Bug 105851 – Web Inspector: show raster tasks on Timeline

• Status: RESOLVED
• Resolution: FIXED
• Component: Web Inspector (Deprecated)

[Andrey Kosyakov]

Add processing of trace events emitted during paint and deferred paint of a layer and display deferred paint (and potential nested events, such as image decode/resize) on the timeline.

[Andrey Kosyakov]

Created attachment 180926 [details]
Patch

[Pavel Feldman]

Comment on attachment 180926 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=180926&action=review

> Source/WebCore/inspector/InspectorTimelineAgent.cpp:281
> +        m_traceEventProcessor->willPaint();

It would be great if there was unidirectional dependency in the instrumentation (i.e. processor -> timeline).

[Andrey Kosyakov]

Created attachment 180938 [details]
Patch

[Pavel Feldman]

Comment on attachment 180938 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=180938&action=review

r- for introducing page pointer and not using constants. Otherwise looks cryptic and good.

> Source/WebCore/inspector/InspectorInstrumentation.cpp:508
> +    TRACE_EVENT_INSTANT1("webkit", "Instrumentation::Paint", "pageId", frame ? reinterpret_cast<unsigned long long>(frame->page()) : 0);

These need constants. Is there anything we could reuse?

> Source/WebCore/inspector/TimelineTraceEventProcessor.cpp:179
> +TimelineTraceEventProcessor::TimelineTraceEventProcessor(InspectorTimelineAgent* timelineAgent, InspectorClient *client, Page* page)

timeline agent has page agent already.

> Source/WebCore/inspector/TimelineTraceEventProcessor.cpp:182
> +    , m_pageId(reinterpret_cast<unsigned long long>(page))

page can be recreated.

> Source/WebCore/inspector/TimelineTraceEventProcessor.cpp:191
> +    m_handlersByType.set("Instrumentation::Paint", EventTypeEntry(&TimelineTraceEventProcessor::onPaint));

Lets have constants for these in InspectorInstrumentation.

[Andrey Kosyakov]

Created attachment 182589 [details]
Patch

[Andrey Kosyakov]

(In reply to comment #4)
> 
> These need constants. Is there anything we could reuse?

Added constants to InspectorInstrumentation.{h,cpp}. I don't think we should be re-using existing trace events, as ours are different in that their names are meaningful to other modules.

> > Source/WebCore/inspector/TimelineTraceEventProcessor.cpp:179
> > +TimelineTraceEventProcessor::TimelineTraceEventProcessor(InspectorTimelineAgent* timelineAgent, InspectorClient *client, Page* page)
> 
> timeline agent has page agent already.

Exposed InspectorTimelineAgent::page()

> 
> > Source/WebCore/inspector/TimelineTraceEventProcessor.cpp:182
> > +    , m_pageId(reinterpret_cast<unsigned long long>(page))
> 
> page can be recreated.

Now within the lifetime of TimelineTraceEventProcessor -- we'll kill it upon stopping the record.

> > Source/WebCore/inspector/TimelineTraceEventProcessor.cpp:191
> > +    m_handlersByType.set("Instrumentation::Paint", EventTypeEntry(&TimelineTraceEventProcessor::onPaint));
> 
> Lets have constants for these in InspectorInstrumentation.

done!

[Andrey Kosyakov]

Created attachment 186012 [details]
Patch

[Andrey Kosyakov]

Created attachment 186871 [details]
Patch

[Andrey Kosyakov]

Committed r144044: <http://trac.webkit.org/changeset/144044>

[Andrey Kosyakov]

Reopening to attach new patch.

[Andrey Kosyakov]

Created attachment 190297 [details]
EWS run

[Early Warning System Bot]

Comment on attachment 190297 [details]
EWS run

Attachment 190297 [details] did not pass qt-ews (qt):
Output: http://webkit-commit-queue.appspot.com/results/16768384

[Early Warning System Bot]

Comment on attachment 190297 [details]
EWS run

Attachment 190297 [details] did not pass qt-wk2-ews (qt):
Output: http://webkit-commit-queue.appspot.com/results/16781163

[Andrey Kosyakov]

Created attachment 190300 [details]
EWS run

[Andrey Kosyakov]

Committed r144154: <http://trac.webkit.org/changeset/144154>