help: gdb and c++ strings

help: gdb and c++ strings

[Murad Nayal]

Hello,

I am wondering if anyone here can help me with this. I am porting code
to linux and trying to use gdb to do the debugging. unfortunately this
is complicated by the fact that gdb does not print c++ string variables.

p s
$1 = (
    const basic_string<char,std::char_traits<char>,std::allocator<char>
> &) @0xbffff3d0: {static npos = Cannot access memory at address 0x8f4b974

I have come upon an older post to bug-gdb list that mentions this
problem
http://sources.redhat.com/ml/bug-gdb/2000-09/msg00012.html

in response Daniel Berlin suggested that this problem might be related
to the stabs format and that using dwarf-2 format (i.e. g++ -gdwarf-2)
would solve it. however it was mentioned back then that for large
programs this problem persists even when compiling with the -gdwarf-2
flag. this is where I am at right now. I have recompiled gdb from
sources (5.2.1) and recompiled most of the libraries I am using
(including the objects I am trying to debug) using the -gdwarf-2 flag. I
am on a pentium 3 platform using g++ 3.2.1, also compiled from sources.
I am not sure what to try next. can you help?

thanks
Murad Nayal

Re: help: gdb and c++ strings

[Daniel Jacobowitz]

On Thu, Dec 05, 2002 at 11:02:24PM -0500, Murad Nayal wrote:
> 
> 
> Hello,
> 
> I am wondering if anyone here can help me with this. I am porting code
> to linux and trying to use gdb to do the debugging. unfortunately this
> is complicated by the fact that gdb does not print c++ string variables.
> 
> p s
> $1 = (
>     const basic_string<char,std::char_traits<char>,std::allocator<char>
> > &) @0xbffff3d0: {static npos = Cannot access memory at address 0x8f4b974
> 
> I have come upon an older post to bug-gdb list that mentions this
> problem
> http://sources.redhat.com/ml/bug-gdb/2000-09/msg00012.html
> 
> in response Daniel Berlin suggested that this problem might be related
> to the stabs format and that using dwarf-2 format (i.e. g++ -gdwarf-2)
> would solve it. however it was mentioned back then that for large
> programs this problem persists even when compiling with the -gdwarf-2
> flag. this is where I am at right now. I have recompiled gdb from
> sources (5.2.1) and recompiled most of the libraries I am using
> (including the objects I am trying to debug) using the -gdwarf-2 flag. I
> am on a pentium 3 platform using g++ 3.2.1, also compiled from sources.
> I am not sure what to try next. can you help?

You need to get a newer snapshot of GDB.  This bug was fixed only a
couple of weeks ago, IIRC.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: help: gdb and c++ strings

[Murad Nayal]

many thanks. this will really help.

is there a packaged snapshot available somewhere or do I need to get the
sources from cvs?

best

Daniel Jacobowitz wrote:
> 
> On Thu, Dec 05, 2002 at 11:02:24PM -0500, Murad Nayal wrote:
> >
> >
> > Hello,
> >
> > I am wondering if anyone here can help me with this. I am porting code
> > to linux and trying to use gdb to do the debugging. unfortunately this
> > is complicated by the fact that gdb does not print c++ string variables.
> >
> > p s
> > $1 = (
> >     const basic_string<char,std::char_traits<char>,std::allocator<char>
> > > &) @0xbffff3d0: {static npos = Cannot access memory at address 0x8f4b974
> >
> > I have come upon an older post to bug-gdb list that mentions this
> > problem
> > http://sources.redhat.com/ml/bug-gdb/2000-09/msg00012.html
> >
> > in response Daniel Berlin suggested that this problem might be related
> > to the stabs format and that using dwarf-2 format (i.e. g++ -gdwarf-2)
> > would solve it. however it was mentioned back then that for large
> > programs this problem persists even when compiling with the -gdwarf-2
> > flag. this is where I am at right now. I have recompiled gdb from
> > sources (5.2.1) and recompiled most of the libraries I am using
> > (including the objects I am trying to debug) using the -gdwarf-2 flag. I
> > am on a pentium 3 platform using g++ 3.2.1, also compiled from sources.
> > I am not sure what to try next. can you help?
> 
> You need to get a newer snapshot of GDB.  This bug was fixed only a
> couple of weeks ago, IIRC.
>

Re: help: gdb and c++ strings

[Murad Nayal]

Hello Michael

the 5.2 snapshot worked. thank you all very much for the help. I haven't
tried the 5.3 branch snapshot (it took forever to download from cvs,
which I started before getting Michael's message).

best


Michael Elizabeth Chastain wrote:
> 
> Hello Murad,
> 
> There are two snapshots available:
> 
>   ftp://sources.redhat.com/pub/gdb/snapshots/branch/gdb-5.2.91.tar.bz2
>   ftp://sources.redhat.com/pub/gdb/snapshots/current/insight+dejagnu-weekly-CVS-20021206.tar.bz2
> 
> The first snapshot is from the 5.3 branch and will become gdb 5.3 soon
> if everything goes well.  I am regression testing this right now versus
> 5.2.1 and there are no big obvious problems yet.
> 
> The second snapshot is from the CVS head.  It's more likely to have the
> bug fix that Daniel J is referring to.  It's also possible that it has
> build problems because the configuration+build machinery is in active
> development right now.  If it builds, though, it is also in good shape
> for gcc v3 dwarf-2 debugging.
> 
> You are on the right track with gcc 3.2.1 and -gdwarf-2.
> 
> I recommend you try the "branch" snapshot first; it is much more stable.
> If it is not featureful enough (does not fix your problems) then try
> the "current" snapshot.
> 
> Hope this helps,
> 
> Michael C
> 
> P. S. I see you are at columbia.edu.  I am on Staten Island.  Isn't
> the city beautiful with today's snow?

Re: help: gdb and c++ strings

[Michael Elizabeth Chastain]

Hello Murad,

There are two snapshots available:

  ftp://sources.redhat.com/pub/gdb/snapshots/branch/gdb-5.2.91.tar.bz2
  ftp://sources.redhat.com/pub/gdb/snapshots/current/insight+dejagnu-weekly-CVS-20021206.tar.bz2

The first snapshot is from the 5.3 branch and will become gdb 5.3 soon
if everything goes well.  I am regression testing this right now versus
5.2.1 and there are no big obvious problems yet.

The second snapshot is from the CVS head.  It's more likely to have the
bug fix that Daniel J is referring to.  It's also possible that it has
build problems because the configuration+build machinery is in active
development right now.  If it builds, though, it is also in good shape
for gcc v3 dwarf-2 debugging.

You are on the right track with gcc 3.2.1 and -gdwarf-2.

I recommend you try the "branch" snapshot first; it is much more stable.
If it is not featureful enough (does not fix your problems) then try
the "current" snapshot.

Hope this helps,

Michael C

P. S. I see you are at columbia.edu.  I am on Staten Island.  Isn't
the city beautiful with today's snow?