RE: C++ nested class bug/questions

["Willgerodt, Felix" <felix.willgerodt@intel.com>]

> -----Original Message-----
> From: Simon Marchi <simon.marchi@polymtl.ca>
> Sent: Dienstag, 22. Juni 2021 21:01
> To: Willgerodt, Felix <felix.willgerodt@intel.com>; gdb@sourceware.org
> Subject: Re: C++ nested class bug/questions
> 
> On 2021-06-22 7:27 a.m., Willgerodt, Felix via Gdb wrote:
> > Hello everyone,
> >
> > I was debugging a program similar to the one attached and noticed a
> > couple of unexpected things happening. I compiled it with ”-g -O0” and
> > g++ 11.1.0 (clang showed similar behaviour) and saw this:
> >
> > (gdb) bt
> > #0  DF::vvv (this=0x7fffffffd120) at foo.cc:11
> > #1  0x0000555555555180 in foo () at foo.cc:16
> > #2  0x00005555555551a7 in main () at foo.cc:20
> > (gdb) ptype abcd
> > type = class DF::FFFF {
> >   private:
> >     int ffff;
> > }
> > (gdb) ptype FFFF
> > No symbol "FFFF" in current context.
> > (gdb) ptype DF::
> > FFFF   vvv()
> > (gdb) ptype DF::FFFF
> > There is no field named FFFF
> > (gdb) ptype 'DF::FFFF'
> > type = class DF::FFFF {
> >   private:
> >     int ffff;
> > }
> > (gdb)
> >
> > (The fourth command shows tab completion.)
> >
> > The problems I see are:
> > 1) I think it shouldn’t be DF::FFFF.
> > 2) 'ptype FFFF’ should just work as it is in the current scope.
> > 3) There shouldn’t be a need to quote DF:FFFF, right? Although that might
> just be a side-effect here.
> > 4) The fact that a suggestion from tab completion doesn’t work is annoying.
> This is probably also be a side-effect.
> >
> > Before debugging this further, I wanted to see if there are any
> comments/suggestions.
> > Am I even right about point 1? I couldn’t really find anything
> > definitive in my favourite cpp reference on this type of nesting. g++
> > and clang output FFFF as foo()::DF::vvv()::FFFF as far as I can see.
> >
> > I don’t see any obvious problems in the dwarf that is produced.
> >
> > Thanks,
> > Felix
> 
> I'm not a reference on this topic, but at first glance I agree with you.
> 
> Simon

I looked into this a bit today, and this patch seems to solve it for me:


diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index 57bcfb713db..fbff05ac153 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -22589,7 +22589,7 @@ determine_prefix (struct die_info *die, struct dwarf2_cu *cu)
                && (dwarf2_name (parent, cu) != NULL))
              return dwarf2_name (parent, cu);
          }
-       return determine_prefix (parent, cu);
+       return "";
       case DW_TAG_enumeration_type:
        parent_type = read_type_die (parent, cu);
        if (parent_type->is_declared_class ())


With this patch I see no regressions in gdb.cp and gdb.fortran.

But I am wondering why the current code even tries to go to the parent of the
subprogram (which is the "first level" parent) to get a prefix? I don't really see a use case for it.
But I am certainly not an expert in all languages that GDB understands and fear that I am
just missing the point. Does anyone have an idea?

I will try to test this more thoroughly and write a regression test in the meantime.

Regards,
Felix
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928