Could the new gud changes move to a branch?

Could the new gud changes move to a branch?

[John Wiegley]

I appreciate whoever is trying to make gud smarter, but unfortunately
is stumbling over its own feet far too often.  It has nearly rendered
debugging from within Emacs impossible.

It is much slower to get started.  It creates three windows, but
expects a certain arrangement, so that every once in a while when I
hit my breakpoint I am left with only the source window -- and must
hit C-x C-b to get back to the *gud* window.

I really like the new disassembly view, but at one point I got myself
into a situation where I simply could not switch my buffer away from
it.  It seemed to take over my minibuffer or something, and I had to
kill the process and restart it.

And lastly it seems that if the input/output window goes away (from
hitting C-x 1), and then I switch to the *gud* window and hit "up" or
"down", that it redisplays my frame so that only the source window and
the input/output window are displayed.  I have to keep switching to
the *gud* buffer.

Anyway, it has felt very complicated, and I used to be such a fan of
debugging with gdb under Emacs.  So could we go back to the old,
simple way until things settle down and the new code is better tested
by the author?  I think I will start using 21.3 for my debugging.

John

Could the new gud changes move to a branch?

[Nick Roberts]

 > I appreciate whoever is trying to make gud smarter, but unfortunately
 > is stumbling over its own feet far too often.  It has nearly rendered
 > debugging from within Emacs impossible.

Are things better or worse after the changes (2004-04-08) that I've just made?
I've removed some constraints for buffer display but maybe there is too much
freedom now.

 > It is much slower to get started.  It creates three windows, but
 > expects a certain arrangement, so that every once in a while when I
 > hit my breakpoint I am left with only the source window -- and must
 > hit C-x C-b to get back to the *gud* window.

I don't why its slower unless you are starting with a large disassembly view.
What are the three windows?

 > I really like the new disassembly view,...

I've included this as other debuggers have it but I don't use it myself>
What is it useful for?

 >                                      ...but at one point I got myself
 > into a situation where I simply could not switch my buffer away from
 > it.  It seemed to take over my minibuffer or something, and I had to
 > kill the process and restart it.
 > 
 > And lastly it seems that if the input/output window goes away (from
 > hitting C-x 1), and then I switch to the *gud* window and hit "up" or
 > "down", that it redisplays my frame so that only the source window and
 > the input/output window are displayed.  I have to keep switching to
 > the *gud* buffer.
 > 
 > Anyway, it has felt very complicated, and I used to be such a fan of
 > debugging with gdb under Emacs.  So could we go back to the old,
 > simple way until things settle down and the new code is better tested
 > by the author?  

The idea behind making "gdb --annotate=3" the default was to get feedback from
others.  Testing it all myself seems a rather sterile approach.

To get the old, simple way all you need to do is set gud-gdb-command-name to
"gdb --fullname" using customise. I've tried quite hard to keep the old
functionality there. Creating a branch for one elisp file doesn't seem
realistic. 

The (last?) remaining problem for GDB-UI seems to be buffer display and this
should be resolved by the time Emacs is released from trunk. If not, then it
will be quite straightforward to reset the default. In the meantime I would
like to keep it as it is to encourage its use in what, after all, is an
experimental version of Emacs.

 > I think I will start using 21.3 for my debugging.

For debugging emacs in CVS that seems a good idea, in any case.

Nick

Re: Could the new gud changes move to a branch?

[Stefan Monnier]

> I appreciate whoever is trying to make gud smarter, but unfortunately
> is stumbling over its own feet far too often.  It has nearly rendered
> debugging from within Emacs impossible.

Just start `gdb --fullname <foo>' instead of the --annotate=3 default.
It will then use good ol' gud code.  You can probably even change
the default to be --fullname, although I don't know offhand how to do it.


        Stefan

Re: Could the new gud changes move to a branch?

[Richard Stallman]

    Anyway, it has felt very complicated, and I used to be such a fan of
    debugging with gdb under Emacs.  So could we go back to the old,
    simple way until things settle down and the new code is better tested
    by the author?

Certainly not.  That would be giving up on it, not the way to make
progress in getting it to work better!

Perhaps we can add an option or options to turn off some of the new
features, those that take the most time to start up.