Bug 44957 – Web Inspector, Extension API: Panel.onSelectionChanged event is never fired

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

[Andrey Kosyakov]

webInspector.panels.elements.onSelectionChanged should be fired when a DOM element is selected in elements panel.

[Andrey Kosyakov]

Created attachment 66064 [details]
patch

- Added Panel.name
- Fixed the way onSelectionChanged is dispatched from ElementsPanel.js
- Added console API into context of code being evaluated by webInspector.inspectedWindow.evaluate() (so we can do inspect() there)
- Added basic test for extension API events

[Andrey Kosyakov]

Created attachment 66071 [details]
patch

- Extracted stringifying wrapper around expressions evaluated by extensions into a more readable function.

[Andrey Kosyakov]

Comment on attachment 66071 [details]
patch

oops, wrong patch. sorry.

[Andrey Kosyakov]

Created attachment 66073 [details]
patch

[Yury Semikhatsky]

Comment on attachment 66073 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=66073&action=prettypatch

> WebCore/inspector/front-end/ExtensionServer.js:225
> +            var result = window.eval(
I believe we can get rid of this eval by providing the function body directly or even call the eval without wrapping it into an anonymous function as we discussed offline. A bigger problem is that this code doesn't allow extensions to declare global functions, r- for this.

[Andrey Kosyakov]

Created attachment 66207 [details]
patch

- changed webInspector.inspectedWindow.evaluaet() to return object instead of its stringified representation
- rewrote wrapper for evaluated code
- added more tests for eval

[Andrey Kosyakov]

(In reply to comment #5)
> (From update of attachment 66073 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=66073&action=prettypatch
> 
> > WebCore/inspector/front-end/ExtensionServer.js:225
> > +            var result = window.eval(
> I believe we can get rid of this eval by providing the function body directly or even call the eval without wrapping it into an anonymous function as we discussed offline. A bigger problem is that this code doesn't allow extensions to declare global functions, r- for this.

The latter happened to be not the case: the code being evaluated still runs in global context. Added a test for that, though, and rewrote wrapper code to a shorter (though less readable) version without outer function.

[Andrey Kosyakov]

Manually committed r66601: http://trac.webkit.org/changeset/66601

[Eric Seidel (no email)]

http://trac.webkit.org/browser/trunk/LayoutTests/inspector/extensions-events.html is flaky on the Leopard Commit bot and is causing the bot to back up. :(

[Eric Seidel (no email)]

1  CONSOLE MESSAGE: line 663: [object HTMLParagraphElement]
 1 CONSOLE MESSAGE: line 667: [object HTMLParagraphElement]

Is the diff.

[Andrey Kosyakov]

(In reply to comment #10)
> 1  CONSOLE MESSAGE: line 663: [object HTMLParagraphElement]
>  1 CONSOLE MESSAGE: line 667: [object HTMLParagraphElement]
> 
> Is the diff.

This should have been fixed by r67394: http://trac.webkit.org/changeset/67394
It is not actual flakiness, rather a stable, though unexpected, side effect from r67385: the inspect() console command line API call uses console.log() internally, which causes line number of a line within InjectedScript.js that called console.log() to appear in test expectations. Hence a change to InjectedScripts.js that causes a call to console.log() to move causes test expectations change.