Re: fix for bug 10994 breaks ido customizations in major way

[Le Wang <l26wang@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1771 bytes --]

I sent this off list by accident.

On Fri, May 3, 2013 at 12:13 PM, Leo Liu <sdl.web@gmail.com> wrote:

> On 2013-05-03 01:57 +0800, Le Wang wrote:
> > bug:
> >
> http://emacs.1067599.n5.nabble.com/bug-10994-23-3-ido-mode-ido-next-match-ido-prev-match-work-wrong-with-same-elements-td300.html
> >
> > The committed fix converts equal to eq, causing any plugin that
> propertizes
> > the completion list to hang Emacs.
> >
> > This bug is filed as: http://debbugs.gnu.org/cgi/bugreport.cgi?bug=14334
> >
> > The original bug report is essentially this:
> >
> > (ido-completing-read "dat is whrong -> " '("2" "3" "3" "3" "4" "5"))
> >
> > What's the expected behaviour in this case?  Shouldn't the duplicates
> from
> > the list just be removed?
>
> I put in the fix. It was based on how ido-matches are built from
> ido-cur-list. So it fixes the bug.
>

> I am not sure what new things you are doing with ido you will have to
> look into this yourself..
>

There are a few ido customizations floating around that propertizes text.
 This will break all of them.  I don't think this fix is acceptable.


> Ideally ido should not modified the list fed to it. i.e.


 That's not happening.  The list passed to ido is not modified.
 ido-set-matches-1 returns a newlist,

(ido-completing-read "dat is whrong -> " '("2" "3" "3" "3" "4" "5"))
>
> should be able to rotate for- band backwards without removing the dups.
> It would be odd-looking if a user supply one list and ido display
> another.
>

I asked for a use-case of a completion list with duplicate strings.   Do
you have one?

If I do a completing read outside of ido from "emacs -Q" duplicates are not
shown.

I tried

    (completing-read ": " '("a" "ab" "a"))

when I press <tab> only two items are shown.


-- 
Le

[-- Attachment #2: Type: text/html, Size: 3994 bytes --]