Bug 40119 - Web Inspector [JSC]: Instrument JS function calls from bindings for Timeline Panel.
Summary: Web Inspector [JSC]: Instrument JS function calls from bindings for Timeline ...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Web Inspector (Deprecated) (show other bugs)
Version: 528+ (Nightly build)
Hardware: All All
: P2 Normal
Assignee: Eli Fidler
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2010-06-03 06:40 PDT by Pavel Feldman
Modified: 2012-01-26 12:39 PST (History)
8 users (show)

See Also:


Attachments
Patch (11.83 KB, patch)
2012-01-19 14:50 PST, Eli Fidler
no flags Details | Formatted Diff | Diff
Patch (11.89 KB, patch)
2012-01-20 15:17 PST, Eli Fidler
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Pavel Feldman 2010-06-03 06:40:33 PDT
timelineAgent->willCallFunction(resourceName, lineNumber); should be called each time some native event handler passes control to the JavaScript function.
Comment 1 Timothy Hatcher 2010-06-03 06:41:34 PDT
This is something V8 has, that we should consider for JSC to match Chrome.
Comment 2 Eli Fidler 2012-01-19 14:50:42 PST
Created attachment 123200 [details]
Patch
Comment 3 Pavel Feldman 2012-01-20 03:33:49 PST
Comment on attachment 123200 [details]
Patch

Web Inspector part looks good. Please have someone from the JSC team look at it (JSCallbackData changes and instrumentedCall source / line fetching).
Comment 4 Eli Fidler 2012-01-20 09:07:06 PST
CC ggaren and olliej for feedback
Comment 5 Geoffrey Garen 2012-01-20 11:40:48 PST
Comment on attachment 123200 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=123200&action=review

> Source/WebCore/bindings/js/JSCallbackData.cpp:73
> +        Page* page = static_cast<JSDOMWindow*>(globalObject())->impl()->frame()->page();
> +        ASSERT(page);

What guarantees that frame() and page() will not be NULL here? I think you need to NULL check both.

> Source/WebCore/bindings/js/JSMainThreadExecState.h:63
> +            String resourceName("undefined");

Creating the string "undefined" should be conditional -- otherwise, it's a malloc per function call.

> Source/WebCore/bindings/js/JSMainThreadExecState.h:76
> +        InspectorInstrumentation::didCallFunction(cookie);

This code will call didCallFunction even if it didn't call willCallFunction. Is that right?
Comment 6 Eli Fidler 2012-01-20 15:17:52 PST
Created attachment 123392 [details]
Patch
Comment 7 Eli Fidler 2012-01-20 15:18:43 PST
addressed ggaren's first two comments.

The didCallFunction even if not willCallFunction paradigm is the same as the V8 implementation. It should work fine.
Comment 8 Geoffrey Garen 2012-01-26 11:25:13 PST
Comment on attachment 123392 [details]
Patch

r=me
Comment 9 WebKit Review Bot 2012-01-26 12:39:23 PST
Comment on attachment 123392 [details]
Patch

Clearing flags on attachment: 123392

Committed r106030: <http://trac.webkit.org/changeset/106030>
Comment 10 WebKit Review Bot 2012-01-26 12:39:28 PST
All reviewed patches have been landed.  Closing bug.