The problem is it /can't/ be a random page for removing or adding from
watchlist. We have to find a page that is already on the watchlist for the
test for removing from watchlist and for the test adding to watchlist we
have to ensure we land on a page which is not previously being watched.
We could possibly change these so that one of the Given statements is
"Given I am already watching the "Foo" article" - this could then
ensure
the article is being watched and vice versa.
A random page cannot be relied on under any circumstances.
On Wed, Mar 19, 2014 at 2:37 PM, Chris McMahon <cmcmahon(a)wikimedia.org>wrote;wrote:
I fixed it.
Since yesterday Cloudbees has unexpectedly given us unlimited Jenkins
build executors where before we could only run two builds at once. This
has made for a few race conditions, of which this was one.
This test should probably create a random page to use for clicking the
watchlist on and off.
-C
On Wed, Mar 19, 2014 at 2:30 PM, Jon Robson <jrobson(a)wikimedia.org> wrote:
They are indeed false positives. I think I know
what's happening here.
Basically the tests are not atomic - they are trying to watch an article
that is already watched and have the consequence of making that article
watched. I have an idea of how to fix this.
Does Watir have a concept of a tear down step?
On Wed, Mar 19, 2014 at 1:45 PM, Arthur Richards <arichards(a)wikimedia.org
wrote:
I tested these manually and think they may be
false positives, but can
someone else take a look to make sure these tests and what's in betalabs is
OK? We should get this resolved asap so we can be sure the state of things
is healthy on betalabs for tomorrow's deployment.
On Wed, Mar 19, 2014 at 12:19 PM, <jenkins-no-reply(a)cloudbees.com>wrote;wrote:
--
Arthur Richards
Software Engineer, Mobile
[[User:Awjrichards]]
IRC: awjr
+1-415-839-6885 x6687