Re: [PATCH] implements MI "-file-list-exec-sections" (updated)

[Andrew Cagney <cagney at gnu dot org>]

> On Sat, May 01, 2004 at 01:07:44PM +0100, Nick Roberts wrote:
>
> > >  > I think everyone agrees, changing the CLI would be the worst possible
> > >  > thing GDB could do.
> > >
> > > I'm not sure if thats true, I was just talking about timing. Its probably
> > > not reasonable or even realistic to insist that CLI output never changes.
>
>
> Well, over time, I can see the CLI changing. However, I would expect
> that to happen after at least one open source project implemented the MI
> front end  (Don't say eclipse, they use a hybrid approach) and the MI
> API was considered stable and capable of doing everything that could
> be done from the CLI.

Regardless of MI, the CLI is going to evolve.  This is because the 
current implementation is limiting GDB's functionality.

A few simple examples:

- signal backtraces.  They currently look something like:
	0x1234 in foo
	<signal handler called>
	0x456 in bar
but are about to be changed to:
	0x1234 in foo
	0x785 in <signal trampoline>
	0x456 in bar
This comes hand-in-hand with the raised expectation that operations such 
as "stepi"ing through a trampoline work (previously it was luck, 
typically bad :-).  For the user this is all good, but I bet there's a 
CLI parsing gui out there that breaks - oops.

- N:M breakpoints
Again, the user will easily adapt.

Andrew