gfx: Implement display-list-based invalidation.

[pcwalton]

This ensures that we don't throw away all tiles if only a few of them need to be repainted. You can test this with something like servo -s 64 -Z paint-flashing tests/html/dlbi.html and mousing over the button.

Requires servo/rust-layers#167 to land first.

r? @glennw

[hoppipolla-critic-bot]

Critic review: https://critic.hoppipolla.co.uk/r/5123

This is an external review system which you may optionally use for the code review of your pull request.

In order to help critic track your changes, please do not make in-place history rewrites (e.g. via git rebase -i or git commit --amend) when updating this pull request.

[waddlesplash]

Just curious: do you have any numbers for how much faster this can be / is in practice?

[bors-servo]

☔ The latest upstream changes (presumably #6216) made this pull request unmergeable. Please resolve the merge conflicts.

[glennw]

Review status: all files reviewed, 4 unresolved discussions, some commit checks failed.
Reviewed files:

• components/compositing/compositor.rs @ r2
• components/compositing/compositor_layer.rs @ r1
• components/gfx/display_list/invalidation.rs @ r1
• components/gfx/display_list/mod.rs @ r2
• components/gfx/paint_task.rs @ r2
• components/layout/display_list_builder.rs @ r2
• components/msg/compositor_msg.rs @ r1
• components/util/range.rs @ r1
• tests/html/dlbi.html @ r1

---

^{components/compositing/compositor.rs, line 658 [r2] (raw file):}
Use unwrap or expect?

---

^{components/compositing/compositor.rs, line 680 [r2] (raw file):}
unwrap or expect?

---

^{components/compositing/compositor.rs, line 1241 [r2] (raw file):}
Perhaps the function should be "invalidate_and_send ... " to make it clear from reading calling code?

---

^{components/gfx/display_list/invalidation.rs, line 316 [r2] (raw file):}
Is there a reason not to destructure the base item as you have for the items below?

---

Comments from the review on Reviewable.io

[glennw]

@pcwalton This seems to break the 3d transforms demos (stuff clips and doesn't update). I'll do some more investigation. tests/ref/iframe/positioning_margin is also failing with this PR (on my laptop at least).

[glennw]

@pcwalton I looked briefly into this - it appears to suffer the same problems as we currently have with clipping on 3d layers. How about updating this PR to disable DLBI on 3d layers (like we do for clipping currently)? That way, we can get this merged since it is useful for 2d layers anyway, and when I get to supporting clipping for 3d layers, I can also fix up the DLBI on 3d layers then?

[bors-servo]

☔ The latest upstream changes (presumably #6434) made this pull request unmergeable. Please resolve the merge conflicts.

[pcwalton]

Rebased and attempted to turn this off for 3D layers.

@glennw How does this look?

[bors-servo]

☔ The latest upstream changes (presumably #6576) made this pull request unmergeable. Please resolve the merge conflicts.

[glennw]

@pcwalton Looks good to me, just needs a rebase.

[pcwalton]

@bors-servo: r=glennw

[pcwalton]

@bors-servo: retry

They may be intermittent. I'm going to retry to see if they are.

[bors-servo]

⚡ Previous build results for android, gonk, linux2, mac1 are reusable. Rebuilding only linux1, linux3, mac2, mac3...

[bors-servo]

💔 Test failed - mac2

[pcwalton]

OK, so positioning_margin is not intermittent, but border_spacing is. That's what I had suspected/worried about. Is the border_spacing one pre-existing, I wonder?

[pcwalton]

Actually, IIRC they're both intermittent; it's just that the iframe ones fail more often. I worry that DLBI is making our known compositor races worse :(

[bors-servo]

☔ The latest upstream changes (presumably #6630) made this pull request unmergeable. Please resolve the merge conflicts.

[pcwalton]

@bors-servo: r=glennw

[bors-servo]

📌 Commit 7667944 has been approved by glennw

[bors-servo]

⌛ Testing commit 7667944 with merge a26a525...

[bors-servo]

💔 Test failed - linux1

[metajack]

/home/servo/buildbot/slave/linux1/build/tests/unit/util/lib.rs:5:35: 5:43 warning: unused or unknown feature, #[warn(unused_features)] on by default
/home/servo/buildbot/slave/linux1/build/tests/unit/util/lib.rs:5 #![feature(plugin, custom_derive, heap_api)]
                                                                                                   ^~~~~~~~
   Compiling servo v0.0.1 (file:///home/servo/buildbot/slave/linux1/build/components/servo)
../../tests/reftest.rs:46:26: 46:32 warning: use of deprecated item: superseded by split_first, #[warn(deprecated)] on by default
../../tests/reftest.rs:46     let mut parts = args.tail().split(|e| &**e == "--");
                                                   ^~~~~~
../../tests/reftest.rs:104:22: 104:42 warning: use of deprecated item: use process::exit instead, #[warn(deprecated)] on by default
../../tests/reftest.rs:104         Ok(false) => env::set_exit_status(1), // tests failed
                                                ^~~~~~~~~~~~~~~~~~~~
../../tests/reftest.rs:105:19: 105:39 warning: use of deprecated item: use process::exit instead, #[warn(deprecated)] on by default
../../tests/reftest.rs:105         Err(_) => env::set_exit_status(2),    // I/O-related failure
                                             ^~~~~~~~~~~~~~~~~~~~

[pcwalton]

That was the reftest I fixed… maybe it's a Linux-only failure now?

[bors-servo]

☔ The latest upstream changes (presumably #6863) made this pull request unmergeable. Please resolve the merge conflicts.

[glennw]

@pcwalton Should we close this and re-open it when you have time to look at this again?

[pcwalton]

Yeah, sure.