[dojo-contributors] Dijit Intern-Conversion

Bill Keese bill at dojotoolkit.org
Tue Dec 23 02:24:46 EST 2014


> 2) The tests for menu_mouse were all written with the state of the last
> test influencing the next test.  This seems to be discouraged in Intern,
> and certainly made test conversions much harder.
>

​When you read that test file, think of each doh.register()​

​call as one test, and each runTest() is just a step in a test.


RE: General Conversion
> Regarding testing, it is going to be a tough balance doing the Dijit
> conversion.  Just the that one file took much longer than I anticipated,
> and I did find some small bugs in the original tests.  Getting some
> additional “plugins” to make dijit testing easier is probably going to be a
> big win for the conversion.  I am not sure you would want to modify the
> code too much on a per test basis. I tried to stay as true to the original
> test as possible.   I do not think there is going to be a “good” way of
> doing the conversion, considering the back port requirements.  Getting them
> done “quickly” is probably the best bet, then running them “backwards” to
> 1.7 – and manually deleting/updating the parts of tests that are “wrong”
> for that version – is probably the only way to go.
>
> Any way you cut it, it is a huge effort.
>
>
​I agree it's a huge effort, and we shouldn't try to boil the ocean by
conflating test conversion with other tasks​ like rearchitecting the tests.

Perhaps it's better to do some test cleanup within the existing
infrastructure first, in preparation for an Intern conversion.  That would
be a bird in the hand, even if we didn't meet the stretch goal of
converting dijit tests to Intern.

You mentioned converting functional tests to unit tests, and also about
making tests more independent.  Those tasks could both be done in the
existing infrastructure.  Also, most of the existing dijit unit tests start
by running the parser on a large chunk of HTML.  That's somewhat awkward
for an Intern unit test; it's more natural to create the widgets
programatically.  So that's another possible refactoring within the
existing infrastructure.

RE: DOH Robot
> ​​I am interested in trying to get DOHRobot running again, but will wait
> until it is signed with a valid cert.  I spent enough time messing with
> certs/self-signing trying to get it working, so I am hesitant to try again,
> with a valid, but untrusted cert since the valid one is coming ’soon'.  I
> have some more stuff to submit, but if I need to write DOHRobot tests for
> them, I want to make sure I have a working DOHRobot setup.


​Yeah, self-signing is tricky and I hope we get the properly signed robot
soon too.  But like I said, I've had good luck with the un​signed jar
(which I put at
 http://bill.dojotoolkit.org/DOHRobot.jar
​), so you should try that, and just add your machine to the list of sites
that java trusts (using the java control panel)

Despite rumors to the contrary, it's definitely possible to have the DOH
tests run, with only occasional intermittent errors.


> ​RE: #18262 – Changing dijit/place may also be an option.  But when I was
> testing with Claro, Nihilo, and Soria – it seemed only Tundra was
> different.  Would you like me to submit a follow up patch moving
> dijit/Place, or add an additional DOHRobot test?


It was just an idea.  Now that we've checked in the change to tundra I'm
not really looking for any further changes
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.dojotoolkit.org/pipermail/dojo-contributors/attachments/20141223/791c424e/attachment.htm 


More information about the dojo-contributors mailing list