launchpad-dev team mailing list archive

[Ian Booth <ian.booth@xxxxxxxxxxxxx>]

Hi Curtis

>>
>> This code creates a "Test status: complete" node on the page when the
>> test runner finishes. The Windmill harness looks for this node and
>> then
>> processes the test results. If you run the tests manually in a
>> browser,
>> everything looks ok. However, without the above code, when invoked as
>> part of our test run, the test appears to fail with a Windmill
>> waits.forElement timeout, even if the tests have all passed. 
> 
> I apologise for not updating the article. I wrote the initial article,
> then about about a year later I wrote the YUI-Windmill test integration.
> I neglected to revisit the article; I had forgotten about it. At the
> time, I felt like I was the only person determined to never write a
> windmill test again.
> 

No problem at all - I didn't even look to see who wrote it. I just think
it's great that more of us are actually starting to write javascript
tests and also correctly structured windmill tests. To me they are a
valuable part of our test coverage and have picked up more than one bug
in my code. Now that they are turned on again, let's hope the
reliability issues have been sufficiently sorted so that we can run them
without affecting our build throughput. Plus, based on Deryck's advice
at the Epic and subsequent emails, much of what we would have previously
written as a windmill tests should be done as a Javascript test anyway
and these are much faster to run too.

My remaining issue with windmill is the apparent unreliability of
jquery. I've found for example that a waits.forElement falsely matches
and then when you go to use the same jquery for a client.click (with the
"[0]" appended), it fails. xpath has no such problems.