support color masking of images

[lsalzman]

This is to help support the necessary fixes for Gecko bug https://bugzilla.mozilla.org/show_bug.cgi?id=1479196

Tofu glyphs will be drawn out of an atlas, and to avoid the need to have to generate a new atlas for every single color or opacity combination, this just allows push_image to specify a color which can then be used for masking or opacity.

Most of the plumbing is already there, just some parameters need to be added to support this.

---

This change is 

[gw3583]

Reviewed 17 of 17 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @lsalzman)

---

webrender/src/batch.rs, line 726 at r1 (raw file):

                                                let prim_header_index = prim_headers.push(&prim_header, [
                                                    uv_rect_address.as_int(),
                                                    (ShaderColorMode::Image as i32) << 16 |

How did this used to work previously and why do we need to change it now on types other than images?

[gw3583]

This basically looks good, but I'm somewhat confused about the shader color mode change on some of the batching code. If I'm reading it correctly it looks like it was wrong previously, and is now right? But if so, how come we don't have test failures from those? I feel like I'm mis-reading part of the patch :)

We'll also need a try run before merging.

[lsalzman]

It worked previously because we always passed in the premultiplied color white to the shader. The old shader color mode only accessed color.a, which if you squint at it, behaves exactly the same as the premultiplied color white, but would break if any other color was passed. So I had to change the color mode to access the RGB components of the color as well, not just the alpha component.

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @lsalzman)

[gw3583]

@lsalzman Ah that makes sense - so it was broken, but not in any user-visible way previously 💩

[gw3583]

r=me once we have a try result posted here.

[lsalzman]

Try run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=f0115f48d25d10d55480f624368ce02e6d53113f

Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @lsalzman)

[gw3583]

That looks like a lot of orange on the try - but it's also not the try configuration I normally use, so maybe those are expected.

I leave it to your wise judgment :)

@bors-servo delegate+

[bors-servo]

✌️ @lsalzman can now approve this pull request

[lsalzman]

Cleaner try run: https://treeherder.mozilla.org/#/jobs?repo=try&revision=d453838d0586955f94608413a8533b6770f708be

[lsalzman]

@bors-servo r+

[bors-servo]

📌 Commit c75c32d has been approved by lsalzman

[bors-servo]

⌛ Testing commit c75c32d with merge f087a9f...

[bors-servo]

☀️ Test successful - status-appveyor, status-taskcluster
Approved by: lsalzman
Pushing f087a9f to master...