From 7afb14407e59dfeaa79c33aca1ffa60e7982e349 Mon Sep 17 00:00:00 2001
From: Behdad Esfahbod <behdad@behdad.org>
Date: Mon, 30 Jul 2012 13:54:46 -0400
Subject: [PATCH] [Indic] Recategorize Telugu length marks

Fixes 8 more Telugu tests.  Failures at 15 (0.00154548%).
---
 src/hb-unicode.cc |  8 ++++++++
 src/indic.cc      | 11 ++++++++---
 2 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/src/hb-unicode.cc b/src/hb-unicode.cc
index 3569b205..b05b290c 100644
--- a/src/hb-unicode.cc
+++ b/src/hb-unicode.cc
@@ -369,6 +369,14 @@ _hb_unicode_modified_combining_class (hb_unicode_funcs_t *ufuncs,
      * Uniscribe does this. */
     c = 104;
   }
+  else if (unlikely (hb_in_range<hb_codepoint_t> (unicode, 0x0C55, 0x0C56)))
+  {
+    /* Telugu length marks.
+     * These are the only matras in the main Indic script range that have
+     * a non-zero ccc.  That makes them reorder with the Halant that is
+     * ccc=9.  Just zero them, we don't need them in our Indic shaper. */
+    c = 0;
+  }
 
   return c;
 }
diff --git a/src/indic.cc b/src/indic.cc
index 3b440761..991a7723 100644
--- a/src/indic.cc
+++ b/src/indic.cc
@@ -39,8 +39,13 @@ main (void)
     unsigned int category = type & 0x0F;
     unsigned int position = type >> 4;
 
-    hb_codepoint_t a, b;
-    if (!hb_unicode_decompose (funcs, u, &a, &b))
-      printf ("U+%04X %x %x\n", u, category, position);
+    hb_unicode_general_category_t cat = hb_unicode_general_category (funcs, u);
+    unsigned int ccc = hb_unicode_combining_class (funcs, u);
+    if (category == OT_M && ccc)
+      printf ("U+%04X %d\n", u, ccc);
+
+//    hb_codepoint_t a, b;
+//    if (!hb_unicode_decompose (funcs, u, &a, &b))
+//      printf ("U+%04X %x %x\n", u, category, position);
   }
 }
-- 
GitLab