"no longer active" registers

"no longer active" registers

[Jeremy Bennett]

I've been using GDB to connect to a remote target via RSP.

After loading a program, an attempt to set $pc will give the message:

        "Value being assigned to is no longer active."

I believe this is a known issue
(http://sourceware.org/bugzilla/show_bug.cgi?id=9342), and there has
been some discussion of why it occurs
(http://sourceware.org/ml/gdb-patches/2006-05/msg00017.html).

I've discovered that the Fedora project has a small patch to valops.c by
Jan Kratochvil, which seems to cure this problem:

        http://cvs.fedoraproject.org/viewvc/devel/gdb/gdb-6.8-bz436037-reg-no-longer-active.patch?revision=1.1&view=markup

I notice that this patch is not part of the main GDB CVS tree. Are there
plans to incorporate this patch into the main CVS tree, or are there
side-effects that mean it is not generally applicable?

Advice on this subject much appreciated.

Thanks,


Jeremy

-- 
Tel:      +44 (1202) 416955
Cell:     +44 (7970) 676050
SkypeID: jeremybennett
Email:   jeremy.bennett@embecosm.com
Web:     www.embecosm.com

Re: "no longer active" registers

[Jan Kratochvil]

On Mon, 13 Apr 2009 10:17:46 +0200, Jeremy Bennett wrote:
> I've discovered that the Fedora project has a small patch to valops.c by
> Jan Kratochvil, which seems to cure this problem:
> 
>         http://cvs.fedoraproject.org/viewvc/devel/gdb/gdb-6.8-bz436037-reg-no-longer-active.patch?revision=1.1&view=markup

Just a correction of the credit, this patch is by Denys Vlasenko.


> I notice that this patch is not part of the main GDB CVS tree. Are there
> plans to incorporate this patch into the main CVS tree, or are there
> side-effects that mean it is not generally applicable?

While I could not find a regression by this patch it is just a fixup of one of
the many places affected by the ambiguous meaning of a NULL frame.

Probably the right fix is to have non-NULL frame value for the unbacktraceable
frame.


Regards,
Jan

Re: "no longer active" registers

[Daniel Jacobowitz]

On Mon, Apr 13, 2009 at 02:40:21PM +0200, Jan Kratochvil wrote:
> > I notice that this patch is not part of the main GDB CVS tree. Are there
> > plans to incorporate this patch into the main CVS tree, or are there
> > side-effects that mean it is not generally applicable?
> 
> While I could not find a regression by this patch it is just a fixup of one of
> the many places affected by the ambiguous meaning of a NULL frame.

I think it's the opposite workaround that should be used.  The NULL ID
is usually the outermost frame, not the innermost.  I don't have a
link handy but I've posted another workaround for this to the list in
the past.

> Probably the right fix is to have non-NULL frame value for the unbacktraceable
> frame.

Yes, I agree.

-- 
Daniel Jacobowitz
CodeSourcery