Last modified: 2014-10-07 08:39:19 UTC
Create a test that loads a typical (existing) items and compares it to the expected output.
We already have many tests tauching items in some way. So what is the point of adding a smoke test? Adding a smoke test when there are more detailed tests already makes little sense. Perhaps the term "smoke test" is used incorrectly here?
It does make sense. We have had too many cases where our specific tests didn't catch issues that would have been caught with this.
(In reply to Jeroen De Dauw from comment #1) > We already have many tests tauching items in some way. So what is the point > of adding a smoke test? Adding a smoke test when there are more detailed > tests already makes little sense. Perhaps the term "smoke test" is used > incorrectly here? It happened several times recently that things broke and were only recognized after it was deployed to test.wikidata.org or in worst cases after we went live. All these issues could have been avoided by just looking at a showcase item on beta because breakage was visible there. Just nobody looked there manually. The purpose of such a test (call it however you want) is to automate this and get instant reports if something that obvious breaks on beta. Hope that explained the task a bit more.
(In reply to tobias.gritschacher from comment #3) > (In reply to Jeroen De Dauw from comment #1) > > We already have many tests tauching items in some way. So what is the point > > of adding a smoke test? Adding a smoke test when there are more detailed > > tests already makes little sense. Perhaps the term "smoke test" is used > > incorrectly here? > > It happened several times recently that things broke and were only > recognized after it was deployed to test.wikidata.org or in worst cases > after we went live. All these issues could have been avoided by just looking > at a showcase item on beta because breakage was visible there. Just nobody > looked there manually. > The purpose of such a test (call it however you want) is to automate this > and get instant reports if something that obvious breaks on beta. > Hope that explained the task a bit more. Addition: the detailed tests we have would not catch breakage that is related to existing items. All the tests we currently have create new items. There is no test that checks an existing item with hundreds of statements, references and qualifiers. This task will close this lack.
https://github.com/wmde/WikidataBrowserTests/pull/11
> Hope that explained the task a bit more. It did.