Re: About pangowin32...

[Hans Breuer <hans breuer org>]

At 13:43 15.09.02 +0300, Tor Lillqvist wrote:
>Hi,
>
>Recently somebody complained that Chinese characters don't render in
>GTK+ 2.0 (i.e. pangowin32). I had a look at it, and I think the
>problem is that pangowin32 only uses a single Win32 font for each
>PangoWin32Font. For "sans" it seems to use Arial, and nothing else. My
>memory is rather hazy, but I think that it originally (over a year
>ago, back when the pangowin32.aliases file was still used) was
>supposed to use a list of Win32 fonts for each Pango font, in order to
>be able to pick characters from different scripts from different Win32
>fonts. (For instance Latin and Hebrew characters from Arial, Chinese
>characters from MingLiU, Thai characters from Angsana, etc.)
>
It appears to still work this way but there needs to be a global alias 
file named 'pango.aliases'. The format has changed as well but

sans = sans
sans += "bitstream cyberbit"

does have the desired visual impact :-)

[
  aliasing sans to sans seems to be required to avoid

** (TESTGTK.EXE:4294826355): WARNING **: Couldn't load font "Sans Italic
12" falling back to "Sans 12"

  when invoking testgtk::labels
]

This appears to be documented only by the implementation though.

>I think this ability was lost at some time, probably with alexl's
>changes over a year ago (2001-08-09: "Also drops the subfont thing,
>thereby simplifying the code immensely")
>
My first attempt was to #define  _WE_WANT_GLOBAL_ALIASES_ but the code
between doesn't compile anymore and appears to be not needed either.
Shouldn't it be deleted ?
 
>Does anybody have any clearer memory of this, or any idea what would
>be the best way to get the multiple-Win32-fonts-per-Pango-font feature
>back? I think it is rather essential that CJK character output should
>work on Windows in GTK+ 2.0, as it has done for a long time in GTK+
>1.3.0 (the gtk-1-3-win32-production branch).
>
No memory but new observations ;-) testgtk::labels does render 
almost all glyphs under win98. They look ok and can be coped to the
testgtk::entry . I don't know anything about entering foreign
languages by keyboard though ... 

>I must admit what little understanding of Pango's internal workings I
>had has rusted considerably during the last year or so...  Rather than
>hacking on pangowin32 (without really understanding what's going on),
>I am tempted to start using the FreeType2 backend on Windows. (Perhaps
>as an alternative to pangowin32, selectable at run time with an
>environment variable or command-line switch.) This would at least mean
>that the work other people have done on pangoft2 would benefit Win32,
>too. (For instance there are Arabic and Hebrew shapers for pangoft2)
>
I would like to have the freetype backend useable for win32 again
but only as alternative. The cvs head version is broken since

'Heads up on Pango HEAD'
http://mail.gnome.org/archives/gtk-devel-list/2002-June/msg00008.html

and I still doubt that it is the right way to port fontconfig to 
win32 to make pangoft2/win32 depend on an 'obscure' configuration
file. This kind of stuff appears to be available through the 
system apis. 
It is intentionally that the pangowin32 version does work reasonable 
without any system specific configuration file.

	Hans

-------- Hans "at" Breuer "dot" Org -----------
Tell me what you need, and I'll tell you how to 
get along without it.                -- Dilbert