[Bug fortran/29139] New: Emit pointer attribute when printing type of Fortran pointers

["nils-christian.kempke at intel dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=29139

            Bug ID: 29139
           Summary: Emit pointer attribute when printing type of Fortran
                    pointers
           Product: gdb
           Version: HEAD
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P2
         Component: fortran
          Assignee: unassigned at sourceware dot org
          Reporter: nils-christian.kempke at intel dot com
  Target Milestone: ---

Currently, printing the type of a Fortran pointer (so an object with the
pointer attribute) and printing the type of an object of the same type without
the pointer attribute does not show any difference within GDB.

Consider the Fortran program

==============================================
program MAIN
  Integer, target :: var
  Integer, pointer :: pvar
  pvar => var
  var = 1

  print *, var, pvar
end program
==============================================

compiled with

==============================================
>> gfortran --version
GNU Fortran (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
==============================================


Running a GDB session and breaking at line 7 (the print statement) we get

==============================================
7         print *, var, pvar
(gdb) p pvar
$1 = (1, 1, 1)
(gdb) ptype pvar
type = integer(kind=4) (3)
(gdb) p var
$2 = (1, 1, 1)
(gdb) ptype var
type = integer(kind=4) (3)
(gdb) p associated(var)
ASSOCIATED can only be applied to pointers
(gdb) p associated(pvar)
$3 = .TRUE.
==============================================

The only way to find out if something is a pointer or not is via the Fortran
intrinsics. I think this could be improved and one could add the 'pointer'
attribute also to the type-print of pvar. The pointer attributed can be derived
from the emitted DWARF and the DW_AT_associated for the respective DIEs (as is
already done for the ASSOCIATED intrinsic)

In this case gfortran generates DWARF like:

==============================================
...
 <2><ab>: Abbrev Number: 8 (DW_TAG_variable)
    <ac>   DW_AT_name        : (indirect string, offset: 0x56): pvar
    <b0>   DW_AT_decl_file   : 1
    <b1>   DW_AT_decl_line   : 3
    <b2>   DW_AT_type        : <0xca>
    <b6>   DW_AT_location    : 3 byte block: 91 80 7f   (DW_OP_fbreg: -128)
...
 <1><ca>: Abbrev Number: 10 (DW_TAG_array_type)
    <cb>   DW_AT_data_location: 2 byte block: 97 6     
(DW_OP_push_object_address; DW_OP_deref)
    <ce>   DW_AT_associated  : 4 byte block: 97 6 30 2e        
(DW_OP_push_object_address; DW_OP_deref; DW_OP_lit0; DW_OP_ne)
    <d3>   DW_AT_type        : <0x74>
    <d7>   DW_AT_sibling     : <0xf1>
...
==============================================

with the additional DW_AT_associated attribute.

Something like

==============================================
(gdb) p pvar
type = interger(kind=4) (3) pointer
==============================================

would give more insight here.

-- 
You are receiving this mail because:
You are on the CC list for the bug.