Re: [PATCH] Fortran DW_TAG_module/DW_TAG_imported_{declaration,module} support, improved DW_TAG_common_block support

[Jakub Jelinek <jakub at redhat dot com>]

On Thu, Aug 21, 2008 at 09:12:28PM +0200, Tobias Burnus wrote:
> > Attached is an updated patch (and interdiff from the last posted patch),
> > which leaves this compatibility cruft only to C++ and lets Fortran
> > define stuff inside of DW_TAG_module.
> 
> Thanks, the test case is now handled properly. However, I have found
> that now module variables outside the only clause "ONLY:" do not work.
> 
> Assume the following module:
> ----------------------
> module m
> implicit none
> real :: i,k,z
> end module m
> ----------------------
> 
> (a) Then the following works now:
>   use m, only: j=>i  ! and also: "only: i" w/o renaming
>   integer(8) :: i    ! or "only: j=>i, z=>i" work
>   j = 5.0
>   i = 7
> 
> (b) However, the following does not:
>   use m
>   i = 5.0
>   k = 10.0
>   z = 19.0
>   print *, i, k, z

That sounds like an idb bug, IMHO the Dwarf3 standard is quite clear
that for use m only DW_TAG_imported_module with DW_AT_import @DW_TAG_module
is enough, no DW_TAG_imported_declaration are needed.  It seems
ifort emits DW_TAG_imported_declaration for all use_assoc vars, not
DW_TAG_imported_module, so perhaps idb just handles what ifort implements.
Can you try say:
module modZ
  integer :: i, j, k, l
contains 
  subroutine init
    i = 1
    j = 2
    k = 3
    l = 4
  end subroutine init
end module modZ
  use modZ
  call init
  print *, i, j
end

with ifort/idb and see whether you can see in the debugger all i, j, k, l
values or just i and j inside of MAIN__?  The right behavior is to be able
to print them all.  I hope gdb will implement this correctly...

	Jakub