1358642 - stylo: evaluate whether the unship of -moz-linear-gradient was successful

Status: RESOLVED WORKSFORME

(Core :: CSS Parsing and Computation, enhancement, P3)

Description

[Boris Zbarsky [:bzbarsky]]

Looks like this is a thing, and it's not quite the same as either linear-gradient or -webkit-linear-gradient.  Bunch of reftest failures from this one not being supported.

Some implementation notes:

CSSParserImpl::ParseLinearGradient has code that switches on (aFlags & eGradient_MozLegacy) which probably needs to be duplicated in stylo.

There's webkit bits there too, but I think bug 1345204 tracks that.

Comment 1

[Bobby Holley (:bholley)]

Nox is working on gradient stuff, so presumably he can knock this one out. High priority given that it's contributing to our reftest failure count.

Comment 2

[Daniel Holbert [:dholbert]]

(In reply to Boris Zbarsky [:bz] (still a bit busy) (if a patch has no decent message, automatic r-) from comment #0)
> Looks like this is a thing, and it's not quite the same as either
> linear-gradient or -webkit-linear-gradient.

We're working on making -moz-linear-gradient *not a thing*, in bug 1337655 (which is just a pref-flip that we can make after its dependencies land), and formerly in bug 1176496.

If that change sticks & doesn't cause too much webcompat pain, then I don't think we want Stylo to bother supporting this.,

Comment 3

[Daniel Holbert [:dholbert]]

For reference, xidorn also posted an intent to un-ship here:
 https://groups.google.com/d/msg/mozilla.dev.platform/egVDMiu86m0/t53-r2efDwAJ

Comment 4

[Bobby Holley (:bholley)]

Ah ok. Makes sense!

Comment 5

[Bobby Holley (:bholley)]

(We can reevaluate in June or July whether we need to do anything here)

Comment 6

[Boris Zbarsky [:bzbarsky]]

I think we want to enable the various gradient tests for stylo (e.g. by switching them to the prefixless gradient syntax) way before June/July...  Daniel, do you know whether someone is actively working on that?

Comment 7

[Daniel Holbert [:dholbert]]

Actively working on which?  Unshipping moz-linear-gradient?

I think xidorn is taking that, via bug 1337655, per his intent-to-ship. (Its dependent bugs all have patches awaiting review, so once those have r+, it's a simple as an r+ and a mass prefixed test-disable-renewal -- which I tentatively assume xidorn is also doing as part of bug 1337655, but I could do too.)

Comment 8

[Daniel Holbert [:dholbert]]

> it's a simple as an r+

Sorry, I mixed up my words -- I meant "as simple as a pref flip"

Comment 9

[Daniel Holbert [:dholbert]]

...and by "test-disable-renewal" I meant "test-disable-or-removal". Sorry. :D I'm sleepy.

Though, you mentioned "switching them [tests] to the prefixless gradient syntax" -- I hadn't thought there was any of that that needed to be done -- I was thinking we just had property_database.js entries to remove, basically.  But now that you mention it, judging by the patch[1] that landed on bug 1176496 {and was later backed out), I suppose we do have a lot of tests to fix up. :(  But I think the patch still mostly applies cleanly, at least for reftests! So I don't think there's a lot to be done there.

I'll file a helper bug for bug 1337655 to cover that.  Thanks for bringing it up! 

[1] https://hg.mozilla.org/mozilla-central/rev/acb7eb7f5ad4

Comment 10

[Boris Zbarsky [:bzbarsky]]

> Actively working on which?  Unshipping moz-linear-gradient?

No, converting our reftests to not use it.  For stylo's "figure out where we still have failures" purposes, that's sufficient.

Landing just the reftest changes from bug 1176496 would be a pretty good step forward, for sure.

Comment 11

[Bobby Holley (:bholley)]

Nazim is fixing up the tests in bug 1358710.

Comment 12

[Bobby Holley (:bholley)]

We implemented -moz-linear-gradient in bug 1366544.