I brought up an XS instance on hpcloud and ran the tests there. I can reproduce this locally as well, but it's somewhat less frequent. Basically the problem is a race condition that leaves hanging firefox processes (created with the "firefox -silent" command, which is used by web driver to reset the fx profile). If I kill those processes manually, the tests then continue and succeed.
Selenium has a WebDriver for PhantomJS, so it plugs into the initial driver creation we have - no changes to our existing tests necessary. I think I spoke too soon regarding the flakiness though… I see qunit failures with the PhantomJS driver as well. I've increased the socket timeouts and wait times and am currently running the tests 100 times to see if that has helped.
I brought up an XS instance on hpcloud and ran the tests there. I can reproduce this locally as well, but it's somewhat less frequent. Basically the problem is a race condition that leaves hanging firefox processes (created with the "firefox -silent" command, which is used by web driver to reset the fx profile). If I kill those processes manually, the tests then continue and succeed.
I applied a variation of the fix used in SST and the hangs stopped: code.ohloh. net/file? fid=OruYKyN- ihGMgE1hoxb7Lkg 1cqI&cid= bSNDWAKLkUw& s=&fp=299267& mp&projSelected =true#L108
http://
Selenium has a WebDriver for PhantomJS, so it plugs into the initial driver creation we have - no changes to our existing tests necessary. I think I spoke too soon regarding the flakiness though… I see qunit failures with the PhantomJS driver as well. I've increased the socket timeouts and wait times and am currently running the tests 100 times to see if that has helped.