Bug 193741 – DidFirstVisuallyNonEmptyLayout milestone should always fire at some point.

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[zalan]

It's important that this milestone gets fired.

[zalan]

<rdar://problem/47135030>

[zalan]

Created attachment 359960 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 359960 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=359960&action=review

> Source/WebCore/page/FrameView.cpp:4550
> +void FrameView::firstVisuallyNonEmptyLayoutMilestoneWatchdogFired()

Firing your watchdog means that you weren't happy with how it was doing its job, so you let it go.

This is a watchdog timer getting fired.

> Source/WebCore/page/FrameView.cpp:4559
> +void FrameView::fireFirstVisuallyNonEmptyLayoutMilestoneIfRequested()

This function seems to do more than fire the milestone; it clears some state, and triggers the DidFirstMeaningfulPaint milestone.

> Source/WebCore/page/FrameView.h:836
> +    Timer m_firstVisuallyNonEmptyMilestoneWatchdog;

m_firstVisuallyNonEmptyMilestoneWatchdogTimer

[zalan]

Created attachment 359983 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 359983 [details]
Patch

Can we layout-test this?

[Simon Fraser (smfr)]

Or API test

[zalan]

(In reply to Simon Fraser (smfr) from comment #6)
> Or API test
We've got some API tests in this area. Need to think about how to simulate this case though.

[zalan]

Created attachment 360014 [details]
Patch

[zalan]

Created attachment 360020 [details]
Patch

[Antti Koivisto]

Comment on attachment 360020 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=360020&action=review

As someone mentioned, a test would be cool.

> Source/WebCore/page/FrameView.cpp:4557
> +void FrameView::handleFirstVisuallyNonEmptyLayoutMilestone()
> +{
> +    ASSERT(!m_isVisuallyNonEmpty);
> +    ASSERT(m_firstVisuallyNonEmptyLayoutCallbackPending);
> +    m_isVisuallyNonEmpty = true;
> +    m_firstVisuallyNonEmptyLayoutCallbackPending = false;
> +    addPaintPendingMilestones(DidFirstMeaningfulPaint);
> +
> +    if (!frame().isMainFrame())
> +        return;
> +    if (!frame().page() || !frame().page()->requestedLayoutMilestones().contains(DidFirstVisuallyNonEmptyLayout))
> +        return;
> +    frame().loader().didReachLayoutMilestone(DidFirstVisuallyNonEmptyLayout);
> +}

Can't we just call fireLayoutRelatedMilestonesIfNeeded() instead of this? I'm not sure why another function doing the firing is needed. I'm not sure this does everything that needs to be done.

[zalan]

(In reply to Antti Koivisto from comment #10)
> Comment on attachment 360020 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=360020&action=review
> 
> As someone mentioned, a test would be cool.
> 
> > Source/WebCore/page/FrameView.cpp:4557
> > +void FrameView::handleFirstVisuallyNonEmptyLayoutMilestone()
> > +{
> > +    ASSERT(!m_isVisuallyNonEmpty);
> > +    ASSERT(m_firstVisuallyNonEmptyLayoutCallbackPending);
> > +    m_isVisuallyNonEmpty = true;
> > +    m_firstVisuallyNonEmptyLayoutCallbackPending = false;
> > +    addPaintPendingMilestones(DidFirstMeaningfulPaint);
> > +
> > +    if (!frame().isMainFrame())
> > +        return;
> > +    if (!frame().page() || !frame().page()->requestedLayoutMilestones().contains(DidFirstVisuallyNonEmptyLayout))
> > +        return;
> > +    frame().loader().didReachLayoutMilestone(DidFirstVisuallyNonEmptyLayout);
> > +}
> 
> Can't we just call fireLayoutRelatedMilestonesIfNeeded() instead of this?
> I'm not sure why another function doing the firing is needed. I'm not sure
> this does everything that needs to be done.
Since this is not an actual layout, I thought it would be incorrect to call fireLayoutRelatedMilestonesIfNeeded() -but obviously it fixes the bug as well. Let's go back to that.

[zalan]

>I'm not sure this does everything that needs to be done.
It does.

[zalan]

Created attachment 360022 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 360022 [details]
Patch

Please rename m_firstVisuallyNonEmptyLayoutCallbackPending. It's not a callback.

[Antti Koivisto]

Comment on attachment 360022 [details]
Patch

r=me

[zalan]

Created attachment 360034 [details]
Patch

[zalan]

Created attachment 360039 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 360039 [details]
Patch

Clearing flags on attachment: 360039

Committed r240450: <https://trac.webkit.org/changeset/240450>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.