Bug 233893 – [GPU Process] [Filters] Make FilterEffect::externalRepresentation() and its overridable functions work iteratively

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Said Abou-Hallawa]

This is a step towards removing the inputEffect() from FilterEffect.

[Said Abou-Hallawa]

Created attachment 446070 [details]
Patch

[Said Abou-Hallawa]

Created attachment 446109 [details]
Patch

[Cameron McCormack (:heycam)]

Comment on attachment 446109 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=446109&action=review

> Source/WebCore/platform/graphics/filters/FEBlend.cpp:63
> +    {

No need for the block (and in all other externalRepresentation functions that don't use an IndentScope).

> Source/WebCore/rendering/CSSFilter.cpp:419
> +        TextStream::IndentScope indentScope(ts, level++);

Is there a need to indent? I don't think it provides any information.

> Source/WebCore/svg/graphics/filters/SVGFilter.cpp:140
> +    auto end = m_expression.rend();
> +
> +    for (auto it = m_expression.rbegin(); it != end; ++it) {

Probably more idiomatic to write this as:

for (auto it = m_expression.rbegin(), end = m_expression.rend(); it != end; ++it)

> Source/WebCore/svg/graphics/filters/SVGFilter.cpp:147
> +        TextStream::IndentScope indentScope(ts, term.level);

Similarly, is it useful to represent the depth in the filter DAG as the indent level? What makes me doubt this is that it doesn't fully encode the dependencies between the nodes in the filter graph, e.g. if two filter effects depend on the one other filter effect.

> Source/WebCore/svg/graphics/filters/SVGFilterBuilder.cpp:220
> +    expression.append({ *effect, effectGeometry(*effect), level });

I don't think we should bother computing and storing the level. Unless it's needed for something other than externalRepresentation.

[Said Abou-Hallawa]

Created attachment 446123 [details]
Patch

[Said Abou-Hallawa]

Comment on attachment 446109 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=446109&action=review

>> Source/WebCore/platform/graphics/filters/FEBlend.cpp:63
>> +    {
> 
> No need for the block (and in all other externalRepresentation functions that don't use an IndentScope).

The block was removed.

>> Source/WebCore/rendering/CSSFilter.cpp:419
>> +        TextStream::IndentScope indentScope(ts, level++);
> 
> Is there a need to indent? I don't think it provides any information.

Removing the indentation will require rebase-line for many layout tests. We can remove the indentation in a future patch.

>> Source/WebCore/svg/graphics/filters/SVGFilter.cpp:140
>> +    for (auto it = m_expression.rbegin(); it != end; ++it) {
> 
> Probably more idiomatic to write this as:
> 
> for (auto it = m_expression.rbegin(), end = m_expression.rend(); it != end; ++it)

Fixed.

[EWS]

Committed r286589 (?): <https://commits.webkit.org/r286589>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 446123 [details].

[Radar WebKit Bug Importer]

<rdar://problem/86141818>