Re: bug#613: 23.0.60; Unicode display problems

[Kenichi Handa <handa@m17n.org>]

In article <18604.9770.893099.425147@priss.frightenedpiglet.com>, sand@blarg.net writes:

> Kenichi Handa writes:
[...]
> > Now I see what is wrong.  Dejavu Sans Mono lacks some of
> > charaters (#x201C #x2200 #x2500) to be used as `symbol'
> > font, so the other arbitrary font that contains all of those
> > characters are selected.  This setting was added by the
> > request in the attached mail, but it seems that the
> > restriction is too strong for symbol fonts.  So, I deleted
> > that restriction.  Please try again with the latest CVS code.

> Removing the symbol restrictions will allow Emacs to pick (incorrect)
> ISO8859-1 fonts for these characters again.  Have you changed
> something in the C code to prevent that?

Now the entry for symbols is this:

     (symbol ,(font-spec :registry "iso10646-1"))

So, at least, Emacs tries to select an iso10646-1 font that
is closest (in family, style, size) to the default font.

> Looking at the DejaVu Sans Mono glyphs in "gucharmap", it appears that
> #x2200 is the problem.  The #x201C, #x2200, #x2500 combination was an
> attempt to find "representative" symbol codepoints, but that
> combination isn't all-or-nothing.  Can you first try leaving just
> #x201C in `script-representative-chars'?  That's LEFT DOUBLE QUOTATION
> MARK, so it should have wide coverage.

I think the typical usage of symbol characters is to use a
few of them in a normal text.  In that case, the most
suitable font for those symbols is the default font
(provided that it supports those symbols) rather than a font
that covers wider range of symbols.  So, I think there's no
merit in restricting fonts in the above way.

As a result, when a sequence of many symbols is displayed,
Emacs will not use the same font for them, but, I think it's
not a big problem.

---
Kenichi Handa
handa@ni.aist.go.jp