Bug 121656 – Avoid calling AtomicString::lower() in makeFontGlyphsCacheKey

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Anders Carlsson]

Avoid calling AtomicString::lower() in makeFontGlyphsCacheKey

[Anders Carlsson]

Created attachment 212122 [details]
Patch

[Darin Adler]

Comment on attachment 212122 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=212122&action=review

> Source/WebCore/platform/graphics/Font.cpp:171
> +        if (!equalIgnoringCase(a.families[i].impl(), b.families[i].impl()))

Why is the impl() needed here? Shouldn't we have functions overloaded so that isn’t needed?

> Source/WebCore/platform/graphics/Font.cpp:206
> +    Vector<unsigned, 7> hashCodes;
> +    hashCodes.reserveInitialCapacity(4 + key.families.size());

Need a comment explaining why 7. Why not, say, more like 16 to cover virtually all non-ridiculous cases without allocation?

> Source/WebCore/platform/graphics/Font.cpp:215
> +    return StringHasher::hashMemory(hashCodes.data(), hashCodes.size() * sizeof(unsigned));

The proper comment for this line is, "Yo dawg! I heard you like hashes, so I hashed your hash."

[Ryosuke Niwa]

Comment on attachment 212122 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=212122&action=review

> Source/WebCore/platform/graphics/Font.cpp:170
> +    for (unsigned i = 0; i < a.families.size(); ++i) {

Why don't we cache the size in a local variable or traverse backwards?

[Anders Carlsson]

Committed r156139: <http://trac.webkit.org/changeset/156139>

[Csaba Osztrogonác]

(In reply to comment #4)
> Committed r156139: <http://trac.webkit.org/changeset/156139>
FYI:It made all test crash everywhere.

[Csaba Osztrogonác]

Fix landed in http://trac.webkit.org/changeset/156142. Thanks.