Package: emacs;
Reported by: Steffan <smias <at> yandex.ru>
Date: Wed, 4 Jul 2012 18:43:12 UTC
Severity: normal
Found in version 24.1
Done: Stefan Kangas <stefan <at> marxist.se>
Bug is archived. No further changes may be made.
Message #62 received at 11860 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Kenichi Handa <handa <at> gnu.org>, Jason Rumney <jasonr <at> gnu.org> Cc: 11860 <at> debbugs.gnu.org, smias <at> yandex.ru Subject: Re: bug#11860: 24.1; Arabic - Harakat (diacritics, short vowels) don't appear Date: Sun, 19 Aug 2012 21:22:40 +0300
> From: Kenichi Handa <handa <at> gnu.org>
> Cc: eliz <at> gnu.org, 11860 <at> debbugs.gnu.org, smias <at> yandex.ru
> Date: Sat, 18 Aug 2012 11:45:27 +0900
>
> So, apparently Emacs on Windows and GNU/Linux uses the
> different metrics of glyphs. As the shaper on GNU/Linux
> (m17n-lib library) works correctly for the same font, and
> the other applications on Windows have no problem, I suspect
> that the problem is in Emacs' interface with uniscribe
> (w32font.c or w32uniscribe.c).
I agree.
> If this problem happens only for bidi scripts
Can you suggest how to test this hypothesis?
> one possibility is that Emacs's rendering engine (xdisp.c) expects
> glyphs in a glyph-string are rendered in that order from left to
> right, but the returned glyph-string on Windows should be rendered
> in reverse order.
You may be right, but it's hard to be sure. At least the advances[]
array returned by ScriptPlace seems to point into that direction.
Here's what I see in the debugger:
Breakpoint 8, uniscribe_shape (lgstring=55041941) at w32uniscribe.c:373
373 LGLYPH_SET_CHAR (lglyph, chars[items[i].iCharPos
(gdb) p items <at> nitems
$1 = {0x35195a0}
(gdb) p items[0]@nitems
$2 = {{
iCharPos = 0,
a = {
eScript = 26,
fRTL = 1,
fLayoutRTL = 1,
fLinkBefore = 0,
fLinkAfter = 0,
fLogicalOrder = 1,
fNoGlyphIndex = 0,
s = {
uBidiLevel = 1,
fOverrideDirection = 0,
fInhibitSymSwap = 0,
fCharShape = 0,
fDigitSubstitute = 0,
fInhibitLigate = 0,
fDisplayZWG = 0,
fArabicNumContext = 0,
fGcpClusters = 0,
fReserved = 0,
fEngineReserved = 0
}
}
}}
(gdb) p nitems
$3 = 1
(gdb) p nglyphs
$4 = 2
(gdb) p advances[0]@nglyphs
$5 = {8, 0}
(gdb) p offsets[0]@nglyphs
$6 = {{
du = 0,
dv = 0
}, {
du = 1,
dv = -2
}}
(gdb) p chars[0]@2
$7 = L"\x639\x652"
(Note that the fRTL member of items[0].a is set to TRUE.) My
understanding of the advances[] array is that it gives, for each glyph
in the cluster, the number of pixels to advance to the right after
drawing the glyph. So the fact that it is 8 for the first (base)
character and zero for the second one tells me that this grapheme
cluster is supposed to be rendered in reverse order: first the Sukun,
then Ayin at the same location, and then advance by 8 pixels for the
next character. Is this correct?
If it is correct, then how come the glyphs shown on GNU/Linux also
have non-zero value of xadvance:
[0 1 1593 969 8 2 8 4 4 nil]
[0 1 1618 760 0 -6 -3 8 -11 [-9 2 0]]
The value 8 after 969 comes directly from xadvance, as this code in
ftfont.c shows:
LGLYPH_SET_WIDTH (lglyph, g->xadv >> 6);
Is the meaning of xadvance in libotf different from its meaning in
Uniscribe? (And why is the glyph string element called WIDTH instead
of ADVANCE?) If not, what am I missing?
> For instance, in the above case, we may have to render glyphs in
> this order (diacritical mark first):
>
> [0 1 1593 760 0 3 6 12 4 [1 -2 0]]
> [0 1 1593 969 8 1 8 12 4 nil]
I tried the naive patch below, but it didn't quite work. It seems
like those changes somehow prevented character composition. Perhaps
Handa-san could give me some guidance here.
> I think the further debugging must be done by those who
> knows uniscribe, w32font.c, and w32uniscribe.c.
It's very hard, given that glyph-string documentation leaves a lot to
be desired, and the way its various components are used during drawing
is also left without clear documentation. E.g., this:
FROM-IDX and TO-IDX are used internally and should not be touched.
is not really helpful for explaining what are FROM-IDX and TO-IDX, so
how can I figure out whether the code you asked about is doing TRT?
And without knowing what is each component of glyph-string used for
during drawing, how can I compare the values produced by Uniscribe
APIs with what glyph-string needs? If someone could explain all those
things, it would make debugging possible. Otherwise, I'm just
randomly poking around...
Here's the patch I tried:
--- src/w32uniscribe.c~ 2012-07-08 07:24:56.000000000 +0300
+++ src/w32uniscribe.c 2012-08-19 15:55:17.323623900 +0300
@@ -331,17 +331,13 @@ uniscribe_shape (Lisp_Object lgstring)
Lisp_Object lglyph = LGSTRING_GLYPH (lgstring, lglyph_index);
ABC char_metric;
unsigned gl;
+ int j1;
if (NILP (lglyph))
{
lglyph = Fmake_vector (make_number (LGLYPH_SIZE), Qnil);
LGSTRING_SET_GLYPH (lgstring, lglyph_index, lglyph);
}
- /* Copy to a 32-bit data type to shut up the
- compiler warning in LGLYPH_SET_CODE about
- comparison being always false. */
- gl = glyphs[j];
- LGLYPH_SET_CODE (lglyph, gl);
/* Detect clusters, for linking codes back to
characters. */
@@ -365,6 +361,16 @@ uniscribe_shape (Lisp_Object lgstring)
}
}
}
+ if (items[i].a.fRTL)
+ j1 = to - (j - from);
+ else
+ j1 = j;
+
+ /* Copy to a 32-bit data type to shut up the
+ compiler warning in LGLYPH_SET_CODE about
+ comparison being always false. */
+ gl = glyphs[j1];
+ LGLYPH_SET_CODE (lglyph, gl);
LGLYPH_SET_CHAR (lglyph, chars[items[i].iCharPos
+ from]);
@@ -372,13 +378,13 @@ uniscribe_shape (Lisp_Object lgstring)
LGLYPH_SET_TO (lglyph, items[i].iCharPos + to);
/* Metrics. */
- LGLYPH_SET_WIDTH (lglyph, advances[j]);
+ LGLYPH_SET_WIDTH (lglyph, advances[j1]);
LGLYPH_SET_ASCENT (lglyph, font->ascent);
LGLYPH_SET_DESCENT (lglyph, font->descent);
result = ScriptGetGlyphABCWidth (context,
&(uniscribe_font->cache),
- glyphs[j], &char_metric);
+ glyphs[j1], &char_metric);
if (result == E_PENDING && !context)
{
/* Cache incomplete... */
@@ -387,7 +393,7 @@ uniscribe_shape (Lisp_Object lgstring)
old_font = SelectObject (context, FONT_HANDLE (font));
result = ScriptGetGlyphABCWidth (context,
&(uniscribe_font->cache),
- glyphs[j], &char_metric);
+ glyphs[j1], &char_metric);
}
if (SUCCEEDED (result))
@@ -399,17 +405,17 @@ uniscribe_shape (Lisp_Object lgstring)
else
{
LGLYPH_SET_LBEARING (lglyph, 0);
- LGLYPH_SET_RBEARING (lglyph, advances[j]);
+ LGLYPH_SET_RBEARING (lglyph, advances[j1]);
}
- if (offsets[j].du || offsets[j].dv)
+ if (offsets[j1].du || offsets[j1].dv)
{
Lisp_Object vec;
vec = Fmake_vector (make_number (3), Qnil);
- ASET (vec, 0, make_number (offsets[j].du));
- ASET (vec, 1, make_number (offsets[j].dv));
+ ASET (vec, 0, make_number (offsets[j1].du));
+ ASET (vec, 1, make_number (offsets[j1].dv));
/* Based on what ftfont.c does... */
- ASET (vec, 2, make_number (advances[j]));
+ ASET (vec, 2, make_number (advances[j1]));
LGLYPH_SET_ADJUSTMENT (lglyph, vec);
}
else
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.