RE: [PATCH 2/2] gdb: Resolve dynamic target types of pointers.

["Kempke, Nils-Christian" <nils-christian.kempke@intel.com>]

> -----Original Message-----
> From: Tom Tromey <tom@tromey.com>
> Sent: Friday, April 15, 2022 6:14 PM
> To: Kempke, Nils-Christian via Gdb-patches <gdb-patches@sourceware.org>
> Cc: Tom Tromey <tom@tromey.com>; Kempke, Nils-Christian <nils-
> christian.kempke@intel.com>
> Subject: Re: [PATCH 2/2] gdb: Resolve dynamic target types of pointers.
> 
> > Thanks for the feedback. Sorry that it took me so long to reply - the mail
> > somehow got lost..
> 
> No need to apologize!  I'm constantly months behind and/or dropping email.
> 
> >> > +	  if (type->code () == TYPE_CODE_PTR && is_dynamic_type (type))
> >> > +	    {
> >> > +	      CORE_ADDR addr;
> >> > +	      if (nullptr != TYPE_DATA_LOCATION (TYPE_TARGET_TYPE (type)))
> >> > +		addr = value_address (val);
> >> > +	      else
> >> > +		addr = value_as_address (val);
> >>
> >> This seems weird to me.  When does value_as_address fail?
> 
> > This if else has been introduced because of the way ifort/icc describe
> pointers
> > vs. gcc/gfortran and icx/ifx.
> 
> > <1><AAA> DW_TAG_variable
> >     <> DW_AT_type: <BBB>
> >     <> DW_AT_location: (location of the pointer)
> > <1><BBB> DW_TAG_pointer_type
> >     DW_AT_type: <CCC>
> > <1><CCC> DW_TAG_array_type
> >     <> DW_AT_type: <DDD>
> >     <> DW_AT_data_location: (location of the array)
> > <2><EEE> DW_TAG_subrange_type
> 
> > Fortran:
> 
> > <1><AAA> DW_TAG_variable
> >     <>   DW_AT_type: <BBB>
> >     <>   DW_AT_location: (location of the pointer)
> > <1><BBB> DW_TAG_array_type
> >     <> DW_AT_type: <CCC>
> >     <> DW_AT_data_location: (location of the array)
> > <2><DDD> DW_TAG_subrange_type
> 
> > When printing this pointer to array in gfortran/ifx the pointer print function
> is
> > not actually called - since the DWARF does not emit a pointer here.
> Instead,
> > the normal array print is invoked which will use the DW_AT_data_location.
> > So we do not run into this same problem
> 
> I still do not understand, sorry.
> 
> For me this looks like two different types.  In one situation, the type
> is a pointer to an array.  In the second, it's just an array.  So, I
> would expect them to print differently.
> 
> Are you saying these should print identically?  If so, that's fine --
> but then it seems like something to do in the Fortran code, not in
> c-valprint.c.

Yes, that is true. So, in Fortran pointers are just aliases for the underlying objects
 
> Maybe in the code the check for TYPE_DATA_LOCATION is a proxy for
> checking that the pointer is a pointer to a Fortran array?  I guess I
> could understand that, but it doesn't seem like a good approach, because
> what if some C compiler starts emitting this?
> 
> >> I'm curious why the code added here is not also needed in f-valprint.c.
> 
> > The reason here is simply that Fortran does not have this implemented for
> > Itself.  Instead the fortran f_langugage inherits from language_defn (the all
> > languages base class) which simply delegates the value-print to
> > c_value_print / so changing it there suffices.
> 
> If Fortran-specific printing is needed, this will have to change.
> 
> Tom
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