[Thai] Adjust SARA AM reordering to match Uniscribe

Adjust the list of marks before SARA AM that get the reordering
treatment.  Also adjust cluster formation to match Uniscribe.

With Wikipedia test data, now I see:

  - For Thai, with the Angsana New font from Win7, I see 54 failures out
    of over 4M tests  (0.00129107%).  Of the 54, two are legitimate
    reordering issues (fix coming soon), and the other 52 are simply
    Uniscribe using a zero-width space char instead of an unknown
    character for missing glyphs.  No idea why.  The missing-glyph
    sequences include one that is a Thai character followed by an Arabic
    Sokun.  Someone confused it with Nikhahit I assume!

  - For Lao, with the Dokchampa font from Win7, 33 tests fail out of
    54k (0.0615167%).  All seem to be insignificant mark positioning
    with two marks on a base.  Have to investigate.

src/hb-ot-shape-complex-misc.cc

@@ -121,19 +121,20 @@ _hb_ot_shape_complex_setup_masks_thai (hb_ot_map_t *map HB_UNUSED,
   /* The following is NOT specified in the MS OT Thai spec, however, it seems
    * to be what Uniscribe and other engines implement.  According to Eric Muller:
    *
-   * When you have a sara am, decompose it in nikhahit + sara a, *and* mode the
+   * When you have a SARA AM, decompose it in NIKHAHIT + SARA AA, *and* move the
-   * nihka hit backwards over any *tone* mark (0E48-0E4B).
+   * NIKHAHIT backwards over any tone mark (0E48-0E4B).
    *
    * <0E14, 0E4B, 0E33> -> <0E14, 0E4D, 0E4B, 0E32>
    *
-   * This reordering is legit only when the nikhahit comes from a sara am, not
+   * This reordering is legit only when the NIKHAHIT comes from a SARA AM, not
    * when it's there to start with. The string <0E14, 0E4B, 0E4D> is probably
-   * not what a u↪ser wanted, but the rendering is nevertheless nikhahit above
+   * not what a user wanted, but the rendering is nevertheless nikhahit above
    * chattawa.
    *
    * Same for Lao.
    */
 
+
   /*
    * Here are the characters of significance:
    *
@@ -142,9 +143,9 @@ _hb_ot_shape_complex_setup_masks_thai (hb_ot_map_t *map HB_UNUSED,
    * SARA AA:		U+0E32	U+0EB2
    * Nikhahit:		U+0E4D	U+0ECD
    *
-   * Tone marks:
+   * Testing shows that Uniscribe reorder the following marks:
-   * Thai:	<0E48..0E4B> CCC=107
+   * Thai:	<0E31..0E37,0E47..0E4E>
-   * Lao:	<0EC8..0ECB> CCC=122
+   * Lao:	<0EB1..0EB7,0EC7..0ECE>
    *
    * Note how the Lao versions are the same as Thai + 0x80.
    */
@@ -154,7 +155,7 @@ _hb_ot_shape_complex_setup_masks_thai (hb_ot_map_t *map HB_UNUSED,
 #define IS_SARA_AM(x) (((x) & ~0x0080) == 0x0E33)
 #define NIKHAHIT_FROM_SARA_AM(x) ((x) - 0xE33 + 0xE4D)
 #define SARA_AA_FROM_SARA_AM(x) ((x) - 1)
-#define IS_TONE_MARK(x) (((x) & ~0x0083) == 0x0E48)
+#define IS_TONE_MARK(x) (hb_in_ranges<hb_codepoint_t> ((x) & ~0x0080, 0x0E31, 0x0E37, 0x0E47, 0x0E4E))
 
   buffer->clear_output ();
   unsigned int count = buffer->len;
@@ -179,14 +180,23 @@ _hb_ot_shape_complex_setup_masks_thai (hb_ot_map_t *map HB_UNUSED,
     while (start > 0 && IS_TONE_MARK (buffer->out_info[start - 1].codepoint))
       start--;
 
-    /* Move Nikhahit (end-2) to the beginning */
+    if (start + 2 < end)
-    hb_glyph_info_t t = buffer->out_info[end - 2];
+    {
-    memmove (buffer->out_info + start + 1,
+      /* Move Nikhahit (end-2) to the beginning */
-	     buffer->out_info + start,
+      buffer->merge_out_clusters (start, end);
-	     sizeof (buffer->out_info[0]) * (end - start - 2));
+      hb_glyph_info_t t = buffer->out_info[end - 2];
-    buffer->out_info[start] = t;
+      memmove (buffer->out_info + start + 1,
-
+	       buffer->out_info + start,
-    buffer->merge_out_clusters (start, end);
+	       sizeof (buffer->out_info[0]) * (end - start - 2));
+      buffer->out_info[start] = t;
+    }
+    else
+    {
+      /* Since we decomposed, and NIKHAHIT is combining, merge clusters with the
+       * previous cluster. */
+      if (start)
+	buffer->merge_out_clusters (start - 1, end);
+    }
   }
   buffer->swap_buffers ();
 }

src/hb-private.hh

@@ -729,6 +729,12 @@ hb_in_range (T u, T lo, T hi)
     return lo <= u && u <= hi;
 }
 
+template <typename T> static inline bool
+hb_in_ranges (T u, T lo1, T hi1, T lo2, T hi2)
+{
+  return hb_in_range (u, lo1, hi1) || hb_in_range (u, lo2, hi2);
+}
+
 
 /* Useful for set-operations on small enums.
  * For example, for testing "x ∈ {x1, x2, x3}" use:

test/shaping/texts/in-tree/shaper-thai/MANIFEST

@@ -1 +1,2 @@
+script-lao
 script-thai

test/shaping/texts/in-tree/shaper-thai/script-lao/MANIFEST

@@ -0,0 +1 @@
+misc

test/shaping/texts/in-tree/shaper-thai/script-lao/misc/MANIFEST

@@ -0,0 +1 @@
+sara-am.txt

test/shaping/texts/in-tree/shaper-thai/script-lao/misc/sara-am.txt

@@ -0,0 +1,20 @@
+ດຳ
+ດ໋ຳ
+ດໍ໋າ
+ດ໋ໍາ
+ມັຳ
+ມິຳ
+ມີຳ
+ມຶຳ
+ມືຳ
+ມຸຳ
+ມູຳ
+ມ຺ຳ
+ມ໇ຳ
+ມ່ຳ
+ມ້ຳ
+ມ໊ຳ
+ມ໋ຳ
+ມ໌ຳ
+ມໍຳ
+ມ໎ຳ

test/shaping/texts/in-tree/shaper-thai/script-thai/misc/sara-am.txt

@@ -1,4 +1,20 @@
-ำ
 ดำ
 ด๋ำ
 ดํ๋า
+ด๋ํา
+มัำ
+มิำ
+มีำ
+มึำ
+มืำ
+มุำ
+มูำ
+มฺำ
+ม็ำ
+ม่ำ
+ม้ำ
+ม๊ำ
+ม๋ำ
+ม์ำ
+มํำ
+ม๎ำ