[Webkit-unassigned] [Bug 43344] Web Inspector: Extensions API tests fail on Qt & GTK

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=43344





--- Comment #7 from Andrey Kosyakov <caseq at chromium.org>  2010-08-03 23:18:29 PST ---
(In reply to comment #6)
> (From update of attachment 63367 [details])
> WebCore/inspector/InspectorController.cpp:464
>  +      if (m_scriptsToEvaluateOnLoad.size()) {
> So we are now hitting this code for every page. 

Yes we do, but what's the problem with this? This is the cheapest check that we're going to have ever. The only way something under this conditional is going to be executed is when inspector previously added scripts to evaluate on load, which currently only happens either (a) for a page being inspected if user has enabled workers debugging or (b) for inspector page if addExtensions has been called.

This also used to work unconditionally on enabled(), just upon a slightly later event for half a year: http://trac.webkit.org/changeset/64458/trunk/WebCore/inspector/InspectorController.cpp

> I never liked this thing, I don't think we should extend its capabilities.

What we essentially need is an ability of inspector to inject some JS to certain pages at the earlist possible moment. We now have a generic means for that, which covers inspector extensions, workers and any possible other case where we could need to emulate some API for page being inspected (we intentially made the scheme generic for that, as you suggested back when it was introdcued). Now, we could have it more specialized for extensions API, but it would be doing roughly the same: setting the API from inspector and injecting it in about the same moment into inspector's page. So, what exactly would be the difference that would justify the additional code?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.