Change 199794 had a related patch set uploaded (by Jdlrobson):
QA: Browser test for edit workflow

https://gerrit.wikimedia.org/r/199794

I'm reviewing this but I'm having trouble running the tests. Investigating...

Change 199794 merged by jenkins-bot:
QA: Browser test for edit workflow

https://gerrit.wikimedia.org/r/199794

I've merged the patch since it works cool, but I have a couple of concerns @Jdlrobson

• Edit collection tests depend on new collection tests

I was having trouble running the tests because the edit collection tests are dependent on having run new collection before (otherwise the public collection does not exist and the test fails).

This works when we run the tests all at once luckily because probably either the selenium user got the public collection on the database at some point or the new_collection tests run before the edit_collection tests, but if you want to run individual features if the collection isn't on the database it doesn't work.

Seems a bit brittle, do you think we should have a task to make test features independent?

• Editing the description with the same text always

Maybe I'm wrong about this one but, this test edits the collection's description always with the same text. If such collection already had that text the test will pass even if editing didn't worked, right?

This will properly catch fails on a new clean environment but when we run the tests locally the selenium's user public collection will probably have already the result description so the test will always be correct even if the editing didn't happen?

We could fix this by performing 2 edits on the collection description, should be an easy fix.

Am I right thinking this is a problem?

@Jhernandez yeh this is all as you can. They are sadly a little brittle. I need to familiarise myself with the ruby api code to make this not the case. Have setup T94701