Implement Houdini worklets #16814

asajeffrey · 2017-05-11T16:17:39Z

This PR implements the current draft Houdini Worklets specification (https://drafts.css-houdini.org/worklets/).

The implementation is intended to provide a responsive environment for worklet execution, and in particular to ensure that the primary worklet executor does not garbage collect, and does not block loading module code. The implementation does this by providing a thread pool, and performing GC and module loading in a backup thread, not in the primary thread.

./mach build -d does not report any errors
./mach test-tidy does not report any errors
These changes fix #16206
There are tests for these changes

This change is

highfive · 2017-05-11T16:17:43Z

Heads up! This PR modifies the following files:

@wafflespeanut: python/tidy/servo_tidy/tidy.py
@fitzgen: components/script/dom/webidls/Console.webidl, components/script/dom/testworkletglobalscope.rs, components/script/dom/mod.rs, components/script/dom/window.rs, components/script/dom/webidls/Window.webidl and 20 more
@KiChjang: components/script/dom/webidls/Console.webidl, components/script/dom/testworkletglobalscope.rs, components/script/dom/mod.rs, components/script/dom/window.rs, components/script/dom/webidls/Window.webidl and 17 more

highfive · 2017-05-11T16:17:45Z

Warning

These commits modify unsafe code. Please review it carefully!

asajeffrey · 2017-05-11T16:17:56Z

r? @jdm

asajeffrey · 2017-05-12T15:55:58Z

cc @tschneidereit @metajack

metajack · 2017-05-12T16:03:07Z

cc @bfgeek

asajeffrey · 2017-05-12T16:12:16Z

Some things that should probably be fixed before this lands:

the methods in globalscope which use downcasting are missing some WorkletGlobalScope cases.
threads should be named.
add time reporting code.
think about how to configure SM for worklet threads (in particular how to discourage GC).

The code is ready for review as-is though.

asajeffrey · 2017-05-12T16:21:40Z

Hi @bfgeek! This is implementation of worklets in Servo, using a thread pool to ensure worklet execution isn't blocked by GC or by module loading. Mostly the spec was pretty straightforward to implement, you'll be pleased to hear!

asajeffrey · 2017-05-12T17:01:33Z

The spec issues raised by implementing worklets (and starting on paint worklets):

Speculative evalution (w3c/css-houdini-drafts#406)
Fetching and invoking a worklet script which reports an exception (w3c/css-houdini-drafts#407)

bfgeek · 2017-05-12T17:51:41Z

cc/ @nhiroki who works on our implementation.

That's great to see! Thanks for filing spec issues.

bors-servo · 2017-05-13T13:13:42Z

☔ The latest upstream changes (presumably #16845) made this pull request unmergeable. Please resolve the merge conflicts.

jdm · 2017-05-16T15:30:39Z

components/script/dom/webidls/Window.webidl

+// https://drafts.css-houdini.org/worklets/#examples
+partial interface Window {
+   [Pref="dom.worklet.testing.enabled", SameObject] readonly attribute Worklet testWorklet;
+   [Pref="dom.worklet.testing.enabled"] DOMString? testWorkletLookup(DOMString key);


What if we made a separate DOM interface with a constructor so we don't need to include test-only fields in Window?

jdm · 2017-05-16T15:30:39Z

components/script/dom/workletglobalscope.rs

+    pub mem_profiler_chan: mem::ProfilerChan,
+    /// Chan to the time profiler
+    pub time_profiler_chan: time::ProfilerChan,
+    /// Chan to devtools


Let's use channel instead of chan in these comments.

jdm · 2017-05-16T15:30:39Z

components/script/dom/worklet.rs

+use style::thread_state;
+use swapper::Swapper;
+use swapper::swapper;
+use uuid::Uuid;


Any reason not to collapse some of these import statements from the same modules?

Minimizing the chance of annoying merge conflicts.

jdm · 2017-05-16T15:30:39Z

components/script/dom/worklet.rs

+        let global = window.upcast::<GlobalScope>();
+        unsafe {
+            WorkletBinding::Wrap(global.get_cx(), global, worklet)
+        }


We should use reflect_dom_object instead.

I don't think you can use reflect_dom_object to create globals can you?

WorkletBinding isn't a global.

Oh good point, we're in Worklet not WorkletGlobalScope.

jdm · 2017-05-16T15:30:39Z

components/script/dom/worklet.rs

+        self.worklet_id
+    }
+
+    #[allow(dead_code)]


This is dead code, just in for future-proofing. I can get rid of it.

jdm · 2017-05-16T15:30:39Z

components/script/dom/worklet.rs

+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, JSTraceable)]
+pub struct WorkletId(Uuid);
+
+known_heap_size!(0, WorkletId);


I would prefer using #[ignore_heap_size_of = "Can't measure uuid"] instead.

jdm · 2017-05-16T15:30:39Z

components/script/dom/worklet.rs

+    }
+
+    fn decrement_counter_by(&self, offset: isize) -> isize {
+        self.0.fetch_sub(offset, Ordering::AcqRel)


Why AcqRel instead of SeqCst?

Using the weakest ordering that does the job. In this case it doesn't matter a huge amount, this isn't hot code.

jdm · 2017-05-16T15:30:39Z

components/script/dom/worklet.rs

+            debug!("Finished adding script.");
+            let old_counter = pending_tasks_struct.decrement_counter_by(1);
+            if old_counter == 1 {
+                // TODO: trigger a reflow?


This happens automatically when events run in the script thread. Did you mean something more specific?

At this point, we have loaded a worklet, but not actually run it (e.g. to get a paint worklet to do its painting).

jdm · 2017-05-16T15:30:39Z

components/script/dom/worklet.rs

+    }
+
+    /// Fetch and invoke a worklet script.
+    ///g https://drafts.css-houdini.org/worklets/#fetch-and-invoke-a-worklet-script


jdm · 2017-05-16T15:31:19Z

As usual, I found these changes to be a pleasant and straightforward read. Good job!

jdm · 2017-05-16T15:34:42Z

components/script/dom/worklet.rs

+            type_: RequestType::Script,
+            destination: Destination::Script,
+            credentials_mode: credentials.into(),
+            .. RequestInit::default()


We should be using CORS mode here, per https://html.spec.whatwg.org/multipage/webappapis.html#fetch-a-single-module-script. We'll need to include a real origin, too.

CORS mode isn't a problem, but I'm not sure which origin to use here. One snag is that this is being done in a different thread, so we don't have access to the MutableOrigin of the window.

Send an ImmutableOrigin from the window as part of the initiation?

asajeffrey

Thanks for the review!

asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/webidls/Window.webidl

+// https://drafts.css-houdini.org/worklets/#examples
+partial interface Window {
+   [Pref="dom.worklet.testing.enabled", SameObject] readonly attribute Worklet testWorklet;
+   [Pref="dom.worklet.testing.enabled"] DOMString? testWorkletLookup(DOMString key);


asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/workletglobalscope.rs

+    pub mem_profiler_chan: mem::ProfilerChan,
+    /// Chan to the time profiler
+    pub time_profiler_chan: time::ProfilerChan,
+    /// Chan to devtools


asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+use style::thread_state;
+use swapper::Swapper;
+use swapper::swapper;
+use uuid::Uuid;


Minimizing the chance of annoying merge conflicts.

asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+        let global = window.upcast::<GlobalScope>();
+        unsafe {
+            WorkletBinding::Wrap(global.get_cx(), global, worklet)
+        }


I don't think you can use reflect_dom_object to create globals can you?

asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+        self.worklet_id
+    }
+
+    #[allow(dead_code)]


This is dead code, just in for future-proofing. I can get rid of it.

asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, JSTraceable)]
+pub struct WorkletId(Uuid);
+
+known_heap_size!(0, WorkletId);


asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+    }
+
+    fn decrement_counter_by(&self, offset: isize) -> isize {
+        self.0.fetch_sub(offset, Ordering::AcqRel)


Using the weakest ordering that does the job. In this case it doesn't matter a huge amount, this isn't hot code.

asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+    }
+
+    /// Fetch and invoke a worklet script.
+    ///g https://drafts.css-houdini.org/worklets/#fetch-and-invoke-a-worklet-script


asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+            type_: RequestType::Script,
+            destination: Destination::Script,
+            credentials_mode: credentials.into(),
+            .. RequestInit::default()


CORS mode isn't a problem, but I'm not sure which origin to use here. One snag is that this is being done in a different thread, so we don't have access to the MutableOrigin of the window.

asajeffrey · 2017-05-16T16:04:08Z

components/script/dom/worklet.rs

+            debug!("Finished adding script.");
+            let old_counter = pending_tasks_struct.decrement_counter_by(1);
+            if old_counter == 1 {
+                // TODO: trigger a reflow?


At this point, we have loaded a worklet, but not actually run it (e.g. to get a paint worklet to do its painting).

jdm · 2017-05-16T19:43:04Z

Squash and r=me.

asajeffrey · 2017-05-16T21:14:29Z

@bors-servo r=jdm

bors-servo · 2017-05-16T21:14:30Z

📌 Commit de96995 has been approved by jdm

bors-servo · 2017-05-17T02:42:06Z

⌛ Testing commit de96995 with merge d8f4b59...


        Auto merge of #16814 - asajeffrey:script-worklets, r=jdm

Implement Houdini worklets  This PR implements the current draft Houdini Worklets specification (https://drafts.css-houdini.org/worklets/). The implementation is intended to provide a responsive environment for worklet execution, and in particular to ensure that the primary worklet executor does not garbage collect, and does not block loading module code. The implementation does this by providing a thread pool, and performing GC and module loading in a backup thread, not in the primary thread. ---  - [x] `./mach build -d` does not report any errors - [x] `./mach test-tidy` does not report any errors - [x] These changes fix #16206 - [x] There are tests for these changes    --- This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/16814)

bors-servo · 2017-05-17T03:13:16Z

💔 Test failed - linux-rel-wpt

jdm · 2017-05-17T03:50:12Z

  ▶ Unexpected subtest result in /_mozilla/mozilla/interfaces.html:
  │ FAIL [expected PASS] Interfaces exposed on the window
  │   → assert_true: If this is failing: DANGER, are you sure you want to expose the new interface Worklet to all webpages as a property on the global? Do not make a change to this file without review from jdm or Ms2ger for that specific change! expected true got false
  │ 
  │ test_interfaces/<@http://web-platform.test:8000/_mozilla/mozilla/interfaces.js:73:7
  │ Test.prototype.step@http://web-platform.test:8000/resources/testharness.js:1406:20
  │ test@http://web-platform.test:8000/resources/testharness.js:497:9
  │ test_interfaces@http://web-platform.test:8000/_mozilla/mozilla/interfaces.js:2:3
  └ @http://web-platform.test:8000/_mozilla/mozilla/interfaces.html:13:1

We can either hide the Worklet interface behind an off-by-default pref or update the test to include the interface in the expected list.


        Implemented Houdini worklets.

asajeffrey · 2017-05-17T14:02:04Z

I added Worklet to the interfaces test. @bors-servo r=jdm

bors-servo · 2017-05-17T14:02:05Z

📌 Commit af8436c has been approved by jdm

bors-servo · 2017-05-17T21:20:43Z

⌛ Testing commit af8436c with merge ac99a48...


        Auto merge of #16814 - asajeffrey:script-worklets, r=jdm

Implement Houdini worklets  This PR implements the current draft Houdini Worklets specification (https://drafts.css-houdini.org/worklets/). The implementation is intended to provide a responsive environment for worklet execution, and in particular to ensure that the primary worklet executor does not garbage collect, and does not block loading module code. The implementation does this by providing a thread pool, and performing GC and module loading in a backup thread, not in the primary thread. ---  - [x] `./mach build -d` does not report any errors - [x] `./mach test-tidy` does not report any errors - [x] These changes fix #16206 - [x] There are tests for these changes    --- This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/16814)

bors-servo · 2017-05-17T23:10:51Z

☀️ 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-msvc-dev
Approved by: jdm
Pushing ac99a48 to master...

asajeffrey added the A-content/script label May 11, 2017

highfive assigned mbrubeck May 11, 2017

highfive added the S-awaiting-review label May 11, 2017

highfive assigned jdm and unassigned mbrubeck May 11, 2017

highfive added the S-needs-rebase label May 13, 2017

jdm requested changes May 16, 2017

View changes

jdm added S-needs-code-changes and removed S-awaiting-review labels May 16, 2017

jdm reviewed May 16, 2017

View changes

asajeffrey reviewed May 16, 2017

View changes

highfive added S-awaiting-review and removed S-needs-code-changes labels May 16, 2017

asajeffrey force-pushed the asajeffrey:script-worklets branch from 10a9a04 to 6a4b4e6 May 16, 2017

jdm added S-needs-squash and removed S-awaiting-review S-needs-rebase labels May 16, 2017

jdm approved these changes May 16, 2017

View changes

highfive added S-awaiting-merge and removed S-needs-squash labels May 16, 2017

asajeffrey force-pushed the asajeffrey:script-worklets branch from 67eed3e to de96995 May 16, 2017

highfive added S-awaiting-review and removed S-awaiting-merge labels May 16, 2017

highfive added S-awaiting-merge and removed S-awaiting-review labels May 16, 2017

highfive added S-tests-failed and removed S-awaiting-merge labels May 17, 2017

Implemented Houdini worklets.

Loading status checks…

af8436c

asajeffrey force-pushed the asajeffrey:script-worklets branch from de96995 to af8436c May 17, 2017

highfive added S-awaiting-review and removed S-tests-failed labels May 17, 2017

highfive added S-awaiting-merge and removed S-awaiting-review labels May 17, 2017

bors-servo mentioned this pull request May 17, 2017

Update app_units to 0.4.1 #16914

Merged

bors-servo merged commit af8436c into servo:master May 17, 2017
3 checks passed

3 checks passed

continuous-integration/appveyor/pr AppVeyor build succeeded
Details

continuous-integration/travis-ci/pr The Travis CI build passed
Details

homu Test successful
Details

highfive removed the S-awaiting-merge label May 17, 2017

bors-servo mentioned this pull request May 17, 2017

Implement WebGL extensions #16893

Merged

3 of 5 tasks complete

servo / servo

Join GitHub today

Implement Houdini worklets #16814

Implement Houdini worklets #16814

Conversation

asajeffrey commented May 11, 2017 • edited by larsbergstrom

highfive commented May 11, 2017

highfive commented May 11, 2017

asajeffrey commented May 11, 2017

asajeffrey commented May 12, 2017

metajack commented May 12, 2017

asajeffrey commented May 12, 2017

asajeffrey commented May 12, 2017

asajeffrey commented May 12, 2017

bfgeek commented May 12, 2017

bors-servo commented May 13, 2017

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

jdm commented May 16, 2017

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

asajeffrey left a comment

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

jdm commented May 16, 2017

asajeffrey commented May 16, 2017

bors-servo commented May 16, 2017

bors-servo commented May 17, 2017

bors-servo commented May 17, 2017

jdm commented May 17, 2017

asajeffrey commented May 17, 2017

bors-servo commented May 17, 2017

bors-servo commented May 17, 2017

bors-servo commented May 17, 2017

asajeffrey commented May 11, 2017 •

edited by larsbergstrom