Re: utf-8 support: much better, a few things still wrong

From: Andrew Dunbar (
Date: Sat Sep 14 2002 - 00:13:45 EDT

  • Next message: Andrew Dunbar: "Re: [Bug 3988] - AbiWord crashes when trying to open en-GB.strings"

     --- Raphael Finkel <> wrote:
    > I just got today's CVS version and compiled on Linux
    > against gtk-2.0, glib-2.0,
    > --enable-xft. I am using Abiword for Yiddish texts.
    > 1. Apparently I no longer need to have a
    > fonts/utf-8 directory, even though I set my language
    > variable to he_IL.utf-8. (good; building such a
    > directory is a true nuisance)

    This is what Xft/fontconfig are supposed to make
    easier. Glad to see it really helps.

    > 2. As soon as I click on the Font menu (where it
    > starts with Times New Roman) I get an abort (not
    > good). Luckily, I can use the ^D menu without
    > problems (good).

    Definitely file a bug report for this. The more
    detail the better. Since you're building ABiWord
    yourself, you ought to build a debug version and
    send us the debugging messages from prior to the

    > 3. By renaming my dictionary to yiddish-yivo.hash
    > (likewise for encoding), I get proper Yiddish spell
    > checking, with underlining of badly spelled words
    > (good).

    This makes me especially happy!

    > 4. When I have an underlined word indicating bad
    > spelling, right click brings up a menu with
    > suggestions. All the suggestions are invisible. I
    > can only figure out what they contain by trying them
    > (not good).

    Open a seperate bug report for this too and answer
    these questions there:
    Are the suggestions completely blank or are they
    questions marks?
    Are you running a UTF-8 locale? A Hebrew locale? An
    English locale?
    My guess is that older GTK only used 8-bit encodings
    for its GUI. The good part is that if your locale is
    UTF-8, so will be the GUI encoding - so all languages
    should "just work" in the GUI - font problems not-
    If you're running a Hebrew locale, I would expect the
    Hebrew letters to display and any special Yiddish
    characters to display question marks.
    A screenshot will help a lot. (In the bug report).

    > 5. The text displays fine with Times New Roman and
    > with Arial, except that combining characters not
    > only modify the previous character (good) but leave
    > an empty rectangle in the next space (not good).

    File another seperate bug report for this one with
    screenshot. Tomas is excellent Hebrew shaping so I'm
    sure he can fix your Yiddish-specific shaping problems

    > 6. The font-choice menu (^d) doesn't show a sample
    > of text until I click on one of the suggested fonts,
    > even though the intially selected font can display
    > the text (not great). This display does not show
    > combining codes combining with the previous letter
    > (not good). Nor does it show the text in proper
    > order (not good).

    Again this sounds like a bug report and screenshot.
    I suspect the bidi ordering problem may be due to not
    having GTK2 support yet. GTK2 uses pango so it can
    handle bidi even in GUI elements.

    > 7. The spelling-correction box I get with F7 shows
    > incorrectly spelled text in a different font (ok) in
    > correct Bidi order (good) with combining marks
    > properly combining (good) and with no empty
    > rectangles after combining marks (good). Double-
    > clicking on one of the suggestions does not have the
    > effect of auto-invoking "change" (ok, but could be
    > improved).

    Wow some of this is quite impressive. The
    problem seems strange to me - that shouldn't care
    the language and has always worked for me.

    By the way, do you think the spelling suggestions
    are good quality choices. I've been expecting UTF-8
    ispell to make poor spelling suggestions but I'd be
    very happy to be wrong.

    > Typing in a replacement in the Change to: box does
    > not auto-invoke "change" when I type the <cr> (ok,
    > but could be improved). I can type in utf-8
    > characters in the "Change to" box (using xmodmap to
    > re-key my keyboard) (good), and the characters are
    > properly displayed (good). But when I activate with
    > "change", I discover that some characters are
    > misunderstood, primarily (but not only) combining
    > characters (not good); they are replaced by other
    > characters.

    More bug reports with screenshots needed (:
    The wrong-characters problem needs deeper
    such as exactly which characters get replaced with
    exactly which wrong characters? Are the errors
    consistent every time or random?

    > 8. I have started to build a yi.po file for
    > Yiddish. I've done about 1/5 of
    > the file. I'll provide it when I am done.

    Can you please provide the Yiddish spelling hash?
    It may be too large to mail to me or to the list but
    if you can give us a URL to download it from I'll get
    it included.

    Looking forward to seeing this all working!
    Andrew Dunbar.
    > Raphael Finkel


    Do You Yahoo!?
    Everything you'll ever need on one web page
    from News and Sport to Email and Music Charts

    This archive was generated by hypermail 2.1.4 : Sat Sep 14 2002 - 00:18:17 EDT