Bug 35900 – [Chromium] Need to send mouse events to plugin when it has mouse capture

• Status: RESOLVED
• Resolution: FIXED
• Component: Platform

[John Abd-El-Malek]

This would allow a plugin to implement scrollbars in its region, for example.

[John Abd-El-Malek]

Created attachment 50270 [details]
Proposed patch

I don't like how I can't set m_haveMouseCapture to false in mouseCaptureLost(), any suggestions?

[WebKit Review Bot]

Attachment 50270 [details] did not build on chromium:
Build output: http://webkit-commit-queue.appspot.com/results/453001

[John Abd-El-Malek]

Created attachment 50276 [details]
Added missing file

[Darin Fisher (:fishd, Google)]

What if WebWidget::mouseCaptureLost() is called?  See the callsite in chrome/renderer/render_widget.cc.

[John Abd-El-Malek]

Right, that's my worry, that the window loses mouse capture without a right mouse button up event.  Is it possible though?

[John Abd-El-Malek]

Created attachment 50350 [details]
Updated patch

Per Darin's advice, reversed the order that the browser sends the events (http://codereview.chromium.org/743003).  This makes this change cleaner.

[Darin Fisher (:fishd, Google)]

Comment on attachment 50350 [details]
Updated patch

> Index: WebKit/chromium/src/WebViewImpl.cpp
...
> +    if (m_haveMouseCapture && WebInputEvent::isMouseEventType(inputEvent.type)) {
> +        Node* node = focusedWebCoreNode();
> +        if (node && node->renderer() && node->renderer()->isEmbeddedObject()) {
> +          AtomicString eventType;
> +            switch (inputEvent.type) {

nit: indentation of AtomicString is off.


>  void WebViewImpl::mouseCaptureLost()
>  {
> +    m_haveMouseCapture = false;
>  }

We may eventually wish to pass this event along to the plugin.

Are there any compatibility concerns with enabling this patch for ordinary plugins?

[John Abd-El-Malek]

My hunch is that there shouldn't be any issue with passing this to NPAPI plugins.  Basic testing with Flash hasn't shown any problems, but pushing this out will tell us conclusively.

[John Abd-El-Malek]

Committed r55758

[Darin Fisher (:fishd, Google)]

(In reply to comment #8)
> My hunch is that there shouldn't be any issue with passing this to NPAPI
> plugins.  Basic testing with Flash hasn't shown any problems, but pushing this
> out will tell us conclusively.

OK, sounds good.