Re: Completion strangeness

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On May 19,  1:29pm, Oliver Kiddle wrote:
} Subject: Re: Completion strangeness
}
} This is unrelated to the bug Bart pointed to.

I was about to say, "then I was mistaken."

} We can cut this down to a minimal setup to see problems:

[snip]

} Patterns like '???' and '????' in compset also work. If compset has
} removed four or more characters, we get the separator character (.)
} inserted. So the problem is in the C code.

I fiddled with this just a little.  It's not always "four or more" --
it's somehow related to the length of the (un?)ambiguous prefix in the
part compset leaves behind.  For example, if you change
	yupa-dmz.valley-of-wind.krose.org
to
	yuppa-dmz.valley-of-wind.krose.org
in your minimal example (extra 'p'), then the problem appears only
after compset has removed five or more characters.  If you change the
set of completions to
	yupa-dmz.valley-of-wind.krose.org \ 
	yipa.krose.org yopa.valley-of-wind.krose.org
then the problem appears only after seven or more characters are
removed.  And you can see it with

% t 123456@y

so it's not related to "kro" forming a substring of the right side.