Add a high-level API for JS typed arrays

[jdm]

This largely a Rust-y port of Gecko's API. No codegen support yet, but it's a lot easier to work with typed arrays properly now.

r? @Ms2ger

[highfive]

Warning

• These commits modify unsafe code. Please review it carefully!

[jdm]

Requires servo/rust-mozjs#180.

[Ms2ger]

The first thing I'd ask is whether some of this could live in rust-mozjs; I sometimes dream of a fully safe api around SM.

---

Reviewed 2 of 2 files at r1, 2 of 2 files at r2, 2 of 2 files at r3, 1 of 1 files at r4.
Review status: 3 of 9 files reviewed at latest revision, 13 unresolved discussions, all commit checks successful.

---

^{components/canvas_traits/lib.rs, line 126 [r3] (raw file):}
I like issue numbers with my TODOs.

---

^{components/script/dom/bindings/typedarray.rs, line 47 [r1] (raw file):}
I don't imagine this can be private.

---

^{components/script/dom/bindings/typedarray.rs, line 137 [r1] (raw file):}
This is what Gecko does, but why not have fn new() -> Result<TypedArrayBase<T>, ()>?

---

^{components/script/dom/bindings/typedarray.rs, line 159 [r1] (raw file):}
This is unsafe; it may invoke UB by creating two objects from the same JS object.

---

^{components/script/dom/bindings/typedarray.rs, line 168 [r1] (raw file):}
Any reason not to do this in new()?

---

^{components/script/dom/bindings/typedarray.rs, line 180 [r1] (raw file):}
Limit to T: TypedArrayElement?

---

^{components/script/dom/bindings/typedarray.rs, line 250 [r1] (raw file):}
GC hazard?

---

^{components/script/dom/bindings/typedarray.rs, line 292 [r1] (raw file):}
Do we need the macro and trait?

---

^{components/script/dom/bindings/typedarray.rs, line 331 [r1] (raw file):}
Ditto

---

^{components/script/dom/bindings/typedarray.rs, line 358 [r1] (raw file):}
Does this cast adjust the length?

---

^{components/script/dom/bindings/typedarray.rs, line 364 [r1] (raw file):}
Ditto for length and UB

---

^{components/script/dom/imagedata.rs, line 51 [r2] (raw file):}
Followup to remove the argument?

---

^{components/script/dom/imagedata.rs, line 54 [r2] (raw file):}
[]

---

Comments from the review on Reviewable.io

[jdm]

The rooting can't live there, unfortunately, due to its reliance on the whole JSReflectable trait stuff and RootedJSTraceableSet.

[Ms2ger]

It probably wouldn't be a good idea to move everything except the rooting there. Maybe it could make sense to move most of trace.rs to rust-mozjs as well, though…

[jdm]

Review status: 2 of 9 files reviewed at latest revision, 13 unresolved discussions, all commit checks successful.

---

^{components/script/dom/bindings/typedarray.rs, line 47 [r1] (raw file):}
Correct. Private type exposed in public type bounds etc.

---

^{components/script/dom/bindings/typedarray.rs, line 168 [r1] (raw file):}
Consider void setValueCurveAtTime(Float32Array values, double startTime);, and code like setValueCurveAtTime(some_float32_array, /* object with getter that neuters some_float32_array */);
When this constructor moves into the bindings and the DOM methods are passed value: &Float32Array, its buffer pointer has been silently invalidated by startTime's getter and we'll now have a UAF.

---

^{components/script/dom/bindings/typedarray.rs, line 292 [r1] (raw file):}
It can be used in the future for SharedArrayBufferViews.

---

Comments from the review on Reviewable.io

[bors-servo]

☔ The latest upstream changes (presumably #6699) made this pull request unmergeable. Please resolve the merge conflicts.

[Ms2ger]

Reviewed 1 of 1 files at r5, 6 of 6 files at r6, 6 of 6 files at r7.
Review status: all files reviewed at latest revision, 4 unresolved discussions, some commit checks failed.

---

^{components/script/dom/bindings/typedarray.rs, line 110 [r7] (raw file):}
Replace by Option<(*mut _, u32)>?

---

^{components/script/dom/bindings/typedarray.rs, line 168 [r1] (raw file):}
Bah. How about we make these types entirely opaque, and give them a method that returns a more transparent struct in a Result?

---

^{components/script/dom/bindings/typedarray.rs, line 188 [r6] (raw file):}
Can't tenacious enforce that?

---

^{components/script/dom/bindings/typedarray.rs, line 442 [r6] (raw file):}
No need for the parens

---

Comments from the review on Reviewable.io

[jdm]

Review status: all files reviewed at latest revision, 4 unresolved discussions, some commit checks failed.

---

^{components/script/dom/bindings/typedarray.rs, line 166 [r1] (raw file):}
That's a really good idea.

---

^{components/script/dom/bindings/typedarray.rs, line 189 [r6] (raw file):}
Paging @Manishearth

---

Comments from the review on Reviewable.io

[Manishearth]

Review status: 2 of 9 files reviewed at latest revision, 4 unresolved discussions, all commit checks successful.

---

^{components/script/dom/bindings/typedarray.rs, line 194 [r6] (raw file):}
Yes, but keep the assert there too.

---

Comments from the review on Reviewable.io

[jdm]

Review status: 2 of 9 files reviewed at latest revision, 4 unresolved discussions, all commit checks successful.

---

^{components/script/dom/bindings/typedarray.rs, line 194 [r6] (raw file):}
I looked into this, and I'd prefer not to rely on tenacious since it requires introducing another intermediary type:

11:48 <jdm> but that makes the client code pretty clunky
11:48 <jdm> let rooter = TypedArrayRooter::new();
11:48 <jdm> let typed_array = TypedArray::from(foo, &mut rooter);
11:48 <jdm> let inited_typed_array = typed_array.init();
11:48 <jdm> let buffer = inited_typed_array.extract();
11:48 <jdm> buffer.as_slice() //woohoo an actual piece of data!

---

Comments from the review on Reviewable.io

[jdm]

@bzbarsky Do you know why the Gecko code uses JS_CallUnbarrieredObjectTracer?

[bzbarsky]

Probably because I was told to do that by terrence? Note that this is only in cases when we're dealing with a stack-based autorooter, right?

[jdm]

Yeah, that's the conclusion that mwu came to as well. We may need to do something different here, since apparently our RootedTraceableSet has some different characteristics from the CustomAutoRooter stuff.

[Ms2ger]

Reviewed 5 of 7 files at r8.
Review status: 7 of 9 files reviewed at latest revision, 5 unresolved discussions, all commit checks successful.

---

^{components/script/dom/bindings/typedarray.rs, line 182 [r8] (raw file):}
I think we should be able to avoid this unwrap.

---

^{components/script/dom/bindings/typedarray.rs, line 385 [r8] (raw file):}
Missing space

---

Comments from the review on Reviewable.io

[jdm]

Updated.

[jdm]

Note: due to the concerns about unbarriered tracing, I'm intending to implement equivalent support for SM's CustomAutoRooter instead.

[bors-servo]

☔ The latest upstream changes (presumably #6974) made this pull request unmergeable. Please resolve the merge conflicts.

[frewsxcv]

@jdm Any updates on this?

[jdm]

Not since my related mailing list post. No particular point in leaving this open.