RemoteWebDriver should not systematically -when an exception is thrown- take a snapshot and send it back

[lukeis]

Originally reported on Google Code with ID 185

Currently whenever the Webdriver is meant to throw an exception, the
RemoteServer takes a snapshot, and sends it back as a json very long string.

Typically doing a FindElement(By.Id()) of a non-existent DOM element is
going to systematically trigger the creation of such a snapshot,
which in turns -no surprisingly- terribly slows down running tests which
rely on frequently checking that some elements are not present.

Reported by alexis.j.vuillemin on 2009-04-23 06:32:17

[lukeis]

Reported by alexis.j.vuillemin on 2009-04-23 06:33:49

• Status changed: New

[lukeis]

Reported by jari.bakken on 2010-06-16 23:43:57

• Labels added: Lang-Java

[lukeis]

I think user should be able to decide whether to capture screenshot or not.

Reported by jttexe on 2011-02-24 22:13:58

[lukeis]

When creating a new session, the ChromeDriver checks for the "takeScreenshotOnError"
capability to decide whether to automatically take a screenshot. This should be adopted
by the other drivers.

Reported by jmleyba on 2011-06-10 17:02:27

[lukeis]

Reported by barancev on 2011-10-12 18:09:06

• Labels added: Component-WebDriver

[lukeis]

Reported by dawagner on 2012-02-05 17:04:40

• Labels added: GettingInvolved

[lukeis]

DefaultSession is used by most clients, and this class does not check takeScreenshotOnError.
It registers a SnapshotScreenListener before it even knows what the value of takeScreenshotOnError
is. In the case of Android driver, it skips appending the listener.

I removed the binding of the listener out in DefaultSessionJava, and it cut 45 minutes
off of our 3.5 hour test run last night:

-      initialDriver.register(new SnapshotScreenListener(this));

Reported by eli.sarver on 2012-04-11 12:48:01