Re: zsh/complist colours improperly handle multibyte characters

[Peter Stephenson <p.stephenson@xxxxxxxxxxx>]

On Fri, 21 Oct 2016 12:09:34 +1100
Danielle McLean <gopsychonauts@xxxxxxxxx> wrote:
> Although zsh patterns usually have full support for multibyte
> characters, a zsh pattern containing certain multibyte characters will
> fail to match if provided to zsh/complist using $ZLS_COLORS or the
> list-colors zstyle.

Yes, unfortunately character-by-character analysis is built into
completion at quite a low level.

I don't know a great deal about the complist code --- except it's
appeared utterly opaque when I have looked at it --- but I did
try to get grips with the matching control, where there are similar
problems, and utterly failed.  The assumption that a byte in the
string on the command line and a byte in the test string can be
considered equivalent is all-pervasive, but the sort of tricks in use
didn't make converting to wide characters a particularly attractive
optin, either. It became clear this was a huge job I didn't have a clue
about.

complist may not be quite that bad, and indeed character *matching*
isn't a big part of it so this may not be a particularly difficult
problem when it comes down to it, but unless anyone makes it their
life's work to get to grips with it it's probably stuck --- it's clear
Bart and I aren't ever going to have that sort of time.

pws