This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Input stream crash with gcc 3.1/cygwin (stack trace incl.)


Hello there

I'm not getting much response regarding my problem on the gcc-help or cygwin
mailing lists so I apologize for being persistent and trying here.

I have taken the very latest 3.1 branch of gcc and compiled it successfully
on the cygwin (1.3.9) platform.  All seems fine until any kind of iostream
reading takes place - even "stringstream >> frog;" where frog is anything.
The program simply segment faults on me.

Does anybody have any ideas on how to track this problem down?  I have
included as best a stack trace I could do under the limited environment
available to me. (c++filt doesn't seem to be working very well for
demangling these addresses)

Thanks in advance

---------------------------------
Q-Games, Dylan Cuthbert.
http://www.q-games.com

> ----- Original Message -----
> From: "Dylan Cuthbert" <dylan@nospamq-games.com>
> Newsgroups: gmane.os.cygwin
> Sent: Tuesday, March 19, 2002 7:16 PM
> Subject: Re: RE: input stream crash with gcc 3.1 (stack trace)
>
>
> > After compiling the stdlibs with -O0 -g I tracked down the crash to
> address:
> > 0x004bdd7d
> >
> > looking in my map file I see this at that address:
> >
> >  .text$_ZNKSt5ctypeIcE2isEcc
> >                 0x004bdd60       0x3c
> > /usr/local/lib/libstdc++.a(misc-inst.o)
> >                 0x004bdd60                __ZNKSt5ctypeIcE2isEcc
> >
> > here is the full de-mapped stack dump:  note: gdb 5.1 is playing up with
> the
> > new gcc 3.1 output format (maybe I need to grab a different branch of
> gdb?)
> > so I can't simply do a symbolic stack trace (it seg faults on me).
> >
> > 0x004bdd7d:
> >
> >  .text$_ZNKSt5ctypeIcE2isEcc
> >                 0x004bdd60       0x3c
> > /usr/local/lib/libstdc++.a(misc-inst.o)
> >                 0x004bdd60                __ZNKSt5ctypeIcE2isEcc
> >
> >
> > 0x004c7c6e:
> >
> > .text$_ZNSi6sentryC1ERSib
> >                 0x004c7b6c      0x184
> > /usr/local/lib/libstdc++.a(misc-inst.o)
> >                 0x004c7b6c                __ZNSi6sentryC1ERSib
> >  .text$_ZNSi6sentryC2ERSib
> >
> > 0x004f5b5f:
> >
> >
>
.text$_ZStrsIcSt11char_traitsIcESaIcEERSt13basic_istreamIT_T0_ES7_RSbIS4_S5_
> > T1_E
> >                 0x004f5b38      0x1f4
> > /usr/local/lib/libstdc++.a(misc-inst.o)
> >                 0x004f5b38
> >
>
__ZStrsIcSt11char_traitsIcESaIcEERSt13basic_istreamIT_T0_ES7_RSbIS4_S5_T1_E
> >
> >
> >
> > Anyone got any ideas??
> >
> > ---------------------------------
> > Q-Games, Dylan Cuthbert.
> > http://www.q-games.com
> >
> > "Dylan Cuthbert" <dylan@nospamq-games.com> wrote in message
> > news:a76gn8$g39$1@main.gmane.org...
> > > Hmm.. I noticed some talk on a mailing list somewhere about problems
> with
> > > locales... could it be to do with input streams trying to look up
locale
> > > info and getting null ptrs as a result?
> > >
> > > I'll try compiling libstdc++-v3 with -g and -O0 and see how far I can
> > get -
> > > does gdb 5.1 work ok with gcc 3.1 output?
> > >
> > > Regards
> > >
> > > ---------------------------------
> > > Q-Games, Dylan Cuthbert.
> > > http://www.q-games.com
> > >
> > > "Billinghurst, David (CRTS)" <David.Billinghurst@riotinto.com> wrote
in
> > > message
> > >
> news:FAC87D7C874EAB46A847604DA4FD5A640346C9@crtsmail.corp.riotinto.org...
> > > I just tried this too. Same result.  I have been chasing (without
> success)
> > > some libstdc++-v3 testsuite failures that are similar.  That discusion
> is
> > on
> > > the libstdc++ mailing list.   The advice was to try recompiling
> > libtsdc++-v3
> > > with CXXFLAGS="-g -O0" and use the debugger.  This got me a little
> further
> > > into the code.
> > >
> > > -----Original Message-----
> > > From: Dylan Cuthbert [mailto:dylan@q-games.com]
> > > Sent: Tuesday, 19 March 2002 2:11
> > > Subject: Re: RE: input stream crash with gcc 3.1
> > >
> > >
> > > I tried this #define (in the newlib bits directory - cygwin uses
newlib
> > > right?), reconfigured and recompiled gcc 3.1, but the same problem
> > occurs -
> > > segmentation fault on the simplest bit of code that uses cin or
ifstream
> > or
> > > stringstream.  ( cout etc. works fine by the way).
> > >
> > > I'll repeat the code snippet for clarification:
> > >
> > > main()
> > > {
> > > std::string frog;
> > > std::cin >> frog;
> > > assert( 0 );
> > > }
> > >
> > > (it doesn't get to the assert and segment faults)
> > >
> > > Any other ideas? :-/
> > >
> > > ---------------------------------
> > > Q-Games, Dylan Cuthbert.
> > > http://www.q-games.com/personal/utils
> > >
> > > ----- Original Message -----
> > > From: "Danny Smith" <danny_r_smith_2001@yahoo.co.nz>
> > > Sent: Tuesday, March 19, 2002 8:08 AM
> > > Subject: Fwd: RE: input stream crash with gcc 3.1
> > >
> > >
> > > > > From: Danny Smith <danny_r_smith_2001@yahoo.co.nz>
> > > > > Subject: RE: input stream crash with gcc 3.1
> > > > > To: dylan@nospamq-games.com
> > > > > CC: cygwin@cygwin.com
> > > > >
> > > > > > I have recently compiled the latest 3.1 branch (grabbed directly
> via
> > > > > cvs)
> > > > > > for cygwin (1.3.9) and am having problems when using anything
> > related
> > > > > at
> > > > > > all
> > > > > > to input streams.
> > > > >
> > > > > See this:
> > > > > http://gcc.gnu.org/ml/gcc-patches/2001-06/msg00841.html
> > > > > then try rebuilding libstdc++ with
> > > > >
> > > > > #define _GLIBCPP_AVOID_FSEEK
> > > > > added to /libstdc++-v3/config/os/newlib/bits/os_defines.h
> > > > >
> > > > > Danny

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]