[webkit-reviews] review denied: [Bug 90951] [Chromium] Fix bugs in HarfBuzzShaper : [Attachment 151644] Patch

[bugzilla-daemon at webkit.org]

Tony Chang <tony at chromium.org> has denied Kenichi Ishibashi
<bashi at chromium.org>'s request for review:
Bug 90951: [Chromium] Fix bugs in HarfBuzzShaper
https://bugs.webkit.org/show_bug.cgi?id=90951

Attachment 151644: Patch
https://bugs.webkit.org/attachment.cgi?id=151644&action=review

------- Additional Comments from Tony Chang <tony at chromium.org>
View in context: https://bugs.webkit.org/attachment.cgi?id=151644&action=review


r- for lack of tests.

> Source/WebCore/ChangeLog:14
> +	   - Cannot render RTL text if the TextRun is divided into more than
two
> +	     HarfBuzzRun.
> +	   - Script handling in TextRun partitioning is incorrect.
> +	   - Inaccurate calculation of selection rect.
> +	   - Wrong rendering position when the first glyph of the TextRun have
> +	     non-zero offsets in terms of HarfBuzz.

Why are there no tests?  Doesn't this code already run on Chromium Mac?  We
should have a test showing that each of these are fixed.

> Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:230
> +    return m_harfbuzzRuns.size() > 0;

!m_harfbuzzRuns.isEmpty()

> Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:256
> +	       hb_buffer_destroy(harfbuzzBuffer);

Should we make a scoped pointer type that calls hb_buffer_destroy in the
destructor?  Maybe for hb_font_t* too?

> Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.cpp:260
> +	   hb_shape(harfbuzzFont, harfbuzzBuffer, m_features.size() > 0 ?
m_features.data() : 0, m_features.size());

Nit: m_features.isEmpty()

> Source/WebCore/platform/graphics/harfbuzz/ng/HarfBuzzShaper.h:111
> +    float m_startOffsetX;
> +    float m_startOffsetY;

Maybe use a FloatPoint here instead?