[HarfBuzz] harfbuzz-ng: Branch 'master' - 21 commits
Behdad Esfahbod
behdad at kemper.freedesktop.org
Fri Jul 20 17:39:59 PDT 2012
src/hb-ot-layout-gpos-table.hh | 4
src/hb-ot-shape-complex-indic-machine.rl | 7
src/hb-ot-shape-complex-indic-private.hh | 5
src/hb-ot-shape-complex-indic.cc | 126 +++++-----
src/hb-private.hh | 1
src/hb-unicode.cc | 10
src/hb-uniscribe.cc | 34 +-
test/shaping/texts/in-tree/shaper-indic/indic/script-bengali/misc/misc.txt | 1
test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/MANIFEST | 1
test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/bindu.txt | 2
10 files changed, 108 insertions(+), 83 deletions(-)
New commits:
commit 3d4c111b7a13700b2f7a0b087eb3992283295f21
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 19:34:39 2012 -0400
Add a test case
diff --git a/test/shaping/texts/in-tree/shaper-indic/indic/script-bengali/misc/misc.txt b/test/shaping/texts/in-tree/shaper-indic/indic/script-bengali/misc/misc.txt
index 843ee4f..35ce952 100644
--- a/test/shaping/texts/in-tree/shaper-indic/indic/script-bengali/misc/misc.txt
+++ b/test/shaping/texts/in-tree/shaper-indic/indic/script-bengali/misc/misc.txt
@@ -49,3 +49,4 @@
অৗ
নà§à¦¤à§à¦°
তà§à¦¯à§
+চà§à¦¯à§à¦°
commit 92a1ad7bef9efb456ab87bd63818cfbed7da3f6f
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 18:38:27 2012 -0400
[Indic] Stop searching for base if a post form is found before below form
Improves Bengali and Gurmukhi. Malayalam regressed a bit. We will deal
with that later.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index b5ad4ae..ad55f77 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -505,6 +505,7 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
{
/* -> starting from the end of the syllable, move backwards */
unsigned int i = end;
+ bool seen_below = false;
do {
i--;
/* -> until a consonant is found */
@@ -513,11 +514,13 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
/* -> that does not have a below-base or post-base form
* (post-base forms have to follow below-base forms), */
if (info[i].indic_position() != POS_BELOW_C &&
- info[i].indic_position() != POS_POST_C)
+ (info[i].indic_position() != POS_POST_C || seen_below))
{
base = i;
break;
}
+ if (info[i].indic_position() == POS_BELOW_C)
+ seen_below = true;
/* -> or that is not a pre-base reordering Ra,
*
commit 4c450c703f8e4618c587bcd7ef46dcc1f2c7947b
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 18:13:04 2012 -0400
[Indic] Recompose Bengali Ya,Nukta
This is a bunch of hacks for now.
Improves Bengali a bit.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 0316691..b5ad4ae 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -407,7 +407,8 @@ hb_ot_shape_normalization_mode_t
_hb_ot_shape_complex_normalization_preference_indic (void)
{
/* We want split matras decomposed by the common shaping logic. */
- return HB_OT_SHAPE_NORMALIZATION_MODE_DECOMPOSED;
+ /* XXX sort this out after adding per-shaper normalizers. */
+ return HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS;
}
@@ -549,7 +550,6 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
if (base < start)
base = start; /* Just in case... */
-
/* -> If the syllable starts with Ra + Halant (in a script that has Reph)
* and has more than one consonant, Ra is excluded from candidates for
* base consonants. */
diff --git a/src/hb-unicode.cc b/src/hb-unicode.cc
index c527340..140f382 100644
--- a/src/hb-unicode.cc
+++ b/src/hb-unicode.cc
@@ -258,6 +258,14 @@ hb_unicode_compose (hb_unicode_funcs_t *ufuncs,
hb_codepoint_t *ab)
{
*ab = 0;
+ /* XXX, this belongs to indic normalizer. */
+ if ((FLAG (hb_unicode_general_category (ufuncs, a)) &
+ (FLAG (HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK) |
+ FLAG (HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK) |
+ FLAG (HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK))))
+ return false;
+ /* XXX, add composition-exclusion exceptions to Indic shaper. */
+ if (a == 0x09AF && b == 0x09BC) { *ab = 0x09DF; return true; }
return ufuncs->func.compose (ufuncs, a, b, ab, ufuncs->user_data.compose);
}
commit e9c0f152a38cb2e76650a3e43f7fdcda266af696
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 17:05:46 2012 -0400
[Uniscribe] Fix script fallback
Gurmukhi failures half now. Others changed slightly.
diff --git a/src/hb-uniscribe.cc b/src/hb-uniscribe.cc
index e5b98a8..d41542b 100644
--- a/src/hb-uniscribe.cc
+++ b/src/hb-uniscribe.cc
@@ -331,10 +331,6 @@ retry:
}
OPENTYPE_TAG language_tag = hb_uint32_swap (hb_ot_tag_from_language (buffer->props.language));
- hb_tag_t buffer_script_tags[2];
- hb_ot_tags_from_script (buffer->props.script,
- &buffer_script_tags[0],
- &buffer_script_tags[1]);
unsigned int glyphs_offset = 0;
unsigned int glyphs_len;
@@ -345,20 +341,11 @@ retry:
unsigned int chars_offset = items[i].iCharPos;
unsigned int item_chars_len = items[i + 1].iCharPos - chars_offset;
- OPENTYPE_TAG script_tag;
- /* We ignore what script tag Uniscribe chose, except to differentiate
- * between old/new tags. Not sure if this picks DFLT up correctly...
- * This also screws things up as the item.analysis also has an opaque
- * script member. */
- if (script_tags[i] == hb_uint32_swap (buffer_script_tags[1]))
- script_tag = hb_uint32_swap (buffer_script_tags[1]);
- else
- script_tag = hb_uint32_swap (buffer_script_tags[0]);
-
+ retry_shape:
hr = ScriptShapeOpenType (font_data->hdc,
&font_data->script_cache,
&items[i].a,
- script_tag,
+ script_tags[i],
language_tag,
range_char_counts,
range_properties,
@@ -373,9 +360,6 @@ retry:
glyph_props + glyphs_offset,
(int *) &glyphs_len);
- for (unsigned int j = chars_offset; j < chars_offset + item_chars_len; j++)
- log_clusters[j] += glyphs_offset;
-
if (unlikely (items[i].a.fNoGlyphIndex))
FAIL ("ScriptShapeOpenType() set fNoGlyphIndex");
if (unlikely (hr == E_OUTOFMEMORY))
@@ -386,14 +370,24 @@ retry:
goto retry;
}
if (unlikely (hr == USP_E_SCRIPT_NOT_IN_FONT))
- FAIL ("ScriptShapeOpenType() failed: Font doesn't support script");
+ {
+ if (items[i].a.eScript == SCRIPT_UNDEFINED)
+ FAIL ("ScriptShapeOpenType() failed: Font doesn't support script");
+ items[i].a.eScript = SCRIPT_UNDEFINED;
+ goto retry_shape;
+ }
if (unlikely (FAILED (hr)))
+ {
FAIL ("ScriptShapeOpenType() failed: 0x%08xL", hr);
+ }
+
+ for (unsigned int j = chars_offset; j < chars_offset + item_chars_len; j++)
+ log_clusters[j] += glyphs_offset;
hr = ScriptPlaceOpenType (font_data->hdc,
&font_data->script_cache,
&items[i].a,
- script_tag,
+ script_tags[i],
language_tag,
range_char_counts,
range_properties,
commit 5791f329159c9863317e2b507514c29321be31a7
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 16:26:55 2012 -0400
[Indic] Allow a ZWNJ after SM's
Malayalam failures go way down. Other scripts benefitted slightly too.
Sinhala had one or two test regressions, but...
diff --git a/src/hb-ot-shape-complex-indic-machine.rl b/src/hb-ot-shape-complex-indic-machine.rl
index 4be7698..62091e2 100644
--- a/src/hb-ot-shape-complex-indic-machine.rl
+++ b/src/hb-ot-shape-complex-indic-machine.rl
@@ -65,7 +65,7 @@ reph = (Ra H | Repha); # possible reph
cn = c.n?;
forced_rakar = ZWJ H ZWJ Ra;
matra_group = z*.M.N?.(H | forced_rakar)?;
-syllable_tail = SM? (Coeng (cn|V))? (VD VD?)?;
+syllable_tail = (SM.ZWNJ?)? (Coeng (cn|V))? (VD VD?)?;
place_holder = NBSP | DOTTEDCIRCLE;
halant_group = (z?.h.ZWJ?);
final_halant_group = halant_group | h.ZWNJ;
commit 34ae336f3fae93ef9372881d545c817bce383041
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 16:17:28 2012 -0400
[Indic] Improve Reph AfterMain positioning
Fixes 20 out of 48 failing Oriya tests. Failure rate down to 0.066% now.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 228fc63..0316691 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -975,8 +975,7 @@ final_reordering_syllable (hb_buffer_t *buffer,
{
new_reph_pos = base;
/* XXX Skip potential pre-base reordering Ra. */
- while (new_reph_pos < end &&
- !( FLAG (info[new_reph_pos + 1].indic_position()) & (FLAG (POS_BELOW_C) | FLAG (POS_POST_C) | FLAG (POS_AFTER_POST) | FLAG (POS_SMVD))))
+ while (new_reph_pos + 1 < end && info[new_reph_pos + 1].indic_position() <= POS_AFTER_MAIN)
new_reph_pos++;
if (new_reph_pos < end)
goto reph_move;
commit bdd080431a40bc941ece3230f338b94a46bd12a2
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 16:03:09 2012 -0400
[Indic] Reposition Oriya Candrabindu
Oriya failures down from 0.65% to 0.20%.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index f75a500..228fc63 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -291,6 +291,7 @@ set_indic_properties (hb_glyph_info_t &info, hb_ot_map_t *map, hb_font_t *font)
pos = POS_SMVD;
}
+ if (unlikely (u == 0x0B01)) pos = POS_BEFORE_SUB; /* Oriya Bindu is BeforeSub in the spec. */
info.indic_category() = cat;
diff --git a/test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/MANIFEST b/test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/MANIFEST
index 29cfb2f..66a2468 100644
--- a/test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/MANIFEST
+++ b/test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/MANIFEST
@@ -1 +1,2 @@
+bindu.txt
misc.txt
diff --git a/test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/bindu.txt b/test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/bindu.txt
new file mode 100644
index 0000000..13de6ee
--- /dev/null
+++ b/test/shaping/texts/in-tree/shaper-indic/indic/script-oriya/misc/bindu.txt
@@ -0,0 +1,2 @@
+ମàà¬
+ମàଂ
commit 5f0eaaad129ff04d56b8756bebf19fbc242718c9
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 15:47:24 2012 -0400
[Indic] Fix base search in final_reordering
Fixes most Malayalam failures. Down from 1.6% to 0.38% now. Fixes a
few more in other scripts too.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 1a75e78..f75a500 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -712,7 +712,7 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
}
/* XXX This will not match for old-Indic spec since the Halant-Ra order is reversed already. */
- if (basic_mask_array[PREF] && base + 3 <= end)
+ if (basic_mask_array[PREF] && base + 2 < end)
{
/* Find a Halant,Ra sequence and mark it fore pre-base reordering processing. */
for (unsigned int i = base + 1; i + 1 < end; i++)
@@ -829,9 +829,13 @@ final_reordering_syllable (hb_buffer_t *buffer,
*/
/* Find base again */
- unsigned int base = end;
- while (start < base && info[base - 1].indic_position() >= POS_BASE_C)
- base--;
+ unsigned int base;
+ for (base = start; base < end; base++)
+ if (info[base].indic_position() >= POS_BASE_C) {
+ if (start < base && info[base].indic_position() > POS_BASE_C)
+ base--;
+ break;
+ }
unsigned int start_of_last_cluster = base;
commit 81202bd860e4034c18d9f80c5a4f33d9f48463a3
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 15:10:02 2012 -0400
[Indic] Don't attach SM/VD to other characters
diff --git a/src/hb-ot-shape-complex-indic-private.hh b/src/hb-ot-shape-complex-indic-private.hh
index 5f4856e..4a4c8c0 100644
--- a/src/hb-ot-shape-complex-indic-private.hh
+++ b/src/hb-ot-shape-complex-indic-private.hh
@@ -83,6 +83,8 @@ enum indic_category_t {
/* Visual positions in a syllable from left to right. */
enum indic_position_t {
+ POS_START,
+
POS_RA_TO_BECOME_REPH,
POS_PRE_M,
POS_PRE_C,
@@ -102,6 +104,7 @@ enum indic_position_t {
POS_FINAL_C,
POS_SMVD,
+
POS_END
};
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 36bf240..1a75e78 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -634,29 +634,32 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
}
}
- /* Attach ZWJ, ZWNJ, nukta, and halant to previous char to move with them. */
- if (!indic_options ().uniscribe_bug_compatible)
+ /* Attach misc marks to previous char to move with them. */
{
- /* Please update the Uniscribe branch when touching this! */
- for (unsigned int i = start + 1; i < end; i++)
- if ((FLAG (info[i].indic_category()) & (FLAG (OT_ZWNJ) | FLAG (OT_ZWJ) | FLAG (OT_N) | FLAG (OT_RS) | FLAG (OT_H))))
- info[i].indic_position() = info[i - 1].indic_position();
- } else {
- /*
- * Uniscribe doesn't move the Halant with Left Matra.
- * TEST: U+092B,U+093F,U+094DE
- */
- /* Please update the non-Uniscribe branch when touching this! */
- for (unsigned int i = start + 1; i < end; i++)
- if ((FLAG (info[i].indic_category()) & (FLAG (OT_ZWNJ) | FLAG (OT_ZWJ) | FLAG (OT_N) | FLAG (OT_RS) | FLAG (OT_H)))) {
- info[i].indic_position() = info[i - 1].indic_position();
- if (info[i].indic_category() == OT_H && info[i].indic_position() == POS_PRE_M)
+ indic_position_t last_pos = POS_START;
+ for (unsigned int i = start; i < end; i++)
+ {
+ if ((FLAG (info[i].indic_category()) & (JOINER_FLAGS | FLAG (OT_N) | FLAG (OT_RS) | HALANT_OR_COENG_FLAGS)))
+ {
+ info[i].indic_position() = last_pos;
+ if (unlikely (indic_options ().uniscribe_bug_compatible &&
+ info[i].indic_category() == OT_H &&
+ info[i].indic_position() == POS_PRE_M))
+ {
+ /*
+ * Uniscribe doesn't move the Halant with Left Matra.
+ * TEST: U+092B,U+093F,U+094DE
+ */
for (unsigned int j = i; j > start; j--)
if (info[j - 1].indic_position() != POS_PRE_M) {
info[i].indic_position() = info[j - 1].indic_position();
break;
}
+ }
+ } else if (info[i].indic_position() != POS_SMVD) {
+ last_pos = (indic_position_t) info[i].indic_position();
}
+ }
}
/* Re-attach ZWJ, ZWNJ, and halant to next char, for after-base consonants. */
{
@@ -666,7 +669,8 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
last_halant = i;
else if (is_consonant (info[i])) {
for (unsigned int j = last_halant; j < i; j++)
- info[j].indic_position() = info[i].indic_position();
+ if (info[j].indic_position() != POS_SMVD)
+ info[j].indic_position() = info[i].indic_position();
}
}
commit efb4ad735691837a52447bedc1a66a87d0d9af51
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 14:27:38 2012 -0400
Fix compiler warnings
If x is not constant, we cannot ASSERT_STATIC on it.
diff --git a/src/hb-private.hh b/src/hb-private.hh
index bdfd3f5..3f710ed 100644
--- a/src/hb-private.hh
+++ b/src/hb-private.hh
@@ -734,7 +734,7 @@ hb_in_range (T u, T lo, T hi)
* For example, for testing "x ∈ {x1, x2, x3}" use:
* (FLAG(x) & (FLAG(x1) | FLAG(x2) | FLAG(x3)))
*/
-#define FLAG(x) (ASSERT_STATIC_EXPR_ZERO((x) < 8 * sizeof(int)) + (1<<(x)))
+#define FLAG(x) (1<<(x))
#define FLAG_RANGE(x,y) (ASSERT_STATIC_EXPR_ZERO ((x) < (y)) + FLAG(y+1) - FLAG(x))
commit f31d97e44eeb6fb141f3de928e27e033fc7b1f47
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 14:13:35 2012 -0400
[Indic] Form Telugu Reph out of Ra,Virama,ZWJ
Apparently this was approved in Feb 2012. No font yet.
diff --git a/src/hb-ot-shape-complex-indic-private.hh b/src/hb-ot-shape-complex-indic-private.hh
index bbf5484..5f4856e 100644
--- a/src/hb-ot-shape-complex-indic-private.hh
+++ b/src/hb-ot-shape-complex-indic-private.hh
@@ -177,7 +177,7 @@ static const hb_codepoint_t ra_chars[] = {
0x0AB0, /* Gujarati */
0x0B30, /* Oriya */
0x0BB0, /* Tamil */ /* No Reph */
- 0x0C30, /* Telugu */ /* No Reph */
+ 0x0C30, /* Telugu */ /* Reph formed only with ZWJ */
0x0CB0, /* Kannada */
0x0D30, /* Malayalam */ /* No Reph, Logical Repha */
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index ffba986..36bf240 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -471,8 +471,8 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
start + 3 <= end &&
info[start].indic_category() == OT_Ra &&
info[start + 1].indic_category() == OT_H &&
- (unlikely (buffer->props.script == HB_SCRIPT_SINHALA) ?
- info[start + 2].indic_category() == OT_ZWJ /* In Sinhala, form Reph only if ZWJ is present */:
+ (unlikely (buffer->props.script == HB_SCRIPT_SINHALA || buffer->props.script == HB_SCRIPT_TELUGU) ?
+ info[start + 2].indic_category() == OT_ZWJ /* In Sinhala & Telugu, form Reph only if ZWJ is present */:
!is_joiner (info[start + 2] /* In other scripts, any joiner blocks Reph formation */ )
))
{
commit 2e193b240ec85cab0d4e2f8a375c5a7f0ef99985
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 14:02:35 2012 -0400
[Indic] Don't split U+0AC9
Althought IndicMatraCategory.txt classifies it as Top_And_Right matra,
it does not have Unicode decomposition, and Uniscribe does not do
anything special about it either.
Gujarati failures down from 0.672% to 0.0130966%.
diff --git a/src/hb-unicode.cc b/src/hb-unicode.cc
index 3e8f807..c527340 100644
--- a/src/hb-unicode.cc
+++ b/src/hb-unicode.cc
@@ -269,7 +269,7 @@ hb_unicode_decompose (hb_unicode_funcs_t *ufuncs,
{
/* XXX FIXME, move these to complex shapers and propagage to normalizer.*/
switch (ab) {
- case 0x0AC9 : *a = 0x0AC5; *b= 0x0ABE; return true;
+ case 0x0AC9 : return false;
case 0x0931 : return false;
case 0x0B94 : return false;
commit 30c3d5e9fc61b49c2c6ad4e744300edd6f3e0261
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 13:56:32 2012 -0400
[Indic] Simplify Uniscribe cluster emulation
Now that we break syllables on Halant,ZWNJ, this code can be simplified.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 7979e24..ffba986 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -1110,18 +1110,11 @@ final_reordering_syllable (hb_buffer_t *buffer,
if (indic_options ().uniscribe_bug_compatible)
{
- /* This is what Uniscribe does. Ie. add cluster boundaries after Halant,ZWNJ.
+ /* Uniscribe merges the entire cluster.
* This means, half forms are submerged into the main consonants cluster.
* This is unnecessary, and makes cursor positioning harder, but that's what
* Uniscribe does. */
- unsigned int cluster_start = start;
- for (unsigned int i = start + 1; i < start_of_last_cluster; i++)
- if (is_halant_or_coeng (info[i - 1]) && info[i].indic_category() == OT_ZWNJ) {
- i++;
- buffer->merge_clusters (cluster_start, i);
- cluster_start = i;
- }
- start_of_last_cluster = cluster_start;
+ start_of_last_cluster = start;
}
buffer->merge_clusters (start_of_last_cluster, end);
commit decf6ffca475fe01ff3151b7641f629f031137d2
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 13:51:31 2012 -0400
[Indic] Minor!
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index e771e57..7979e24 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -1108,7 +1108,7 @@ final_reordering_syllable (hb_buffer_t *buffer,
* Finish off the clusters and go home!
*/
- if (!indic_options ().uniscribe_bug_compatible)
+ if (indic_options ().uniscribe_bug_compatible)
{
/* This is what Uniscribe does. Ie. add cluster boundaries after Halant,ZWNJ.
* This means, half forms are submerged into the main consonants cluster.
commit 9e4f94a72cea6d65a6a7ba5a47db92e00dbfbb91
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 13:48:03 2012 -0400
[Indic] Break syllables at Halant,ZWNJ
That's really what Uniscribe does, and explains a lot of pecularities of
Halant,ZWNJ before the base.
Sent Telugu from 1% failures to 0.03%. Improved Kannada and Malayalam
slightly. Fixed half of Bengali, and did NOT break anything!
diff --git a/src/hb-ot-shape-complex-indic-machine.rl b/src/hb-ot-shape-complex-indic-machine.rl
index 4501773..4be7698 100644
--- a/src/hb-ot-shape-complex-indic-machine.rl
+++ b/src/hb-ot-shape-complex-indic-machine.rl
@@ -67,8 +67,9 @@ forced_rakar = ZWJ H ZWJ Ra;
matra_group = z*.M.N?.(H | forced_rakar)?;
syllable_tail = SM? (Coeng (cn|V))? (VD VD?)?;
place_holder = NBSP | DOTTEDCIRCLE;
-halant_group = (z?.h.z?);
-halant_or_matra_group = (halant_group | matra_group*);
+halant_group = (z?.h.ZWJ?);
+final_halant_group = halant_group | h.ZWNJ;
+halant_or_matra_group = (final_halant_group | matra_group*);
consonant_syllable = Repha? (cn.halant_group)* cn A? halant_or_matra_group? syllable_tail;
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 42a7e8d..e771e57 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -531,9 +531,8 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
}
else
{
- /* A ZWJ at the end of syllable, or any ZWJ/ZWNJ in other places, stop the base
- * search (to request explicit half or halant forms. */
- if (is_joiner (info[i]) && (i + 1 < end || info[i].indic_category() == OT_ZWJ))
+ /* A ZWJ stops the base search, and requests an explicit half form. */
+ if (info[i].indic_category() == OT_ZWJ)
break;
}
} while (i > limit);
commit 2c372b80f6befad69e216e3f218b38640b8cc044
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 13:37:48 2012 -0400
[Indic] Better check for applying 'init'
Specifically, don't apply 'init' if previous char is a joiner.
Fixes some more of Bengali.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index bbb881e..42a7e8d 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -1101,14 +1101,7 @@ final_reordering_syllable (hb_buffer_t *buffer,
if (info[start].indic_position () == POS_PRE_M &&
(!start ||
!(FLAG (_hb_glyph_info_get_general_category (&info[start - 1])) &
- (FLAG (HB_UNICODE_GENERAL_CATEGORY_LOWERCASE_LETTER) |
- FLAG (HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER) |
- FLAG (HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER) |
- FLAG (HB_UNICODE_GENERAL_CATEGORY_TITLECASE_LETTER) |
- FLAG (HB_UNICODE_GENERAL_CATEGORY_UPPERCASE_LETTER) |
- FLAG (HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK) |
- FLAG (HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK) |
- FLAG (HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK)))))
+ FLAG_RANGE (HB_UNICODE_GENERAL_CATEGORY_FORMAT, HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK))))
info[start].mask |= init_mask;
diff --git a/src/hb-private.hh b/src/hb-private.hh
index 7b778b7..bdfd3f5 100644
--- a/src/hb-private.hh
+++ b/src/hb-private.hh
@@ -734,7 +734,8 @@ hb_in_range (T u, T lo, T hi)
* For example, for testing "x ∈ {x1, x2, x3}" use:
* (FLAG(x) & (FLAG(x1) | FLAG(x2) | FLAG(x3)))
*/
-#define FLAG(x) (1<<(x))
+#define FLAG(x) (ASSERT_STATIC_EXPR_ZERO((x) < 8 * sizeof(int)) + (1<<(x)))
+#define FLAG_RANGE(x,y) (ASSERT_STATIC_EXPR_ZERO ((x) < (y)) + FLAG(y+1) - FLAG(x))
template <typename T, typename T2> inline void
commit 34a7440b7c6c6e53394ddbdbedaad57b23f85105
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 12:32:59 2012 -0400
[GPOS] Don't zero mark advances
Fixes more of Telugu, Kannada, and Oriya.
May break things (outside Indic...), but we cannot think of any font relying
on this immediately.
diff --git a/src/hb-ot-layout-gpos-table.hh b/src/hb-ot-layout-gpos-table.hh
index 94055b3..9eadbd6 100644
--- a/src/hb-ot-layout-gpos-table.hh
+++ b/src/hb-ot-layout-gpos-table.hh
@@ -1486,8 +1486,8 @@ fix_mark_attachment (hb_glyph_position_t *pos, unsigned int i, hb_direction_t di
unsigned int j = i - pos[i].attach_lookback();
- pos[i].x_advance = 0;
- pos[i].y_advance = 0;
+// pos[i].x_advance = 0;
+// pos[i].y_advance = 0;
pos[i].x_offset += pos[j].x_offset;
pos[i].y_offset += pos[j].y_offset;
commit 8ed248de77e5d2ed978e55c0ce1a11727bc9e34c
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 11:42:24 2012 -0400
[Indic] Minor
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index ae3af2c..bbb881e 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -1112,8 +1112,9 @@ final_reordering_syllable (hb_buffer_t *buffer,
info[start].mask |= init_mask;
-
- /* Finish off the clusters and go home! */
+ /*
+ * Finish off the clusters and go home!
+ */
if (!indic_options ().uniscribe_bug_compatible)
{
commit d0e68dbd0b9fc9a42c4280d01c8ffd9c5015d550
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 11:25:41 2012 -0400
[Indic] Implement reph positioning step 5
Not tuned, just copied from step 2. Fixes another 0.5% of Kannada
failures. 1% to go.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index ffae430..ae3af2c 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -999,7 +999,17 @@ final_reordering_syllable (hb_buffer_t *buffer,
*/
reph_step_5:
{
- /* XXX */
+ /* Copied from step 2. */
+ new_reph_pos = start + 1;
+ while (new_reph_pos < base && !is_halant_or_coeng (info[new_reph_pos]))
+ new_reph_pos++;
+
+ if (new_reph_pos < base && is_halant_or_coeng (info[new_reph_pos])) {
+ /* ->If ZWJ or ZWNJ are following this halant, position is moved after it. */
+ if (new_reph_pos + 1 < base && is_joiner (info[new_reph_pos + 1]))
+ new_reph_pos++;
+ goto reph_move;
+ }
}
/* 6. Otherwise, reorder reph to the end of the syllable.
commit a9e45c32e4a0d6da33c52f8427aa694e57f52eb9
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 11:04:15 2012 -0400
[Indic] Don't let ZWNJ at the end of syllable affect base search
Fixes a few Devanagari, half of remaining Kannada failures, quarter for
Telugu, and others slightly improved or unchanged.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index a3f20b1..ffae430 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -530,8 +530,12 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
base = i;
}
else
- if (is_joiner (info[i]))
+ {
+ /* A ZWJ at the end of syllable, or any ZWJ/ZWNJ in other places, stop the base
+ * search (to request explicit half or halant forms. */
+ if (is_joiner (info[i]) && (i + 1 < end || info[i].indic_category() == OT_ZWJ))
break;
+ }
} while (i > limit);
}
else
commit 20b68e699f73e6ce046c0ec143d40b3d6d48e06b
Author: Behdad Esfahbod <behdad at behdad.org>
Date: Fri Jul 20 10:47:46 2012 -0400
[Indic] Apply 'cjct' globally
Fixes 5 Devanagari failures, and no regressions.
diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc
index 42e0f70..a3f20b1 100644
--- a/src/hb-ot-shape-complex-indic.cc
+++ b/src/hb-ot-shape-complex-indic.cc
@@ -322,7 +322,7 @@ indic_basic_features[] =
{HB_TAG('a','b','v','f'), false},
{HB_TAG('p','s','t','f'), false},
{HB_TAG('c','f','a','r'), false},
- {HB_TAG('c','j','c','t'), false},
+ {HB_TAG('c','j','c','t'), true},
{HB_TAG('v','a','t','u'), true},
};
@@ -338,7 +338,7 @@ enum {
ABVF,
PSTF,
CFAR,
- CJCT,
+ _CJCT,
VATU
};
@@ -691,15 +691,15 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
info[i].mask |= basic_mask_array[RPHF];
/* Pre-base */
- mask = basic_mask_array[HALF] | basic_mask_array[CJCT];
+ mask = basic_mask_array[HALF];
for (unsigned int i = start; i < base; i++)
info[i].mask |= mask;
/* Base */
- mask = basic_mask_array[CJCT];
+ mask = 0;
if (base < end)
info[base].mask |= mask;
/* Post-base */
- mask = basic_mask_array[BLWF] | basic_mask_array[ABVF] | basic_mask_array[PSTF] | basic_mask_array[CJCT];
+ mask = basic_mask_array[BLWF] | basic_mask_array[ABVF] | basic_mask_array[PSTF];
for (unsigned int i = base + 1; i < end; i++)
info[i].mask |= mask;
}
@@ -737,7 +737,10 @@ initial_reordering_consonant_syllable (const hb_ot_map_t *map, hb_buffer_t *buff
do {
j--;
- info[j].mask &= ~basic_mask_array[CJCT];
+ /* A ZWJ disables CJCT, however, it's mere presence is enough
+ * to disable ligation. No explicit action needed. */
+
+ /* A ZWNJ disables HALF. */
if (non_joiner)
info[j].mask &= ~basic_mask_array[HALF];
More information about the HarfBuzz
mailing list