Make default linear-gradient() direction 'to bottom', not 'to top'.

[heycam]

Found this while looking at failing Firefox reftests with stylo. If a linear-gradient() value doesn't specify a direction, then it should go from top to bottom.

r? @shinglyu

---

• ./mach build -d does not report any errors
• ./mach test-tidy does not report any errors
• These changes fix #__ (github issue number if applicable).

• There are tests for these changes OR
• These changes do not require tests because _____

---

This change is 

[highfive]

Heads up! This PR modifies the following files:

• @bholley: components/style/values/computed/image.rs
• @emilio: components/style/values/computed/image.rs

[highfive]

Warning

• These commits modify style code, but no tests are modified. Please consider adding a test!

[DominoTree]

Think this is bug #14745 and was fixed in #14746

[DominoTree]

@heycam What reftests were failing? Or was it not related to the incorrect direction?

[heycam]

@DominoTree Oh, I guess you just got in before me. :-) There are a bunch of tests under tests/wpt/css-tests/css-images-3_dev/ that fail, but because that directory of tests is skipped by default, and there is no expectation metadata, it's hard to see which ones were failing before but would now be passing. Closing in favour of your PR.

[heycam]

My mistake, there aren't any linear-gradient() tests under that directory.

[DominoTree]

Do you think those would be the right place to add tests for this? I haven't really delved into those yet, and I added a unit test.

[heycam]

I'm not sure what the process is for adding tests under tests/wpt/ that should be upstreamed, unfortunately. (In Firefox, we have a script that periodically upstreams changes to imported test suites like this, but I'm not sure how it works in Servo.)

[heycam]

(Also I realise earlier you probably were asking about which Firefox reftests were failing -- the one I noticed was a print preview reftest, where we have a dark grey gradient shown behind the pages to be printed. There would be other reftests that more directly test linear-gradient() that would be failing too, but the process for sorting out all the currently failing stylo reftests is going to be a long one so I didn't notice them yet.)

[DominoTree]

I was mainly asking because I was hoping you'd stumbled on some linear-gradient tests that I'd missed :D

I guess I'll dive into these ref tests and see if I can make heads or tails of how to roll my own, and go from there :)

[shinglyu]

@DominoTree IIRC, if your test is general enough to be upstreamed, please submit a PR to WPT test, after it merged, you can use the method described here to update the tests.