Re: Size of DW_FORM_ref_addr in DWARF 3 and the archer-jankratochvil-fedora13 branch

[Joost van der Sluis <joost@cnoc.nl>]

On Thu, 2010-04-29 at 19:39 +0200, Jan Kratochvil wrote:
> On Tue, 20 Apr 2010 17:38:53 +0200, Joost van der Sluis wrote:
> > Since Dwarf 3 the length of a DW_FORM_ref_addr entry is not dependent on
> > the pointer size of the target platform, but on the used Dwarf-format
> > (32 bit or 64 bit) for the current compilation section.
> 
> This is a HEAD patch
> 
> commit e3fa87d5f376db4523a844acc01199785d11b582
> Author: Tom Tromey <tromey@redhat.com>
> Date:   Thu Mar 4 22:01:09 2010 +0000
> 
>         * dwarf2read.c (skip_one_die) <DW_FORM_ref_addr>: Use offset size
>         in DWARF 3 and later.
>         (read_attribute_value) <DW_FORM_ref_addr>: Likewise.
> 
> checked-in after gdb_7_1-branch has been created but still before the 7.1
> release by:
> 
> commit c6c011eebb4006a89d4344864f560762f0aa45f3
> Author: Tom Tromey <tromey@redhat.com>
> Date:   Fri Mar 5 15:37:33 2010 +0000
> 
>         * dwarf2read.c (skip_one_die) <DW_FORM_ref_addr>: Use offset size
>         in DWARF 3 and later.
>         (read_attribute_value) <DW_FORM_ref_addr>: Likewise.
> 
> 
> > When I load a project with Dwarf-3 debug information with fsf-gdb head,
> > it works. But not with the archer-jankratochvil-fedora13 branch. That
> > version assumes that the length of a DW_FORM_ref_addr on 64 bit is
> > always 64 bits, which is not the case.
> 
> It works with FSF GDB HEAD, FSF GDB 7.1 and Fedora-13 gdb-7.1.
> It does not work with archer-jankratochvil-fedora13 as it is just an
> intermediate step for Fedora-13 gdb-7.1.
> 
> Tried now to merge archer-jankratochvil-fedora13 with gdb_7_1-branch but there
> is a problem it would need changes for the FSF-accepted using-directive
> patchset.  One cannot cheerry-pick the patch as diff against gdb_7_1-branch
> would then clash on that patch duplicity with FSF 7.1 release.
> 
> 
> > Is it planned to merge this fix from fsf-gdb to
> > archer-jankratochvil-fedora13? How is this normally handled?
> 
> archer-jankratochvil-fedora13 is now probably completely frozen (=dead) as
> I already add rather patches directly to the F-13 CVS.  One cannot change F-13
> much as it is far in its release cycle.
> 
> There should be some archer-jankratochvil-fedora14 branch created soon
> (=roughly in about two months).  I have to admit there is currently no fresh
> Archer-merged branch.

Thanks for the information. As long as this problem is not in Fedora-13
I'm happy.

I can work around this for now. Thanks for the information.

Regards,

Joost.