[#SRC-383] After test is timed out, browser window is not closed and proxy setting in InetOptions is not restored.

Selenium Remote Control

After test is timed out, browser window is not closed and proxy setting in InetOptions is not restored.

Created: 09/Nov/07 05:08 PM Updated: 14/Nov/08 08:42 PM

Component/s:

Launcher - Internet Explorer

Affects Version/s:

0.9.2

Fix Version/s:

None

Original Estimate:	Unknown	Remaining Estimate:	Unknown	Time Spent:	Unknown
File Attachments:	1. src-383-patch-no-refactoring.diff (9 kb) 2. src-383-patch-with-refactoring.diff (37 kb)
Environment:	IE6/WinXP, IE7/WinXP, IE7/Vista

Description

« Hide

Occasionally, an automated test stops/hangs/or otherwise does not continue past the initial
browser launch. The test is then terminated with a stop(). The server is still in a usable state
but the browser remains open and the proxy directory to use is still selected in the
Inet Options -> Connections -> LAN Settings menu. Following tests are then unable to
launch because the proxy settings are no longer correct.
The RC server in this case is run in -avoidProxy mode, using *iexplore

All

Comments

Work Log

Change History

Sort Order:

[ Permlink | « Hide ]

Jennifer Bevan - 09/Nov/07 05:11 PM

I don't have a good test to expose this yet, but I did need an issue ID number for tracking purposes.

[ Permlink | « Hide ]

Zach Leatherman - 19/May/08 05:00 PM

I think this bug is a little bit more serious that suggested here. If the server is ended prematurely (using CTRL+C, which may be the only option for a failed test that isn't going to return data to the server), the Proxy settings have been overwritten and are not restored to their original state. Depending on your proxy restrictions, this could disable all other web browsing activities in all new Internet Explorer instances.

If the app fails or is ended, at least restore the original registry key for the proxy.

The registry key in question here is described at:
http://support.microsoft.com/kb/819961/

[ Permlink | « Hide ]

Jennifer Bevan - 14/Nov/08 05:27 PM

This is the preliminary patch that ensures that the environment is restored if browser launch fails. Note the similarity in the change. Moving the logic in launch() up to AbstractBrowserLauncher would be a good way to enforce this across all the other launchers.

[ Permlink | « Hide ]

Jennifer Bevan - 14/Nov/08 08:39 PM

This is the much larger patch with the refactoring done. Note that this only deals with BrowserLaunchers that extend AbstractBrowserLauncher: it does not therefore cover the HTABrowserLauncher.

The unit tests for the changed files pass, it builds, but I haven't run through the functional tests yet. I will start testing this on *chrome/linux, *iexplore/winxp, and *safari/macos; others will want to try this on other configurations (or our new continuous integration system :))

-Jen