gdb namespaces/wrapper bug?

gdb namespaces/wrapper bug?

[Klaas Gadeyne]

Hi,

I encountered a strange problem while debugging and I'm not sure wether
the problem is coming from gcc-3.x or gdb?

Consider the following code, consisting of 3 files
=========================================================================
=========================================================================
[klaas@c014 /tmp]$ cat wrapper_object.h
#include <vector>

namespace BASE
{
  class Foo
  {
  public:
    Foo(){};
    virtual ~Foo(){};
    virtual void Method();
  };

}
=========================================================================
[klaas@c014 /tmp]$ cat wrapper_object.cpp
#include "wrapper_object.h"

namespace BASE
{
  void
  Foo::Method(){};
}
==========================================================================
[klaas@c014 /tmp]$ cat wrapper.cpp
#include "wrapper_object.h"

namespace WRAPPED
{
  class Foo : public BASE::Foo
  {
  public:
    Foo() : BASE::Foo(){};
    virtual ~Foo(){};
  };
}

using namespace WRAPPED;

int main()
{
  Foo mywrapped;
  return 0;
}
==========================================================================
==========================================================================

This compiles and runs just fine (compile wrapper_object.cpp, compile
wrapper.cpp linking with wrapper_object.o), but when debugging the code
with gdb (I only tried with both 5.3-debian and 6.0-debian (the latest
from testing and unstable)) and trying to display the mywrapped variable,
gdb enters in a infinite loop:

(gdb) display mywrapped
1: mywrapped = {<Foo> = {<Foo> = ....

and finally segfaults.

Used compilers (all debian packages):

[klaas@c014 /tmp]$ g++ --version
g++ (GCC) 3.3.3 20040110 (prerelease) (Debian)

I also tried with 3.3.2, but that gives the same result.

The problem does not exist when:
- The code is compiled with g++ 2.95.4
- All code is inlined in 1 file (yes, that's why there are 3 files above :)
- I omit the member function method and put the constructor implementation
  in the cpp file.
- When Foo in namespace WRAPPED is renamed into Boo

Is this a gdb problem? or is is related to gcc 3.x?  Or am I missing
something important here?

thanks,

klaas

Re: gdb namespaces/wrapper bug?

[Daniel Jacobowitz]

On Thu, Jan 15, 2004 at 01:37:45PM +0100, Klaas Gadeyne wrote:
> namespace BASE
> {
>   class Foo


> namespace WRAPPED
> {
>   class Foo : public BASE::Foo


> This compiles and runs just fine (compile wrapper_object.cpp, compile
> wrapper.cpp linking with wrapper_object.o), but when debugging the code
> with gdb (I only tried with both 5.3-debian and 6.0-debian (the latest
> from testing and unstable)) and trying to display the mywrapped variable,
> gdb enters in a infinite loop:
> 
> (gdb) display mywrapped
> 1: mywrapped = {<Foo> = {<Foo> = ....
> 
> and finally segfaults.

Could you try CVS gdb?  I think the segfault has been fixed.  I may be
mistaken about that, though.  This is related to
lookup_transparent_type, which David has been doing some work on (and
has more pending).

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: gdb namespaces/wrapper bug?

[Elena Zannoni]

Daniel Jacobowitz writes:
 > 
 > Could you try CVS gdb?  I think the segfault has been fixed.  I may be
 > mistaken about that, though.  This is related to
 > lookup_transparent_type, which David has been doing some work on (and
 > has more pending).

I actually was just looking at a similar bug against RH gdb, which I just
verified with current CVS:

Do 
b main
run 
p foo

with this, and you get the infinite loop + segfault:


struct derived;

struct base
{
  static derived bar;
};

struct derived : base
{
};

derived base::bar;

base foo;

int main()
{
  return 0;
}



the problem seems to be cp_print_static_field. If you set print
static-member off of course it all works. (gcc 3.2.2)

elena

Re: gdb namespaces/wrapper bug?

[Klaas Gadeyne]

> Could you try CVS gdb?  I think the segfault has been fixed.  I may be
> mistaken about that, though.  This is related to
> lookup_transparent_type, which David has been doing some work on (and
> has more pending).

Hi Daniel,

Indeed, the infinite loop/segfault is gone with CVS gdb, however, the
behaviour is not yet as desired :)

====================================================================
[klaas@c014 /tmp]$ /usr/local/bin/gdb ./a.out
GNU gdb 2004-01-15-cvs
This GDB was configured as "i686-pc-linux-gnu"...Using host libthread_db
library "/lib/libthread_db.so.1".

(gdb) break wrapper.cpp:17
Breakpoint 1 at 0x80484f4: file wrapper.cpp, line 17.
(gdb) run
Starting program: /tmp/a.out

Breakpoint 1, main () at wrapper.cpp:17
17        Foo mywrapped;
(gdb) next
18        return 0;
(gdb) display mywrapped
1: mywrapped = {<> = {<No data fields>}, <No data fields>}
======================================================================

So mywrapped is now reported to be empty, where I expected to see
something like

1: mywrapped = {<Foo> = {<No data fields>}, <No data fields>}

regards,

klaas