Re: Removing breakpoints/debugging stuff from CVS head

[Neil Jerram]

Mikael Djurfeldt wrote:
> 
> I think that, from the point of view of the user, debugging tools are
> very central.

Can I just check that you're not misunderstanding me?  I am not at all
suggesting the removal of any of the core infrastructure that was there
before I started playing with things, or of ice-9/debugger.scm as it has
existed for some time, or the ability to type (debug) and get a command
line debugger.  I am only talking of the Scheme and Emacs Lisp pieces
that I have been building experimentally on top of that infrastructure
(and in CVS 1.7.x HEAD only, which means that my stuff has never been
released yet, except in the 2 preview snapshots).

> So, it would be unfortunate if the core guile release
> tar archives (guile-X.X.X.tar.gz) did not contain the Guile debugger.
> It is also my guess that RMS would strongly agree on this.

Once the basic development is done, I completely agree that debugging
tools belong in the core distribution.  While the work is still in
progress, however, it's tactically preferable not to make what I can do
dependent on the core release status and schedule.

What this means in practice is two things.  Firstly, I have been able to
target my guile-debugging work so far at 1.6, which makes sense to me in
general because 1.6 is the most widely installed distribution, and in
particular because I have been using 1.6 for a project at work.  I would
not be able to do this work in 1.6 core CVS, because 1.6.x is stable and
we are not putting new features into it.

Secondly, it means that I can work on 1.8 support in my own time once
1.8 is released, as opposed to either (i) holding up the 1.8 release or
(ii) trying to complete what's needed more quickly than I'd like to meet
a release date.  (When the 1.8 support comes, there will probably be
associated 1.8 core bug fixes as well, but that's OK because they can go
in during the 1.8 stable series.)

Do you agree that this makes sense?

(These things might count less if we had more frequent releases ... but
we don't.)

> And, if you decide to make guile-core releases without the debugger,
> it might be a good idea if Marius or you consult with RMS first.

I find it difficult to see why this should be needed.  Do you still
think this, taking into account my comments above?

        Neil