Re: About the zh_CN.GB2312 locale

[Cyrille Chepelov <cyrille chepelov org>]

Le Tue, May 28, 2002, à 11:32:48AM +0800, Zhang Lin-bo a écrit:

> I have modified po/zh_CN.po following your instructions.
> Dia still crashes, even if I set the locale to en_US.
> So the problem should be somewhere else, either
> within my system or within dia. Is there something
> I can do in order to locate the problem?

Uh. Now I'm puzzled. I guess it'll take being more defensive on the loop you
have highlighted in your previous message. Please still forward the .po
changes to the zh_CN.po maintainer, unless the Postscript font names are the
what the contents of the translated strings were.
 
> Also, I noticed some bad translations in zh_CN.po
> (e.g., the string "Dia" got translated, and it shouldn't be?).

I don't know.
 
> Since freetype does not work on my system,
> I spent some time trying to make Chinese characters
> in the eps/ps files to show up. Since I'm not familiar
> with PS fonts and cmaps, I did not succeed. But I found
> out that the TTF Chinese fonts shipped with rh7.3/rh7.2
> work well with the UTF-8 encoding, so I modified dia to
> not reencode the characters in the PS file, but instead output
> raw utf8 characters - the generated PS files seem to work pretty
> well, at least for the two hardcoded simplified Chinese fonts
> and for the western fonts I have so far tested (redhat distributions

When you mean "Western", do you mean latin1, latin0 or do you mean ASCII ?

I'm afraid outside Ghostscript, UTF-8 is something undefined.
However, it would be very nice if we could work out something using
the re-encoding and your existing fonts (I'll need samples of working and
non-working files, relevant fonts, and a PNG screen shot of expected
results).

> I think it may be a useful feature to allow users to choose
> whether to reencode the characters in the PS files.
> The patch file against dia-0.90.RC2 is attached.
> I'd like that someone looks at my code and decides whether
> it's worth to integrate the patch into dia, by either adding
> a 'configure' option or a runtime dialog box, to allow users
> to choose to use 8-bit utf8 characters in their PS files.

Until you can show me a spec from Adobe saying that UTF-8 is fine in
Postscript, the answer is 'no'. I won't let dia generate
Ghostscript-specific code. However, I do want to make dia work for all
locales using all scripts; I would prefer the map-switching code to be made
working for zh_CN (Akira TAGOH made a lot of work in this area during the
spring, and it was supposed to work for all CJK languages)

Specifically, I don't understand why the \uni1234 notation doesn't work in
your case -- it definitely should.

A couple test cases would be worth trying:
	* a diagram with a few symbols (less than 256)
	* a diagram with a lot of symbols (more than 256)

You may want however to test whether enabling only the first half of your
patch (the lib/font.c section) and leaving the lib/ps-utf8.c code is enough 
to solve the problem ?


PS: don't forget to subscribe to the list. Currently, an administrator has
to unlock each of your posts.

	-- Cyrille


-- 
Grumpf.