webmessaging tests with postMessage API about wpt HOT 6 CLOSED

I made a glance at the tests, and I noticed one suspicious point: in these tests, iframe has an onload handler that calls a function, and test() is used in that function. Is this allowed? I guess async_test() should be used in such cases.

from wpt.

Well, to be clear, I think the test should use the following pattern:

var test = async_test(...);
var iframe = document.createElement('iframe');
// Set iframe's attributes, attach to document...
iframe.onload = test.step_func(function (evt) {
    // assert_whatever...
    test.done();
});

from wpt.

This git issue serves as a reminder that the event.*.htm tests were not working when I last tried them as the postMessage from testharness.js was mistakenly assumed in the test to carry the expected test data.

@yutak I don't think initiating the iFrame from JS or using the iFrame tag really matters, if we are discussing syntax and code style for example, I would more concerned about;

https://github.com/w3c/web-platform-tests/blob/master/webmessaging/event.data.htm

• var DATA is using ALL CAPS, WHY!?
• var ExpectedResult Uses Unexpected Caps someTimes On the First letter of a Variable name Instead of camelCase
• ==, should be ===, always.
  ... anyway. No one is perfect when it comes to code style, and there is no style guide for writing W3C tests (yet?).

This week it will make it my duty to finish off the minimum required work to get these tests to not fail.
Would be interested to know why testharness.js needs to use postMessages...

from wpt.

@aogilvie OK, I think I understand your problem. I'm not sure about why testharness.js is using postMessage, either.

FYI, see https://github.com/w3c/web-platform-tests/blob/master/webmessaging/postMessage_ArrayBuffer.htm for the problem I talked about; this test calls test() function inside the onload event handler, and I'm not sure if this is legitimate.

from wpt.

testharness.js uses postMessage to collect results cross-origin, see w3c/testharness.js#8

from wpt.

Sounds like this was about bugs in the tests at https://wpt.fyi/webmessaging to begin with? I've taken a look at it seems like all tests pass in at least one browser, which is better than expected. Are there still bugs to be fixed here?

@mdittmer @lukebjerring this is a typical case where the interop view would have come in handy :)

from wpt.