This change has two parts which depend on each other:
1. An early exit in the layout process, which allows for skipping
display list construction entirely when nothing would change.
2. A simplification and unification of the way that "fake" animation
frames are triggered. Now this happens on an entire ScriptThread at
once and is based on whether or not any Pipeline triggered a display
list update.
Animations are never canceled in the compositor when the Pipeline
isn't updating, instead the fake animation frame is triggered far
enough in the future that an unexpected compositor tick will cancel
it. This could happen, for instance, if some other Pipeline in some
other ScriptThread produced a new display list for a tick. This makes
everything simpler about these ticks.
The goal is that in a future change the ScriptThread-based animation
ticks will be made more generic so that they can throttle the number of
"update the rendering" calls triggered by script.
This should make Servo do a lot less work when moving the cursor over a
page. Before it would constantly produce new display lists.
Fixes: #17029.
Testing: This should not cause any web observable changes. The fact that
all WPT tests keep passing is the test for this change.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Co-authored-by: Oriol Brufau <obrufau@igalia.com>
Profiling speedometer 2.1 on a quad core Intel N100, we spend quite some
time in Vec::insert when adding new timers. This is mostly because one
of the tests creates a large number of timers (> 10k).
Switching to a VecDeque solves that and gets a ~2% score improvement on
that device.
Signed-off-by: webbeef <me@webbeef.org>
Disabling the TestBinding-related bindings and implementations saves 2mb
in a release build. Also, we lost the related test preferences when we
turned the layout-2020 tests on, so the tests have all been failing for
a long time.
Testing: Existing tests in CI.
---------
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
This turned out to be a full rabbit hole. The new header
is parsed in the new `parse_csp_list_from_metadata` which
sets `disposition` to `report.
I was testing this with
`script-src-report-only-policy-works-with-external-hash-policy.html`
which was blocking the script incorrectly. Turns out that there
were multiple bugs in the CSP library, as well as a missing
check in `fetch` to report violations.
Additionally, in several locations we were manually reporting csp
violations, instead of the new `global.report_csp_violations`. As
a result of that, they would double report, since the report-only
header would be appended as a policy and now would report twice.
Now, all callsides use `global.report_csp_violations`. As a nice
side-effect, I added the code to set source file information,
since that was already present for the `eval` check, but nowhere
else.
Part of #36437
Requires servo/rust-content-security-policy#5
---------
Signed-off-by: Tim van der Lippe <tvanderlippe@gmail.com>
Signed-off-by: Tim van der Lippe <TimvdLippe@users.noreply.github.com>
Move parsing of Refresh values to Document.
Send Refresh header to Document and have meta tags reuse the logic.
I transplanted the existing Regex and made some updates so that it
passed all the existing parser tests.
I added the comments that made sense but it is not very clean to add
many comments within the regex.
Testing: There are existing WPT tests
---------
Signed-off-by: Sebastian C <sebsebmc@gmail.com>
Flatten and simplify Servo's preferences code. In addition, have both
preferences and options passed in as arguments to `Servo::new()` and
make sure not to use the globally set preferences in `servoshell` (as
much as possible now).
Instead of a complex procedural macro to generate preferences, just
expose a very simple derive macro that adds string based getters and
setters.
- All command-line parsing is moved to servoshell.
- There is no longer the concept of a missing preference.
- Preferences no longer have to be part of the resources bundle because
they now have reasonable default values.
- servoshell specific preferences are no longer part of the preferences
exposed by the Servo API.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
No longer hide errors while queueing tasks on the main thread. This
requires creating two types of `TaskSource`s: one for the main thread
and one that can be sent to other threads. This makes queueing a bit
more efficient on the main thread and more importantly, no longer hides
task queue errors.
Fixes#25688.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Co-authored-by: Mukilan Thiyagarajan <mukilan@igalia.com>
Before each `OneShotTimers` would have a `TimerListener` for its
lifetime, holding the timer `TaskSource`. The issue with this is that
the `TaskSource` for dedicated workers keeps the main thread object
alive, so as long as the `OneShotTimers` alive (until the worker thread
exists), the main thread object would never be garbage collected.
This change makes the creation of the listener on-demand, avoiding the
long-lived handle to the main thread object and slightly simplifying
`OneShotTimers` at the expense of some more operations when scheduling a
timer.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Before all timers were managed by the Constellation process, meaning
that they had to trigger IPC calls to be scheduled and fired. Currently,
timers are only used in the `ScriptThread`, so it makes sense that they
are per-process.
This change restores the timer thread functionality that existed before
avoided entirely. Completion is done using a callback that is sent to
the timer thread similarly to how fetch is done. This allows reusing the
existing task queue without making any new channels.
Fixes#15219.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
* Make generated bindings that return a WebIDL `any` value use out parameters.
Returning raw JSVal values makes it easier to create GC hazards in code
that calls these methods. Accepting a MutableHandle argument instead
ensures that the values are rooted by the caller.
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
* Update mozjs.
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
* Fix clippy warnings.
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
---------
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
This switches to using `std::time` types for DOM timer operations, which
allows removing our custom time units in favor of `Duration`.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
* Fixed some clippy warnings in components/script/timers.rs
* Formatted changes in components/script/timers.rs
* Updated changes in components/script/timers.rs
* Updated Default implementation of JsTimers in components/script/timers.rs
* UPDATED DEFAULT METHOD IMPLEMENTATION OF JsTimers in components/script/timers.rs
* Add `no_trace` option to JSTraceable derive
* NoTrace wrapper
* Port some types to no_trace schematics
* Fixing my unsafe mistakes (not tracing traceables)
* Add docs & safety guards for no_trace
Safety guards (trait shenanigans) guarantees safety usage of `no_trace`
* Port canvas_traits to no_trace
* Port servo_media to no_trace
* Port net_traits to no_trace
* Port style to no_trace
* Port webgpu to no_trace
* Port script_traits to no_trace
* Port canvas_traits, devtools_traits, embedder_traits, profile_traits to no_trace
* unrooted_must_root lint in seperate file
* Add trace_in_no_trace_lint as script_plugin
* Composable types in must_not_have_traceable
* Introduced HashMapTracedValues wrapper
* `HashMap<NoTrace<K>,V>`->`HashMapTracedValues<K,V>`
* Port rest of servo's types to no_trace
* Port html5ever, euclid, mime and http to no_trace
* Port remaining externals to no_trace
* Port webxr and Arc<Mutex<_>>
* Fix spelling in notrace doc
Web developers can use `Dynamic Import` in a classic script; thus, we
need to save the script's private reference so that we can reuse it when
we're going to fetch a dynamic import module for a classic script.
Besides, because it's possible to use different executing context for a
dynamic import module (like `dynamic-import/string-compilation-other-document.html` WPT test),
we can't initialize a module owner at the timing of `SetScriptPrivate`;
thus, if the private module script doesn't hold an owner, we'll use a
DynamicImport owner for it.
Servo currently uses `heapsize`, but Stylo/Gecko use `malloc_size_of`.
`malloc_size_of` is better -- it handles various cases that `heapsize` does not
-- so this patch changes Servo to use `malloc_size_of`.
This patch makes the following changes to the `malloc_size_of` crate.
- Adds `MallocSizeOf` trait implementations for numerous types, some built-in
(e.g. `VecDeque`), some external and Servo-only (e.g. `string_cache`).
- Makes `enclosing_size_of_op` optional, because vanilla jemalloc doesn't
support that operation.
- For `HashSet`/`HashMap`, falls back to a computed estimate when
`enclosing_size_of_op` isn't available.
- Adds an extern "C" `malloc_size_of` function that does the actual heap
measurement; this is based on the same functions from the `heapsize` crate.
This patch makes the following changes elsewhere.
- Converts all the uses of `heapsize` to instead use `malloc_size_of`.
- Disables the "heapsize"/"heap_size" feature for the external crates that
provide it.
- Removes the `HeapSizeOf` implementation from `hashglobe`.
- Adds `ignore` annotations to a few `Rc`/`Arc`, because `malloc_size_of`
doesn't derive those types, unlike `heapsize`.
I don't want to do such a gratuitous rename, but with all the other types
now having "Dom" as part of their name, and especially with "DomOnceCell",
I feel like the other cell type that we already have should also follow
the convention. That argument loses weight though when we realise there
is still DOMString and other things.
