[pango] Bug 664125 - Zero-width spaces cause missing characters

[Kristian Rietveld <kristian src gnome org>]

commit dbf40154eb5804f4e8c582f12b30b8291c9c3532
Author: Kristian Rietveld <kris lanedo com>
Date:   Sat Nov 26 17:27:43 2011 +0100

    Bug 664125 - Zero-width spaces cause missing characters
    
    This is fixed by using the characters obtained from CoreText's mapping from
    glyphs back to the original string instead of blindly iterating over the
    original gchar string. Take notice that we have to convert these indices
    from the mapping back to a byte index in the original UTF8 string.
    
    Zero-width spaces are already removed from the list of CGGlyphs, so we were
    mismatching CGGlyphs and characters from the original string. Using
    CoreText's list of string indices instead eliminates the mismatches.

 modules/basic/basic-coretext.c |   80 ++++++++++++++++++---------------------
 1 files changed, 37 insertions(+), 43 deletions(-)
---
diff --git a/modules/basic/basic-coretext.c b/modules/basic/basic-coretext.c
index 8dc3dba..991743e 100644
--- a/modules/basic/basic-coretext.c
+++ b/modules/basic/basic-coretext.c
@@ -76,7 +76,6 @@ basic_engine_shape (PangoEngineShape    *engine,
 		    const PangoAnalysis *analysis,
 		    PangoGlyphString    *glyphs)
 {
-  const char *p;
   char *copy;
   CTLineRef line;
   CFStringRef cstr;
@@ -88,6 +87,7 @@ basic_engine_shape (PangoEngineShape    *engine,
   CTRunRef run;
   CTRunStatus run_status;
   CFIndex i, glyph_count;
+  CFIndex *indices = NULL;
   const CGGlyph *cgglyphs;
 
   CFTypeRef keys[] = {
@@ -128,7 +128,9 @@ basic_engine_shape (PangoEngineShape    *engine,
   glyph_count = CTRunGetGlyphCount (run);
   cgglyphs = CTRunGetGlyphsPtr (run);
 
-  p = text;
+  indices = malloc (sizeof (CFIndex *) * glyph_count);
+  CTRunGetStringIndices (run, CFRangeMake (0, glyph_count), indices);
+
   pango_glyph_string_set_size (glyphs, glyph_count);
   coverage = pango_font_get_coverage (PANGO_FONT (cfont),
                                       analysis->language);
@@ -138,12 +140,7 @@ basic_engine_shape (PangoEngineShape    *engine,
       CFIndex real_i, prev_i;
       gunichar wc;
       gunichar mirrored_ch;
-
-      wc = g_utf8_get_char (p);
-
-      if (analysis->level % 2)
-	if (pango_get_mirror_char (wc, &mirrored_ch))
-	  wc = mirrored_ch;
+      PangoCoverageLevel result;
 
       if (run_status & kCTRunStatusRightToLeft)
         {
@@ -156,53 +153,50 @@ basic_engine_shape (PangoEngineShape    *engine,
           prev_i = real_i - 1;
         }
 
+      wc = CFStringGetCharacterAtIndex (cstr, indices[real_i]);
+
+      if (analysis->level % 2)
+	if (pango_get_mirror_char (wc, &mirrored_ch))
+	  wc = mirrored_ch;
+
       if (wc == 0xa0)	/* non-break-space */
 	wc = 0x20;
 
-      if (pango_is_zero_width (wc))
-	{
-	  set_glyph (font, glyphs, real_i, p - text, PANGO_GLYPH_EMPTY);
-	}
-      else
-	{
-          PangoCoverageLevel result;
+      result = pango_coverage_get (coverage, wc);
 
-          result = pango_coverage_get (coverage, wc);
+      if (result != PANGO_COVERAGE_NONE)
+        {
+          set_glyph (font, glyphs, real_i,
+                     g_utf8_offset_to_pointer (text, indices[real_i]) - text,
+                     cgglyphs[real_i]);
 
-          if (result != PANGO_COVERAGE_NONE)
+          if (g_unichar_type (wc) == G_UNICODE_NON_SPACING_MARK)
             {
-              set_glyph (font, glyphs, real_i, p - text, cgglyphs[real_i]);
-
-              if (g_unichar_type (wc) == G_UNICODE_NON_SPACING_MARK)
+              if (i > 0)
                 {
-                  if (i > 0)
-                    {
-                      PangoRectangle logical_rect, ink_rect;
-
-                      glyphs->glyphs[real_i].geometry.width = MAX (glyphs->glyphs[prev_i].geometry.width,
-                                                                   glyphs->glyphs[prev_i].geometry.width);
-                      glyphs->glyphs[prev_i].geometry.width = 0;
-                      glyphs->log_clusters[real_i] = glyphs->log_clusters[prev_i];
-
-                      /* Some heuristics to try to guess how overstrike glyphs are
-                       * done and compensate
-                       */
-                      pango_font_get_glyph_extents (font, glyphs->glyphs[real_i].glyph, &ink_rect, &logical_rect);
-                      if (logical_rect.width == 0 && ink_rect.x == 0)
-                        glyphs->glyphs[real_i].geometry.x_offset = (glyphs->glyphs[real_i].geometry.width - ink_rect.width) / 2;
-                    }
+                  PangoRectangle logical_rect, ink_rect;
+
+                  glyphs->glyphs[real_i].geometry.width = MAX (glyphs->glyphs[prev_i].geometry.width,
+                                                               glyphs->glyphs[real_i].geometry.width);
+                  glyphs->glyphs[prev_i].geometry.width = 0;
+                  glyphs->log_clusters[real_i] = glyphs->log_clusters[prev_i];
+
+                  /* Some heuristics to try to guess how overstrike glyphs are
+                   * done and compensate
+                   */
+                  pango_font_get_glyph_extents (font, glyphs->glyphs[real_i].glyph, &ink_rect, &logical_rect);
+                  if (logical_rect.width == 0 && ink_rect.x == 0)
+                    glyphs->glyphs[real_i].geometry.x_offset = (glyphs->glyphs[real_i].geometry.width - ink_rect.width) / 2;
                 }
             }
-          else
-            {
-              set_glyph (font, glyphs, real_i, p - text,
-                         PANGO_GET_UNKNOWN_GLYPH (wc));
-            }
         }
-
-      p = g_utf8_next_char (p);
+      else
+        set_glyph (font, glyphs, real_i,
+                   g_utf8_offset_to_pointer (text, indices[real_i]) - text,
+                   PANGO_GET_UNKNOWN_GLYPH (wc));
     }
 
+  free (indices);
   CFRelease (line);
   CFRelease (attstr);
   CFRelease (cstr);