Fix fuzzer-found issues (Part 1)

Cargo.toml

@@ -23,7 +23,8 @@ unicode-properties = { version = "0.1.0", default-features = false, features = [
 unicode-script = "0.5.2"
 
 [dependencies.ttf-parser]
-version = "0.24.0"
+git = "https://github.com/RazrFalcon/ttf-parser"
+rev = "85c1ff2"
 default-features = false
 features = [
     "opentype-layout",

scripts/gen-indic-table.py

@@ -319,7 +319,7 @@
   0x109B: 'SM',
   0x109C: 'SM',
 
-  0x104B: 'PLACEHOLDER',
+  0x104A: 'PLACEHOLDER',
 }
 
 position_overrides = {

scripts/gen-shaping-tests.py

@@ -36,8 +36,6 @@
     # Resource exhaustion tests with large outputs
     "morx_34_001",
     "morx_36_001",
-    # ttf-parser uses different rounding, not a bug
-    "fallback_positioning_001",
     # Requires support for the ltag table.
     "macos_002",
     # Custom MacOS test. A shortened version of `macos_013`, but with `--show-flags`.

scripts/gen-unicode-norm-table.py

@@ -6,6 +6,9 @@
 URL = 'https://www.unicode.org/Public/UCD/latest/ucd/UnicodeData.txt'
 FILE_NAME = 'UnicodeData.txt'
 
+EXCLUSIONS_URL = 'https://www.unicode.org/Public/draft/UCD/ucd/CompositionExclusions.txt'
+EXCLUSIONS_FILE_NAME = 'CompositionExclusions.txt'
+
 
 def hex_to_char_rs(c):
     return f"'\\u{{{c}}}'"
@@ -14,6 +17,20 @@ def hex_to_char_rs(c):
 if not os.path.exists(FILE_NAME):
     urllib.request.urlretrieve(URL, FILE_NAME)
 
+if not os.path.exists(EXCLUSIONS_FILE_NAME):
+    urllib.request.urlretrieve(EXCLUSIONS_URL, EXCLUSIONS_FILE_NAME)
+
+exclusions = set()
+
+with open(EXCLUSIONS_FILE_NAME) as f:
+    import re
+    pattern = re.compile(r'^([0-9a-fA-F]+)\s+#\s+(.*)$')
+
+    for line in f:
+        match = pattern.match(line.strip())
+        if match:
+            exclusions.add(match.group(1))
+
 
 print('// WARNING: this file was generated by ../scripts/gen-unicode-norm-table.py')
 print()
@@ -76,6 +93,7 @@ def hex_to_char_rs(c):
 assert len(pairs) == len(needles)
 
 for pair in pairs:
-    print(f'    ({pair[0]}, {hex_to_char_rs(pair[1])}),')
+    if pair[1] not in exclusions:
+        print(f'    ({pair[0]}, {hex_to_char_rs(pair[1])}),')
 
 print('];')

src/hb/buffer.rs

@@ -668,6 +668,7 @@ impl hb_buffer_t {
             let pos: Vec<hb_glyph_info_t> = bytemuck::cast_vec(core::mem::take(&mut self.pos));
             self.pos = info;
             self.info = pos;
+            self.have_separate_output = false;
         }
 
         self.len = self.out_len;

src/hb/face.rs

@@ -313,7 +313,11 @@ impl<'a> hb_font_t<'a> {
             return ret;
         }
 
-        let bbox = self.ttfp_face.glyph_bounding_box(glyph);
+        let mut bbox = None;
+
+        if let Some(glyf) = self.ttfp_face.tables().glyf {
+            bbox = glyf.bbox(glyph);
+        }
 
         // See https://github.com/RazrFalcon/rustybuzz/pull/98#issuecomment-1948430785
         if self.ttfp_face.tables().glyf.is_some() && bbox.is_none() {

src/hb/ot_layout_gsubgpos.rs

@@ -1182,16 +1182,13 @@ pub mod OT {
 
             if has_glyph_classes {
                 props &= GlyphPropsFlags::PRESERVE.bits();
-                props =
-                    (props & !GlyphPropsFlags::CLASS_MASK.bits()) | self.face.glyph_props(glyph_id);
+                cur.set_glyph_props(props | self.face.glyph_props(glyph_id));
             } else if !class_guess.is_empty() {
                 props &= GlyphPropsFlags::PRESERVE.bits();
-                props = (props & !GlyphPropsFlags::CLASS_MASK.bits()) | class_guess.bits();
+                cur.set_glyph_props(props | class_guess.bits());
             } else {
-                props = props & !GlyphPropsFlags::CLASS_MASK.bits();
+                cur.set_glyph_props(props);
             }
-
-            cur.set_glyph_props(props);
         }
 
         pub fn replace_glyph(&mut self, glyph_id: GlyphId) {

src/hb/ot_shape.rs

@@ -805,6 +805,10 @@ fn map_glyphs_fast(buffer: &mut hb_buffer_t) {
     for info in &mut buffer.info[..len] {
         info.glyph_id = info.glyph_index();
     }
+
+    for info in &mut buffer.out_info_mut()[..len] {
+        info.glyph_id = info.glyph_index();
+    }
 }
 
 fn hb_synthesize_glyph_classes(buffer: &mut hb_buffer_t) {

src/hb/ot_shaper_indic_table.rs

@@ -21,7 +21,7 @@ use OT_MR    as _OT_MR;         /*   1 chars; MR */
 use OT_MW    as _OT_MW;         /*   2 chars; MW */
 use OT_MY    as _OT_MY;         /*   3 chars; MY */
 use OT_N     as _OT_N;          /*  17 chars; N */
-use OT_PLACEHOLDER as _OT_GB;         /* 164 chars; PLACEHOLDER */
+use OT_PLACEHOLDER as _OT_GB;         /* 165 chars; PLACEHOLDER */
 use OT_PT    as _OT_PT;         /*   8 chars; PT */
 use OT_Ra    as _OT_R;          /*  14 chars; Ra */
 use OT_Repha as _OT_Rf;         /*   1 chars; Repha */
@@ -44,7 +44,7 @@ use POS_ABOVE_C as _POS_T;         /*  22 chars; ABOVE_C */
 use POS_AFTER_MAIN as _POS_A;         /*   3 chars; AFTER_MAIN */
 use POS_AFTER_POST as _POS_AP;        /*  50 chars; AFTER_POST */
 use POS_AFTER_SUB as _POS_AS;        /*  51 chars; AFTER_SUB */
-use POS_BASE_C as _POS_C;         /* 832 chars; BASE_C */
+use POS_BASE_C as _POS_C;         /* 833 chars; BASE_C */
 use POS_BEFORE_SUB as _POS_BS;        /*  25 chars; BEFORE_SUB */
 use POS_BELOW_C as _POS_B;         /*  13 chars; BELOW_C */
 use POS_END  as _POS_X;         /*  71 chars; END */
@@ -257,7 +257,7 @@ const TABLE: &[(SyllabicCategory, MatraCategory)] = &[
   /* 1030 */(_OT_VB,_POS_B),(_OT_VL,_POS_L),(_OT_A,_POS_SM),(_OT_VA,_POS_T),(_OT_VA,_POS_T),(_OT_VA,_POS_T),(_OT_A,_POS_SM), (_OT_N,_POS_X),
   /* 1038 */(_OT_SM,_POS_SM), (_OT_H,_POS_X),(_OT_As,_POS_X),(_OT_MY,_POS_X),(_OT_MR,_POS_X),(_OT_MW,_POS_X),(_OT_MH,_POS_X), (_OT_C,_POS_C),
   /* 1040 */(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C),
-  /* 1048 */(_OT_GB,_POS_C),(_OT_GB,_POS_C), (_OT_X,_POS_X),(_OT_GB,_POS_C), (_OT_X,_POS_X), (_OT_X,_POS_X), (_OT_C,_POS_C), (_OT_X,_POS_X),
+  /* 1048 */(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C),(_OT_GB,_POS_C), (_OT_X,_POS_X), (_OT_X,_POS_X), (_OT_C,_POS_C), (_OT_X,_POS_X),
   /* 1050 */ (_OT_C,_POS_C), (_OT_C,_POS_C), (_OT_V,_POS_C), (_OT_V,_POS_C), (_OT_V,_POS_C), (_OT_V,_POS_C),(_OT_VR,_POS_R),(_OT_VR,_POS_R),
   /* 1058 */(_OT_VB,_POS_B),(_OT_VB,_POS_B), (_OT_R,_POS_C), (_OT_C,_POS_C), (_OT_C,_POS_C), (_OT_C,_POS_C),(_OT_MY,_POS_X),(_OT_MY,_POS_X),
   /* 1060 */(_OT_ML,_POS_X), (_OT_C,_POS_C),(_OT_VR,_POS_R),(_OT_PT,_POS_X),(_OT_PT,_POS_X), (_OT_C,_POS_C), (_OT_C,_POS_C),(_OT_VR,_POS_R),