bug#17375: 24.4.50; minibuffer completion by SPC

bug#17375: 24.4.50; minibuffer completion by SPC

[Katsumi Yamaoka]

Hi,

When I visit a hidden buffer (i.e., the name begins with SPC),
now I need to type:

C-x b C-q SPC bla bla...

But IIRC, C-q was not necessary till recently.  Even so, SPC for
the second or later did the completion.  Is there a way to get
the old behavior?

Thanks.

In GNU Emacs 24.4.50.1 (i686-pc-cygwin, GTK+ Version 3.10.7)
 of 2014-04-30 on localhost
Repository revision: 117031 monnier@iro.umontreal.ca-20140429151607-qnkgbymwfaj5ut08
Windowing system distributor `The Cygwin/X Project', version 11.0.11501000
Configured using:
 `configure --verbose --with-x-toolkit=gtk3'

bug#17375: 24.4.50; minibuffer completion by SPC

[Stefan Monnier]

> When I visit a hidden buffer (i.e., the name begins with SPC),
> now I need to type:
> C-x b C-q SPC bla bla...

Why?  What happens when you hit C-x b SPC?


        Stefan

bug#17375: 24.4.50; minibuffer completion by SPC

[Katsumi Yamaoka]

Stefan Monnier <monnier@iro.umontreal.ca> wrote:
>> When I visit a hidden buffer (i.e., the name begins with SPC),
>> now I need to type:
>> C-x b C-q SPC bla bla...

> Why?  What happens when you hit C-x b SPC?

It showed, as the candidates, all buffers of which the names don't
begin with SPC.  (I'll verify it again on Cygwin nine hours after.)

bug#17375: 24.4.50; minibuffer completion by SPC

[Katsumi Yamaoka]

Katsumi Yamaoka <yamaoka@jpl.org> wrote:
> Stefan Monnier <monnier@iro.umontreal.ca> wrote:
>>> When I visit a hidden buffer (i.e., the name begins with SPC),
>>> now I need to type:
>>> C-x b C-q SPC bla bla...

>> Why?  What happens when you hit C-x b SPC?

> It showed, as the candidates, all buffers of which the names don't
> begin with SPC.  (I'll verify it again on Cygwin nine hours after.)

I verified that it doesn't happen on the most fresh Emacs built
on Linux.

In GNU Emacs 24.4.50.1 (x86_64-unknown-linux-gnu, X toolkit, Xaw3d scroll bars)
 of 2014-04-30 on localhost.localdomain
Repository revision: 117034 yamaoka@jpl.org-20140430092109-eivcc464i01u0ogk
Windowing system distributor `Fedora Project', version 11.0.11303000
Configured using:
 `configure --verbose --with-x-toolkit=lucid --without-xim
 --without-selinux'

bug#17375: 24.4.50; minibuffer completion by SPC

[Stefan Monnier]

> I verified that it doesn't happen on the most fresh Emacs built
> on Linux.

It probably depends on some particular configuration or situation.
If you bump into it again, try to see if you can reproduce it, or if it
depends on, I don't know, some particular buffer names or something.


        Stefan

bug#17375: 24.4.50; minibuffer completion by SPC

[Katsumi Yamaoka]

On Wed, 30 Apr 2014 14:07:00 -0400, Stefan Monnier wrote:
>> I verified that it doesn't happen on the most fresh Emacs built
>> on Linux.

> It probably depends on some particular configuration or situation.
> If you bump into it again, try to see if you can reproduce it, or if it
> depends on, I don't know, some particular buffer names or something.

I haven't tested it with `emacs -Q', sorry.  In not only that case
but also the case I launch emacs with no argument, I confirmed it
doesn't happen.  But at last I found a recipe to reproduce this:

(get-buffer-create "foo-bar")

That's all!

Could you try this on `emacs -Q'?  And you will see that the first
SPC for `C-x b' doesn't insert a space (I'm not quite sure if it
happens on every platform, though).
Maybe a buffer-name containing a hyphen triggers this problem.

Thanks.

In GNU Emacs 24.4.50.1 (i686-pc-cygwin, GTK+ Version 3.10.7)
 of 2014-05-01 on localhost
Repository revision: 117036 rrt@sc3d.org-20140430213957-ilgmnyhte50ipdt2
Windowing system distributor `The Cygwin/X Project', version 11.0.11501000
Configured using:
 `configure --verbose --with-x-toolkit=gtk3'

bug#17375: 24.4.50; minibuffer completion by SPC

[Stefan Monnier]

> (get-buffer-create "foo-bar")
> That's all!
> Could you try this on `emacs -Q'?  And you will see that the first
> SPC for `C-x b' doesn't insert a space (I'm not quite sure if it
> happens on every platform, though).
> Maybe a buffer-name containing a hyphen triggers this problem.

Indeed, that's it.
It's due to the fix for http://debbugs.gnu.org/15980.

I.e. if we have buffer names " toto" and "foo-bar", then both " " and
"-" have valid completions (remember: C-x b uses substring completion),
so bug#15980 says that instead of choosing arbitrarily between " " and
"-", we should just leave the input unchanged.

I must say I dislike "word completion".  Both because its behavior is
largely ill-defined (e.g. what does it mean to "complete at most
a single word" when the completion table completes "a-b-$FOO-a" to
"a-b-c-a-d-a"?), and because I'd rather leave SPC as a self-insert key.

So in this case, I'd tend to prefer reverting the fix for bug#15980, so
that " " is preferred over "-" on the premise that it makes SPC behave
a bit more like self-insert-command.


        Stefan

bug#17375: 24.4.50; minibuffer completion by SPC

[Stefan Monnier]

> So in this case, I'd tend to prefer reverting the fix for bug#15980, so
> that " " is preferred over "-" on the premise that it makes SPC behave
> a bit more like self-insert-command.

I've just done that, thank you,


        Stefan