Bug 143925 – CSSParser::parseValue() copies the m_parsedProperties vector at addParsedProperties()

• Status: RESOLVED
• Resolution: FIXED
• Component: CSS

[Simon Fraser (smfr)]

MutableStyleProperties::addParsedProperties() takes a const Vector<CSSProperty>& but m_parsedProperties is a Vector<CSSProperty, 256>, and apparently this forces a copy.

[Chris Dumez]

Very nice find :) We really supposed to use ParsedPropertyVector type here.

[Chris Dumez]

Created attachment 251171 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 251171 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=251171&action=review

> Source/WebCore/css/CSSParser.h:84
> +    using ParsedPropertyVector = Vector<CSSProperty, 256>;

Does the compiler enforce a match between this and the typedef?

> Source/WebCore/css/StyleProperties.cpp:770
> +    for (auto& property : properties)

const auto&?

[Chris Dumez]

Comment on attachment 251171 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=251171&action=review

>> Source/WebCore/css/CSSParser.h:84
>> +    using ParsedPropertyVector = Vector<CSSProperty, 256>;
> 
> Does the compiler enforce a match between this and the typedef?

Which typedef? I just made the typedef public so it could be used in StyleProperties. I used a 'using' statement but it is equivalent to a typedef in this case.

>> Source/WebCore/css/StyleProperties.cpp:770
>> +    for (auto& property : properties)
> 
> const auto&?

auto will be a const CSSProperty so it is equivalent but OK.

[Chris Dumez]

Created attachment 251179 [details]
Patch

[Chris Dumez]

To avoid mistakes like this in the future, I think we should make the following constructor explicit:
template<size_t otherCapacity, typename otherOverflowBehaviour>
Vector(const Vector<T, otherCapacity, otherOverflowBehaviour>&);

I can look into this in a follow-up patch as it likely is a larger change.

[Chris Dumez]

Comment on attachment 251179 [details]
Patch

Clearing flags on attachment: 251179

Committed r183024: <http://trac.webkit.org/changeset/183024>

[Chris Dumez]

All reviewed patches have been landed.  Closing bug.

[Chris Dumez]

(In reply to comment #6)
> To avoid mistakes like this in the future, I think we should make the
> following constructor explicit:
> template<size_t otherCapacity, typename otherOverflowBehaviour>
> Vector(const Vector<T, otherCapacity, otherOverflowBehaviour>&);
> 
> I can look into this in a follow-up patch as it likely is a larger change.

-> https://bugs.webkit.org/show_bug.cgi?id=143970