#20 Need API for evaluate_javascript command

[Anonymous]

Originally created by: dpodwall

Currently the evaluate_javascript command is supported internally but not exposed as API. Attached is a
patch to add an evaluateJavaScript() method to the org.chromium.SDK.BrowserTab interface.

I took this approach because it seemed that the standalone v8 vm might not support it, so it could not be in
a command interface. The idea is that a client would do a runtime instanceof check on
org.chromium.sdk.JavascriptVm to see if it implements BrowserTab, and then cast to BrowserTab to invoke
evaluateJavaScript().

[Anonymous]

Originally posted by: peter.ry...@gmail.com

The proposed API might not be generic enough. The evaluate command does not return
result.

Please provide a rationale why there should be use-cases for such responseless
method.

There is a reason behind the internal method not having response. Response includes
object ref id, that is used further object property lookups. The ref id lives only
during suspend-continue period. Thus without suspended state there is no way you can
browse object's properties.

You also may use a work-around here:
javascriptVm.suspend()
...
debugContext.getGlobalEvaluateContext().evaluate...  // this API is in HEAD only
...
debugContext.continueVm();

Note that this API is fully available in HEAD only (hasn't made its way into release
builds yet).

[Anonymous]

Originally posted by: dpodwall

From the ChromeDevTools protocol doc I'd understood that evaluate_javascript doesn't
return any result, so the proposed API just matches that.

Here's my use case for the API. The desired scenario is that before starting a debug
session, user may set a breakpoint in code that will be executed early, e.g.
window.onload. We want the breakpoint to get hit the first time that code is executed, not
require a refresh.

In order to initiate the debug session we need to target a tab which has a specified URL. If
in starting up the browser we gave the URL for the user's page then their code would
execute before we'd actually attached to the tab, so breakpoints could not get install.

So the approach we're using is to initially point the browser to a static startup page. Once
we've successfully attached we use evaluate_javascript to get the browser to navigate to
the user's page.

I'll try the workaround you've suggested. But that requires an extra suspend and continue.
Since evaluate_javascript is in the protocol and works well for at least this use case it'd be
nice to have the plugins support it.

Thanks!

[Anonymous]

Originally posted by: peter.ry...@gmail.com

This bug should be covered by
https://code.google.com/p/chromedevtools/issues/detail?id=42

Status: WontFix