Re: [rfc, frame] Move the corrupt frame checks earlier

[Daniel Jacobowitz <drow@false.org>]

On Tue, Aug 22, 2006 at 10:22:24PM +0200, Mark Kettenis wrote:
> I don't think this is a small difference.  I've certainly made use of
> it in the past, for both debugging stack corruption in a program being
> and tracking down problems with gdb's unwinders.  Even if you know
> where the saved registers are, you have to remember how big they are
> before you can print them.

I'm not sure what my opinion is on this one any more :-(

> Well, my feeling is that they are part of the backtrace and that it is
> a good thing to show that a backtrace it corrupt.

Well sure.  That's what motivated the previous (stop reason) patch:
notify the user that the backtrace is corrupt, and hint them where
to look for more information.  I could add a manual section near
"backtrace" which referenced the prefix for backtrace errors.
I think this is actually more useful than the extra frame, for a
bunch of reasons - like the "don't show too much information" we
just discussed.

For instance, here's another option that I like.  We could show saved
register values in info frame in addition to addresses.  I think that
would actually be convenient - I'd certainly use it, the first thing I
do after info frame is usually a bunch of x/ commands.  But I'm mildly
worried about what it will do to your SPARC64 example, at the same time.
We do paginate, though, and most people use the GDB console in
something with a scrollback buffer...

-- 
Daniel Jacobowitz
CodeSourcery