x0804a7fe in _init ()
> #8  0x40154c24 in __libc_start_main () from /lib/libc.so.6
> (gdb)
> 
> Does this look familiar to anybody?

According to Gwenole Beauchesne this is the case with 3.1.0 and 3.1.1 too,
ie. IMHO not a show stopper for 3.2.

	Jakub

Re: SIGSEGVs with g++3.2

[Richard Guenther]

On Fri, 2 Aug 2002, Jakub Jelinek wrote:

> On Fri, Aug 02, 2002 at 03:44:38PM +0200, Richard Guenther wrote:
> > Hi!
> >
> > I just tried todays gcc 3.2 branch and encountered SIGSEGVs in many
> > programs which look like
> >
> > (gdb) run
> > Starting program: /net/bellatrix/home/rguenth/src/pooma-work/rhalk/Shock1d
> >
> > Program received signal SIGSEGV, Segmentation fault.
> > 0x080d4fa4 in std::basic_ios<char, std::char_traits<char>
> > >::_M_cache_facets(std::locale const&) (this=0x81402e4, __loc=@0x1) at
> > basic_ios.tcc:171
> > 171           if (has_facet<__ctype_type>(__loc))
> > (gdb) bt
> > #0  0x080d4fa4 in std::basic_ios<char, std::char_traits<char>
> > >::_M_cache_facets(std::locale const&) (this=0x81402e4, __loc=@0x1) at
> > basic_ios.tcc:171
> >
> > Does this look familiar to anybody?
>
> According to Gwenole Beauchesne this is the case with 3.1.0 and 3.1.1 too,
> ie. IMHO not a show stopper for 3.2.

Didnt try 3.1 - but I'm curious, is there any solution to this problem? Is
this perhaps a glibc version or binutils problem?

bellatrix:~$ rpm -q glibc
glibc-2.2.2-60
bellatrix:~$ rpm -q binutils
binutils-2.10.91.0.4-40

Richard.

--
Richard Guenther <richard.guenther@uni-tuebingen.de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/
The GLAME Project: http://www.glame.de/

Re: SIGSEGVs with g++3.2

[Joe Buck]

Richard Guenther writes:
> > I just tried todays gcc 3.2 branch and encountered SIGSEGVs in many
> > programs which look like
> > 
> > (gdb) run
> > Starting program: /net/bellatrix/home/rguenth/src/pooma-work/rhalk/Shock1d
> > 
> > Program received signal SIGSEGV, Segmentation fault.
> > 0x080d4fa4 in std::basic_ios<char, std::char_traits<char>
> > >::_M_cache_facets(std::locale const&) (this=0x81402e4, __loc=@0x1) at
> > basic_ios.tcc:171
> > 171           if (has_facet<__ctype_type>(__loc))
...
> > Does this look familiar to anybody?

Jakub Jelinek writes:
> According to Gwenole Beauchesne this is the case with 3.1.0 and 3.1.1 too,
> ie. IMHO not a show stopper for 3.2.

Do we have a GNATS case for this?  It's hard to tell how significant a
problem is if there is no test case, and since we are fixing at least 8
libstdc++ PRs in the 3.2 release, it may be appropriate to try to fix this
one as well if there is a straightforward solution.

Re: SIGSEGVs with g++3.2

[Gwenole Beauchesne]

On Fri, 2 Aug 2002, Jakub Jelinek wrote:

> According to Gwenole Beauchesne this is the case with 3.1.0 and 3.1.1 too,
> ie. IMHO not a show stopper for 3.2.

That was with HDF 5, a library written in C and I left out the C++
bindings in my test builds. Just curious, isn't David Edelson or Paolo
regularly regression testing gcc with Pooma? At least for 3.1-branch?

Bye,
Gwenole.

Re: SIGSEGVs with g++3.2

[Richard Guenther]

On Fri, 2 Aug 2002, Gwenole Beauchesne wrote:

> On Fri, 2 Aug 2002, Jakub Jelinek wrote:
> 
> > According to Gwenole Beauchesne this is the case with 3.1.0 and 3.1.1 too,
> > ie. IMHO not a show stopper for 3.2.
> 
> That was with HDF 5, a library written in C and I left out the C++
> bindings in my test builds. Just curious, isn't David Edelson or Paolo
> regularly regression testing gcc with Pooma? At least for 3.1-branch?

My problems were with pooma & hdf5, too (both recompiled with gcc3.2, hdf5
was compiled with plain C support only). Interesting to note is that
programs that do not use hdf5 work fine!?

Richard.

--
Richard Guenther <richard.guenther@uni-tuebingen.de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/
The GLAME Project: http://www.glame.de/

Re: SIGSEGVs with g++3.2

[Benjamin Kosnik]

> According to Gwenole Beauchesne this is the case with 3.1.0 and 3.1.1
> too, ie. IMHO not a show stopper for 3.2.

I'm not familiar with this, but the __loc argument looks suspect in the
reported stack trace.

Could one of you please send me a self-contained testcase (minimal) that
reproduces this issue? Also, enter it in GNATS?

thanks,
benjamin