EnyoJS sampler doesn't work.

[glennw]

I did some investigation into it - info below for anyone else who looks into it:

To set up a local version you can test with:

git clone https://github.com/enyojs/sampler
git submodule init
git submodule update

Then build the JS (you need node + grunt installed):
./tools/deploy.sh

By default, this produces a minified / obfuscated build. To work around that, I hacked the compressJsFile function in tools/minifier/minify.js to be:

return fs.readFileSync(inPath, "utf8");

Then you get concatenated JS that is debuggable.

The first problem I ran into is our lack of history implementation - but this can easily be commented out.

After that, the same problem remains. The code appears to do all ajax correctly and get the right sample information. It then calls into pushSampleList. As far as I can tell, servo behaves the same to this point as FF.

[glennw]

cc @metajack

[glennw]

@metajack The enyo libraries rely on (at least some parts of) cssom to work. The sampler fails due to missing support for https://github.com/mozilla/gecko-dev/blob/master/dom/webidl/HTMLElement.webidl#L84.

[glennw]

@metajack enyo also relies on MouseEvent implementing the (non standard) 'which' field of MouseEvent (see https://developer.mozilla.org/en-US/docs/Web/API/MouseEvent/which and https://dxr.mozilla.org/mozilla-central/source/dom/events/MouseEvent.h#40).

This should be trivial to implement (although I'm not sure if we support implementing binding methods without a WebIDL declaration, which is what the current gecko code does). Are we happy to implement this non-standard field? The MDN page says it's supported in every other browser (well, IE9+ anyway).

[metajack]

@Ms2ger @jdm any ideas what we should do about MouseEvent?

[Ms2ger]

(non standard)

https://w3c.github.io/uievents/#KeyboardEvent-supplemental-interface

implementing binding methods without a WebIDL declaration, which is what the current gecko code does

Gecko implements it on UIEvent for some reason.

[glennw]

@metajack enyo also seems to rely on transition callbacks too, which I don't believe we support yet. It does use animation + animation callbacks, but I'm not sure to what extent yet.

[glennw]

@metajack Ah, part of the reason why none of the button presses are working is because the target of our mousedown events is different than what FF sets, so they get dispatched incorrectly. I'll look into how our hit detection is working.

The servo hit detection doesn't work with (2d or 3d) transforms, which explains the discrepancy here. With the way enyo works, this places all panels at (0,0) so the hit detection is wrong for buttons etc.

[jdm]

Filed https://bugzilla.mozilla.org/show_bug.cgi?id=1186125 about Gecko's noncompliance here, fwiw.

[glennw]

enyo relies heavily on transforms for doing layout in js. But some of the transforms are failing to parse - looking into that now before fixing the hit test code.

[glennw]

Partial fix for hit testing (needs servo integration) servo/euclid#98

Fix for mouseevent.which #6691