Completion in GUD buffer

Completion in GUD buffer

[Nick Roberts]

A long time back completion in the GUD buffer changed.  Previously if you
did something like:

(gdb) b <TAB>

to complete on all possible locations, when there were too many to fit in
the completion window, you could click on it to explore down and *up*.
(I think you can only go down from outside it).

Now if you click on the completion window you either get an error message:

  Wrong type argument: window-live-p, #<window 7>

or the window simply disappears.

I've traced this change to:

  revision 1.294
  date: 2004-04-14 18:57:45 +0000;  author: monnier;  state: Exp;  lines: +185 -195
  Doc fixes.
  (comint-last-input-start, comint-last-input-end)
  (comint-last-output-start): Give them values.
  (comint-arguments): Avoid warning.
  (comint-skip-prompt): Use line-end-position.
  (comint-backward-matching-input): Rename arg to agree with doc.
  (comint-extract-string): Use syntax-ppss.
  (comint-dynamic-simple-complete): Delete useless list construction.
  (comint-redirect-subvert-readonly): New.
  (comint-redirect-preoutput-filter): Use it.

but there are 91 difference between this file and the previous one.

Before I wade through them, I'd like to ask if people agree that this change
in behaviour should be reverted?


-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: Completion in GUD buffer

[Stefan Monnier]

> A long time back completion in the GUD buffer changed.  Previously if you
> did something like:

> (gdb) b <TAB>

> to complete on all possible locations, when there were too many to fit in
> the completion window, you could click on it to explore down and *up*.
> (I think you can only go down from outside it).

> Now if you click on the completion window you either get an error message:

>   Wrong type argument: window-live-p, #<window 7>

> or the window simply disappears.

> I've traced this change to:

>   revision 1.294
>   date: 2004-04-14 18:57:45 +0000;  author: monnier;  state: Exp;  lines: +185 -195
>   Doc fixes.
>   (comint-last-input-start, comint-last-input-end)
>   (comint-last-output-start): Give them values.
>   (comint-arguments): Avoid warning.
>   (comint-skip-prompt): Use line-end-position.
>   (comint-backward-matching-input): Rename arg to agree with doc.
>   (comint-extract-string): Use syntax-ppss.
>   (comint-dynamic-simple-complete): Delete useless list construction.
>   (comint-redirect-subvert-readonly): New.
>   (comint-redirect-preoutput-filter): Use it.

> but there are 91 difference between this file and the previous one.

> Before I wade through them, I'd like to ask if people agree that this
> change in behaviour should be reverted?

I'm not sure which change of behavior you're talking about.  AFAIK there's
just a bug, especially the "Wrong type argument: window-live-p, #<window 7>".


        Stefan

Re: Completion in GUD buffer

[Nick Roberts]

 > > but there are 91 difference between this file and the previous one.
 > 
 > > Before I wade through them, I'd like to ask if people agree that this
 > > change in behaviour should be reverted?
 > 
 > I'm not sure which change of behavior you're talking about.  AFAIK there's
 > just a bug, especially the "Wrong type argument: window-live-p, #<window 7>".

That part is a bug.  What I mean is do we want the user to able to select the
completions window like before, or do we only want to allow him to select an
item in the window with mouse-2?

-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: Completion in GUD buffer

[Stefan Monnier]

>> > but there are 91 difference between this file and the previous one.
>> 
>> > Before I wade through them, I'd like to ask if people agree that this
>> > change in behaviour should be reverted?
>> 
>> I'm not sure which change of behavior you're talking about.  AFAIK there's
>> just a bug, especially the "Wrong type argument: window-live-p, #<window 7>".

> That part is a bug.  What I mean is do we want the user to able to select the
> completions window like before, or do we only want to allow him to select an
> item in the window with mouse-2?

I don't see from the ChangeLog how the mentioned patch would have such
a consequence on the behavior of completion, so I suspect that once we've
fixed the bug, the question will disappear.


        Stefan

Re: Completion in GUD buffer

[Nick Roberts]

 > > That part is a bug.  What I mean is do we want the user to able to select
 > > the completions window like before, or do we only want to allow him to
 > > select an item in the window with mouse-2?
 > 
 > I don't see from the ChangeLog how the mentioned patch would have such
 > a consequence on the behavior of completion, so I suspect that once we've
 > fixed the bug, the question will disappear.

At first glance comint-dynamic-simple-complete is a candidate (just because
its got the word complete in it).  If I evaluate comint.el.~1.294~ and
comint-update-fence from current comint.el and run current M-x gdb I see
current behaviour. If I evaluate comint.el.~1.293~ and comint-update-fence
from current comint.el and run current M-x gdb I see the old behaviour.
I think I'm looking at the right ChangeLog entry.  I'll explore further.


-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: Completion in GUD buffer

[Stefan Monnier]

>> > That part is a bug.  What I mean is do we want the user to able to select
>> > the completions window like before, or do we only want to allow him to
>> > select an item in the window with mouse-2?
>> 
>> I don't see from the ChangeLog how the mentioned patch would have such
>> a consequence on the behavior of completion, so I suspect that once we've
>> fixed the bug, the question will disappear.

> At first glance comint-dynamic-simple-complete is a candidate (just because
> its got the word complete in it).  If I evaluate comint.el.~1.294~ and
> comint-update-fence from current comint.el and run current M-x gdb I see
> current behaviour. If I evaluate comint.el.~1.293~ and comint-update-fence
> from current comint.el and run current M-x gdb I see the old behaviour.
> I think I'm looking at the right ChangeLog entry.  I'll explore further.

Sorry, I meant to say that the log entry doesn't indicate that the change in
behavior was intended.  Which is why fixing the bug will probably answer the
question as well.


        Stefan

Re: Completion in GUD buffer

[Nick Roberts]

 > > At first glance comint-dynamic-simple-complete is a candidate (just because
 > > its got the word complete in it).  If I evaluate comint.el.~1.294~ and
 > > comint-update-fence from current comint.el and run current M-x gdb I see
 > > current behaviour. If I evaluate comint.el.~1.293~ and comint-update-fence
 > > from current comint.el and run current M-x gdb I see the old behaviour.
 > > I think I'm looking at the right ChangeLog entry.  I'll explore further.
 > 
 > Sorry, I meant to say that the log entry doesn't indicate that the change in
 > behavior was intended.  Which is why fixing the bug will probably answer the
 > question as well.

I think I've found bug now, which was actually introduced in version 1.292.
I didn't mean to finger point, its just that sometimes posting the change
generates some insight (I'm not familiar with this file).  However it would
help if I pointed to the right change.  Sorry about that but it should work
now, at least.

-- 
Nick                                           http://www.inet.net.nz/~nickrob