Changeset 220797 in webkit

Timestamp:

Aug 16, 2017 9:36:35 AM (7 years ago)

Author:

commit-queue@webkit.org

Message:

[HarfBuzz] Decomposed Vietnamese characters are rendered incorrectly
​https://bugs.webkit.org/show_bug.cgi?id=174418

Patch by Fujii Hironori <Fujii Hironori> on 2017-08-16
Reviewed by Michael Catanzaro.

Source/WebCore:

HarfBuzzShaper should normalize the input text before collecting
HarfBuzzRuns. Actually, HarfBuzzShaper::setNormalizedBuffer does
the task. But, this function hasn't been called from anywhere
since Bug 90951.

Test: fast/text/international/vietnamese-nfd.html

imported/blink/fast/text/international/text-shaping-arabic-diffs.html

• platform/graphics/harfbuzz/HarfBuzzShaper.cpp:

(WebCore::HarfBuzzShaper::HarfBuzzShaper):
Call setNormalizedBuffer instead of normalizeCharacters.
(WebCore::normalizeCharacters): Deleted.
(WebCore::normalizeSpacesAndMirrorChars) Use
FontCascade::treatAsZeroWidthSpaceInComplexScript instead of
FontCascade::treatAsZeroWidthSpace to preserve ZWJ and ZWNJ.

LayoutTests:

• fast/text/international/vietnamese-nfd-expected.html: Added.
• fast/text/international/vietnamese-nfd.html: Added.

Location:

trunk

Files:

• LayoutTests/ChangeLog (modified) (1 diff)
• LayoutTests/fast/text/international/vietnamese-nfd-expected.html (added)
• LayoutTests/fast/text/international/vietnamese-nfd.html (added)
• Source/WebCore/ChangeLog (modified) (1 diff)
• Source/WebCore/platform/graphics/harfbuzz/HarfBuzzShaper.cpp (modified) (3 diffs)

trunk/LayoutTests/ChangeLog

@@ +1 @@
+2017-08-16  Fujii Hironori  <Hironori.Fujii@sony.com>
+
+        [HarfBuzz] Decomposed Vietnamese characters are rendered incorrectly
+        https://bugs.webkit.org/show_bug.cgi?id=174418
+
+        Reviewed by Michael Catanzaro.
+
+        * fast/text/international/vietnamese-nfd-expected.html: Added.
+        * fast/text/international/vietnamese-nfd.html: Added.
+
 2017-08-16  Chris Dumez  <cdumez@apple.com>
 

trunk/Source/WebCore/ChangeLog

@@ +1 @@
+2017-08-16  Fujii Hironori  <Hironori.Fujii@sony.com>
+
+        [HarfBuzz] Decomposed Vietnamese characters are rendered incorrectly
+        https://bugs.webkit.org/show_bug.cgi?id=174418
+
+        Reviewed by Michael Catanzaro.
+
+        HarfBuzzShaper should normalize the input text before collecting
+        HarfBuzzRuns. Actually, HarfBuzzShaper::setNormalizedBuffer does
+        the task. But, this function hasn't been called from anywhere
+        since Bug 90951.
+
+        Test: fast/text/international/vietnamese-nfd.html
+              imported/blink/fast/text/international/text-shaping-arabic-diffs.html
+
+        * platform/graphics/harfbuzz/HarfBuzzShaper.cpp:
+        (WebCore::HarfBuzzShaper::HarfBuzzShaper):
+        Call setNormalizedBuffer instead of normalizeCharacters.
+        (WebCore::normalizeCharacters): Deleted.
+        (WebCore::normalizeSpacesAndMirrorChars) Use
+        FontCascade::treatAsZeroWidthSpaceInComplexScript instead of
+        FontCascade::treatAsZeroWidthSpace to preserve ZWJ and ZWNJ.
+
 2017-08-16  Antti Koivisto  <antti@apple.com>
 

trunk/Source/WebCore/platform/graphics/harfbuzz/HarfBuzzShaper.cpp

@@ -161 +161 @@
 }
 
-static void normalizeCharacters(const TextRun& run, UChar* destination, unsigned length)
-{
-    unsigned position = 0;
-    bool error = false;
-    const UChar* source;
-    String stringFor8BitRun;
-    if (run.is8Bit()) {
-        stringFor8BitRun = String::make16BitFrom8BitSource(run.characters8(), run.length());
-        source = stringFor8BitRun.characters16();
-    } else
-        source = run.characters16();
-
-    while (position < length) {
-        UChar32 character;
-        unsigned nextPosition = position;
-        U16_NEXT(source, nextPosition, length, character);
-        // Don't normalize tabs as they are not treated as spaces for word-end.
-        if (FontCascade::treatAsSpace(character) && character != '\t')
-            character = ' ';
-        else if (FontCascade::treatAsZeroWidthSpaceInComplexScript(character))
-            character = zeroWidthSpace;
-        U16_APPEND(destination, position, length, character, error);
-        ASSERT_UNUSED(error, !error);
-        position = nextPosition;
-    }
-}
-
 HarfBuzzShaper::HarfBuzzShaper(const FontCascade* font, const TextRun& run)
     : m_font(font)
@@ -198 +171 @@
     , m_letterSpacing(font->letterSpacing())
 {
-    m_normalizedBuffer = std::make_unique<UChar[]>(m_run.length() + 1);
-    m_normalizedBufferLength = m_run.length();
-    normalizeCharacters(m_run, m_normalizedBuffer.get(), m_normalizedBufferLength);
+    setNormalizedBuffer();
     setPadding(m_run.expansion());
     setFontFeatures();
@@ -221 +192 @@
         if (FontCascade::treatAsSpace(character) && character != '\t')
             character = ' ';
-        else if (FontCascade::treatAsZeroWidthSpace(character))
+        else if (FontCascade::treatAsZeroWidthSpaceInComplexScript(character))
             character = zeroWidthSpace;
         else if (normalizeMode == HarfBuzzShaper::NormalizeMirrorChars)