Re: [PATCH] DWARF: Don't add nameless modules to partial symbol table.

[Bernhard Heckel <bernhard dot heckel at intel dot com>]

On 02/11/2016 17:41, Yao Qi wrote:
> On Mon, Oct 31, 2016 at 9:05 AM, Bernhard Heckel
> <bernhard.heckel@intel.com> wrote:
> > On 21/10/2016 17:37, Yao Qi wrote:
> > > Bernhard Heckel <bernhard.heckel@intel.com> writes:
> > >
> > > > A name for BLOCK DATA in Fortran is optional. If no
> > > > name has been assigned, GDB will crash during read-in of DWARF
> > > > when BLOCK DATA is represented via DW_TAG_module.
> > > > BLOCK DATA is used for one-time initialization of
> > > > non-pointer variables in named common blocks.
> > > I compiled your test case block-data.f, but there is no DW_TAG_module in
> > > debug info.  It has DW_TAG_common_block.  IOW, I run block-data.exp
> > > without your fix in dwarf2read.c, there is no crash in GDB.  My gfotran
> > > is 4.9.3.
> > It happen when compiled with Ifort. I attached a dump of the debug info,
> > line 80+ is important.
> I don't understand why BLOCK DATA is represented as a
> DW_TAG_module when program is compiled with ifort.  In DWARF 5,
> Modula-2 definition module and Fortran 90 module can be represented
> as DW_TAG_module.
>
> IMO, "BLOCK DATA" should be represented as a DW_TAG_subprogram,
> because it is a program unit initializes data that appears in common
> block.
From DWARF, I still believe DW_TAG_module could be used although it is 
not explicitly mentioned.

Nevertheless, Fortran specification say's this:
    24 subprogram (2.2) : A function subprogram or a subroutine 
subprogram. Note that in Fortran 77, a
    25 block data program unit was called a subprogram.

Sorry for the noise and thanks for your input


Intel Deutschland GmbH
Registered Address: Am Campeon 10-12, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de
Managing Directors: Christin Eisenschmid, Christian Lamprechter
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928