No longer allow clipping items by complex clips

[mrobinson]

This is part of an API simplification which should hopefully lead to
less code complexity in the future.

Fixes #1742.

---

This change is 

[mrobinson]

Gecko try job is here: #2512

[staktrace]

Presumably you meant https://treeherder.mozilla.org/#/jobs?repo=try&revision=ad251afbee824f371097a3ea2c378faa31043be7

[bors-servo]

☔ The latest upstream changes (presumably #2507) made this pull request unmergeable. Please resolve the merge conflicts.

[mrobinson]

@staktrace Thanks for provided the correct link. I was in a bit of a rush this morning.

The branch should be rebased and ready to review now.

[glennw]

Is there a Gecko / Servo patch you could post here for the API update, that we can use while doing try runs in the interim?

[kvark]

Looks great! A few questions, otherwise 🚢 it

---

Reviewed 12 of 12 files at r1.
Review status: all files reviewed at latest revision, 2 unresolved discussions.

---

direct-composition/src/main_windows.rs, line 135 at r1 (raw file):

        let rect = euclid::TypedRect::new(euclid::TypedPoint2D::zero(), layout_size);
        builder.push_rect(
            &api::PrimitiveInfo::with_clip(

any reason we aren't doing push_clip_id with this complex clip?

---

wrench/src/rawtest.rs, line 606 at r1 (raw file):

        // Add a rounded 100x100 rectangle at 200,0.
        let rect = LayoutRect::new(LayoutPoint::new(200., 0.), LayoutSize::new(100., 100.));

any reason this is removed?

---

Comments from Reviewable

[mrobinson]

Review status: all files reviewed at latest revision, 2 unresolved discussions.

---

direct-composition/src/main_windows.rs, line 135 at r1 (raw file):

Previously, kvark (Dzmitry Malyshau) wrote…

any reason we aren't doing push_clip_id with this complex clip?

It looked like the rounded rectangle here wasn't important for the example, but I can add it back. I didn't want to make too many changes here because I don't have a Windows environment ready for testing. I can spend a little time getting the rendered content back to its original state though.

---

wrench/src/rawtest.rs, line 606 at r1 (raw file):

Previously, kvark (Dzmitry Malyshau) wrote…

any reason this is removed?

I added this one to explicitly test items with their own rounded rectangle local clips. Since I removed that feature, I also removed the test for it. Directly below is the same test, but using separate clipping nodes.

---

Comments from Reviewable

[mrobinson]

@glennw I don't think there are any code changes necessary for Gecko (it always uses the PrimitiveInfo constructors) and here is a branch with the changes necessary for Servo: https://github.com/mrobinson/servo/tree/changes-for-localclip-changes. I haven't tested this one, but I'm fairly certain there is no change in behavior.

[kvark]

@mrobinson looks like this is ready to go?

[mrobinson]

@kvark Yep. Is it okay to land it?

[kvark]

@mrobinson well, the try looks good, code is approved by me, and Gecko doesn't (seemingly?) appear to need any changes. Shipit!

[mrobinson]

@kvark Thanks!

@bors-servo r=kvark

[bors-servo]

📌 Commit d398096 has been approved by kvark

[bors-servo]

⌛ Testing commit d398096 with merge 0291c44...

[glennw]

@bors-servo retry

[bors-servo]

⌛ Testing commit d398096 with merge 45a3b7b...

[bors-servo]

💔 Test failed - status-travis

[glennw]

Blocked by #2522.

[mrobinson]

@bors-servo retry

[bors-servo]

⌛ Testing commit d398096 with merge e0fbae7...

[bors-servo]

☀️ Test successful - status-appveyor, status-taskcluster, status-travis
Approved by: kvark
Pushing e0fbae7 to master...