Created attachment 109412 [details]
Test glyphs in pango

The program tries to get the glyph using both FcFreeTypeCharIndex() and pango_shape() but the results are not always the same. The error could be in pango_font_get_glyph(), copied from WebKit, but it seems ok to me.

Run it with (the ja_JP.UTF-8 locale must be enabled to test it):
$ gcc -o lang-test `pkg-config --cflags --libs gtk+-2.0 cairo-ft fontconfig freetype2` lang-test.c

$ LANG=C ./lang-test
freetype: 43, pango: 43

$ LANG=ja_JP.UTF-8 ./lang-test
freetype: 42, pango: 43

You have noticed that the font passed into your pango_font_get_glyph() isn't
used, right?

Basically, the pango_itemize() is picking a different font from the
"first choice" font that pango_font_map_load_font() returns.

Nothing using a function like your pango_font_get_glyph() should be called
a "Pango" backend, though maybe there was some significant distortion
in the translation to a test case. None of the capabilities of Pango
(combining fonts, shaping international text, kerning, etc, etc, etc)
will be available.

(Also, don't put even static functions into the pango_ namespace.
If they conflict with future pango functions, that isn't our problem.)

Sorry for the useless bug report, but I really didn't notice it.

There is some significant distortion from the code in WebKit, that code is only used for "simple" (and optimised in some ways I don't want to know) text rendering. For more complex things 

I didn't write the code, just trying to fix a bug, so I didn't choose the name but probably it should be a C++ private method to follow WebKit conventions.

Thanks.

Gcc should have warned you that the font argument to your pango_font_get_glyph() function is unused.  That's quite telling.