Update Rust to 1.15.0-nightly (71c06a56a 2016-12-18)

[nox]

This change is 

[highfive]

Heads up! This PR modifies the following files:

• @fitzgen: components/script/dom/treewalker.rs
• @KiChjang: components/script/dom/treewalker.rs, components/net/image_cache_thread.rs

[highfive]

Warning

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

[nox]

@bors-servo p=17

[SimonSapin]

@bors-servo r+

[bors-servo]

📌 Commit c5bd4e1 has been approved by SimonSapin

[bors-servo]

⌛ Testing commit c5bd4e1 with merge afaafc3...

[bors-servo]

💔 Test failed - linux-rel-wpt

[SimonSapin]

@bors-servo try

[KiChjang]

@bors-servo r- clean

[KiChjang]

@bors-servo try

[KiChjang]

@bors-servo r- try- clean

[KiChjang]

@bors-servo try retry

[bors-servo]

⌛ Trying commit c5bd4e1 with merge 6517409...

[bors-servo]

💔 Test failed - linux-rel-wpt

[SimonSapin]

So, that’s two runs with identical failures:

{"status": "FAIL", "stack": null, "subtest": null, "action": "test_result", "test": "/_mozilla/css/border_radius_asymmetric_sizes_a.html", "line": 13216, "message": "/_mozilla/css/border_radius_asymmetric_sizes_a.html 327908e8e5639151bbfceb0c4a78876203042ed1\n/_mozilla/css/border_radius_asymmetric_sizes_ref.html 5a9a64f75ad410d2c74e45ceda1c9efa61203c14\nTesting 327908e8e5639151bbfceb0c4a78876203042ed1 == 5a9a64f75ad410d2c74e45ceda1c9efa61203c14", "expected": "PASS"}
{"status": "PASS", "stack": null, "subtest": null, "action": "test_result", "test": "/_mozilla/css/text_decoration_underline_subpx_a.html", "line": 16535, "message": null, "expected": "FAIL"}
{"status": "FAIL", "stack": null, "subtest": null, "action": "test_result", "test": "/_mozilla/css/word-break-keep-all-005.htm", "line": 30050, "message": "/_mozilla/css/word-break-keep-all-005.htm 553453fad4db496e2aab49f65a8af2ec63909388\n/_mozilla/css/word-break-keep-all-ref-005.htm ec3b8afcf107bfa4cd93ad6b32c9238dfec0356d\nTesting 553453fad4db496e2aab49f65a8af2ec63909388 == ec3b8afcf107bfa4cd93ad6b32c9238dfec0356d", "expected": "PASS"}

[SimonSapin]

I can reproduce text_decoration_underline_subpx_a.html PASS expected FAIL, both on this branch and on master. The failure is marked expected on Linux only. I don’t know what happened exactly, but maybe we can just remove it?

I can reliably reproduce the border_radius_asymmetric_sizes_a.html failure (on this branch but not on master), but not the word-break-keep-all-005.htm one…

Let’s try it once more to see what happens.

@bors-servo try

[SimonSapin]

@bors-servo try retry

[bors-servo]

⌛ Trying commit c5bd4e1 with merge 31948ec...

[bors-servo]

⌛ Trying commit 28cc9b0 with merge 6d7c020...

[bors-servo]

💔 Test failed - linux-rel-wpt

[SimonSapin]

Screenshots bellow are of border_radius_asymmetric_sizes_ref.html, scrolled to the bottom because things are move visible there. Each squareish thing is a single element with borders of different colors on each four sides.

The rendering of these borders in Servo is messed up, but that could be an unrelated WebRender bug. What’s more worrying to me right now is that the text became bold just by upgrading the compiler.

Could it be that #[repr(C)] is missing on some type used with Freetype or Fontconfig? (Since rust-lang/rust#37429 "struct field reordering" is among the compiler changes.)

---

On Servo master, or this PR downgraded to rustc 8f02c429a 2016-12-15 (they give identical rendering):

In this PR (rustc 8f02c429a 2016-12-15):

In Firefox:

[eddyb]

FWIW servo/rust-freetype@c796066 fixes the bold problem for me (except for me it's the lack of bold on <h2> as opposed to other things being bold).

[nox]

@bors-servo try

[bors-servo]

⌛ Trying commit e73c1cb with merge 57c1dd7...

[bors-servo]

☀️ Test successful - android, arm32, arm64, linux-dev, linux-rel-css, linux-rel-wpt, mac-dev-unit, mac-rel-css, mac-rel-wpt1, mac-rel-wpt2, windows-dev

[nox]

@bors-servo r=SimonSapin,nox

[bors-servo]

📌 Commit 2cc1b84 has been approved by SimonSapin,nox

[bors-servo]

⌛ Testing commit 2cc1b84 with merge df667f1...

[bors-servo]

☀️ Test successful - android, arm32, arm64, linux-dev, linux-rel-css, linux-rel-wpt, mac-dev-unit, mac-rel-css, mac-rel-wpt1, mac-rel-wpt2, windows-dev

[Manishearth]

@jdm @Ms2ger Doesn't rust-lang/rust#37429 mean that our DOM inheritance thing might need some fixes? We used to rely on the fact that the first field is always the parent. I did update some code to use &foo.firstfieldname instead of transmuting at some point but I don't think that's all of it.

[eddyb]

@Manishearth You used #[repr(C)] with that, right? Riiiight? Anything else is irresponsible.

[Manishearth]

We didn't :) Like I said, I tried to get rid of the transmutes at some point, but it's changed a lot since then and I don't even remember.

[nox]

@eddyb I plan to do that today or tomorrow, but wasn't going to block rustup on it since the tests pass.