Previously, senders and receivers to different kinds of event loops (the
main `ScriptThread`, different types of workers) used a rust `trait`
mechanism to implement dynamic behavior. This led to having many unused
implementations of this `trait`. This change moves to using an `enum`
based approach for these senders and receivers and removes all of the
dead code.
In addition, to allowing for use of rust's dead code detection, it
simplifies the code a great deal. All of these generic senders and
receivers are moved to the `messaging.rs` file and given proper
documentation.
Finally, empty an `JSTraceable` implementation is made for all
crossbeam `Sender<...>`s to avoid having to manually skip them everytime
they are included in structs. The pre-existing empty `MallocSizeOf`
implementation is used more thoroughly.
Other unecessary wrappers around these senders and receivers are removed
as well.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Expose two new helpers and start using them as much as possible.
- `NodeTraits::owner_global`: which gets the `GlobalScope` that currenty
owns a `Node`. This may be different than `.global()` in the case that
the `Node` was adopted by a different `Document`.
- `Window::as_global_scope`: A helper to avoid having to cast so much
when treating a `Window` like a `GlobalScope`.
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>
This is a simplification of the internal `TaskQueue` API that moves the
`TaskManager` to the `GlobalScope` itself. In addition, the handling of
cancellers is moved to the `TaskManager` as well. This means that no
arguments other than the `task` are necessary for queueing tasks, which
makes the API a lot easier to use and cleaner.
`TaskSource` now also keeps a copy of the canceller with it, so that
they always know the proper way to cancel any tasks queued on them.
There is one complication here. The event loop `sender` for dedicated
workers is constantly changing as it is set to `None` when not handling
messages. This is because this sender keeps a handle to the main
thread's `Worker` object, preventing garbage collection while any
messages are still in flight or being handled. This change allows
setting the `sender` on the `TaskManager` to `None` to allow proper
garbabge collection.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
This puts a few commonly used `Node` helpers into a trait (`NodeTraits`)
and gives them more descriptive names and documentation. The renames:
- `document_from_node` -> `NodeTraits::owner_document`
- `window_from_node` -> `NodeTraits::owner_window`
- `stylesheets_owner_from_node<T:` -> `NodeTraits::stylesheet_list_owner`
- `containing_shadow_root` -> `NodeTraits::containing_shadow_root`
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Instead of creating a type for each `TaskSource` variety have each `TaskSource`
hold the same kind of sender (this was inconsistent before, but each
sender was effectively the same trait object), a pipeline, and a
`TaskSourceName`. This elminates the need to reimplement the same
queuing code for every task source.
In addition, have workers hold their own `TaskManager`. This allows just
exposing the manager on the `GlobalScope`. Currently the `TaskCanceller`
is different, but this will also be eliminated in a followup change.
This is a the first step toward having a shared set of `Sender`s on
`GlobalScope`.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Makes use of the is_initial_about_blank property on Document in order to
determine whether the navigation should create a history entry, or replace
the current history entry.
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
Create two new data structures in the `script` crate to hold senders and
receiver:
- `ScriptThreadSenders`: holds all outgoing channels from the
`ScriptThread` including a channel to the `ScriptThread` itself. The
ultimate goal with this is to reduce duplication by giving a boxed
version of this this to `Window`s.
- `ScriptThradReceivers`: holds all incoming channels to the
`ScriptThread`. This isn't cloenable like the senders. This is used to
abstract away `recv()` and `try_recv()` methods used to make the
`ScriptThread` event loop easier to read.
In addition:
- The many duplicated `ScriptThread` self-senders for the `TaskManager`
have been removed and, in general, a lot of boilerplate is removed as
well.
- Visibilty of all methods affected by this change is changed to
`pub(crate)` in order to take advantage of dead code detection. Some
dead code produced from macros is removed.
- Some conversion code is refactord into implementations of the `From`
trait.
- The names of channels uses a standard "sender" and "receiver" naming
as well as trying to be descriptive of where they go in `ScriptThread`
as well as `InitialScriptState`
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
This change creates a new struct `IFrameCollection` that is used to
cache the list of `<iframe>`s in a `Document` as long as the
`Document`'s DOM has not changed. This prevent constantly iterating the
entire DOM during *update the rendering*, which runs up to 60 times per
second as well as for other operations.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Expose a new constructor on `HTMLCollection`, `new_with_filter_fn`, that
filters elements using a simple static function -- a common pattern.
This allows more easily creating this `HTMLCollection` without having to
create a new data structure. Since the constructor takes a static
function, no data should be captured preventing garbage collection
issues.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
These all happen now in *update the rendering*, typically after the
message that triggered this code is processed, though in two cases
reflow needs to be triggered explicitly. This makes `ReflowReason`
redundant though perhaps `ReflowCondition` can be expanded later to give
more insight into why the page is dirty.
- Handling of the "reflow timer" concept has been explained a bit more via
data structures and rustdoc comments.
- Theme changes are cleaned up a little to simplify what happens during
reflow and to avoid unecessary reflows when the theme doesn't change.
Notably, layout queries and scrolling still trigger normal reflows and
don't update the rendering. This needs more investigation as it's
unclear to me currently whether or not they should update the rendering
and simply delay event dispatch or only reflow.
In general, this is a simplfication of the code.
Fixes#31871.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
A rendering opportunity is now unconditionally triggered by handling IPC
messages in the `ScriptThread`, unless animations are running in which
case it's driven by the compositor. We can now remove calls to
`note_rendering_opportunity` and `rendering_opportunity`.
There is one tricky case, which is when a promise completion during a
microtask checkpoint dirties the page again. In this case we need to
trigger a new rendering opportunity, unless animations are running. In
a followup change, when not driven by the compositor, rendering
opportunities will be driven by a timed task, meaning we can remove this
workaround.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
This removes two explicit calls to reflow to detect rAFs that do not
modify the DOM and to trigger reflows when the page isn't dirty. This
can cause extra reflows, especially when animations are running. This
change removes them, relying on *update the rendering* to properly
trigger reflows, shortly after running rAF callbacks and after
animations are updated.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
The `pending reflow` concept isn't necessary now that *update the
rendering* is taking care of triggering reflows at the correct time.
`Window::reflow` already avoids reflows if the page is not dirty, so
pending reflows is now just an extraneous check as long as *update the
rendering* runs properly.
This change also removes some explicit reflows, which now wait until the
appropriate moment during *update the rendering*. This should remove
some extra reflows that are not necessary.
Servo needs some way to track that resizing the web view needs to
re-layout due to the initial containing block changing. Move handling
of `Document::needs_paint` to the script thread and use this, expanding
the rustdoc to explain what it is for a bit more clearly.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
This changes fixes two issues:
1. A reflow of all `Document`s currently done unconditionally after
receving IPC messages in the `ScriptThread`. Reflowing without first
updating the animation timeline can lead to transitions finshing as
soon as they start (because it looks like time advancement is
measaured between calls to `update-the-rendering`).
2. Fix an issue where not all `Pipeline`s were updated during *update
the rendering*. The previous code only took into account top level
frames and their children. It's not guaranteed that a particular
`ScriptThread` is managing any top level frames, depending on the
origens of those frames. We should update the rendering of those
non-top-level iframes regardless.
The new code attempts to order the frames according to the
specification as much as possible without knowing the entire frame
tree, without skipping any documents managed by the `ScriptThread` in
question.
In addition, `Documents` is pulled out the `script_thread.rs` and
renamed to `DocumentCollection`.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
* Add XPath parser/evaluator
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* Correctly annotate XPathEvaluator IDL
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review]: have bindings pass in `can_gc`
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review]: add docstrings
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review]: implement PartialEq for Value for readability
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review]: add docstrings for CoreFunctions
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review]: simplify node test code
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review]: add unit tests for string handling xpath functions
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* put xpath features behind dom.xpath.enabled pref
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review] remove rstest and insta dev-deps
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* update wpt test expectations
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* [PR review]: tweak metadata files
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
* update wpt test expectations AGAIN
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
---------
Signed-off-by: Ville Lindholm <ville@lindholm.dev>
Before, `update_the_rendering()` would update all animations for all
Documents once per-Document. Apart from being generally wrong, the
specification says this should be done once per-Document. This
theoretically means that `update_the_rendering()` is just doing less
work every time it runs.
In addition:
- Don't redirty animations nodes when running rAF callbacks. They
should already be dirty when animations are updated.
- Perform a microtask checkpoint while updating animations as dictacted
by the specification.
- Update comments to reflect the specification text.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
* Remove the referrer policy from document and rely on its policy container
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* Make ReferrerPolicy non-optional, instead using a new enum value to represent the empty string case
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* Fix clippy issue
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* Fix usage of Option<ReferrerPolicy> in unit test
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
---------
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* Implement PolicyContainer
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* implement small parts of fetch that interact with policy container
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* fix: allow policy container's csp list to be unset
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* fix: use the correct default policy when parsing from a token
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
---------
Signed-off-by: Shane Handley <shanehandley@fastmail.com>
* Make CanGc non-sendable, and add documentation.
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
* Update CanGc usage to fix usages that were moved between threads/tasks.
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
---------
Signed-off-by: Josh Matthews <josh@joshmatthews.net>
* More files with CanGc fixes
Signed-off-by: L Ashwin B <lashwinib@gmail.com>
* removed the can_gc inside !task
Signed-off-by: L Ashwin B <lashwinib@gmail.com>
---------
Signed-off-by: L Ashwin B <lashwinib@gmail.com>
Instead of creating a `ROUTER` for each fetch, create a fetch thread
which handles all incoming and outcoming fetch requests. Now messages
involving fetches carry a "request id" which indicates which fetch is
being addressed by the message. This greatly reduces the number of file
descriptors used by fetch.
In addition, the interface for kicking off fetches is simplified when
using the `Listener` with `Document`s and the `GlobalScope`.
This does not fix all leaked file descriptors / mach ports, but greatly
eliminates the number used. Now tests can be run without limiting
procesess on modern macOS systems.
Followup work:
1. There are more instances where fetch is done using the old method.
Some of these require more changes in order to be converted to the
`FetchThread` approach.
2. Eliminate usage of IPC channels when doing redirects.
3. Also eliminate the IPC channel used for cancel handling.
4. This change opens up the possiblity of controlling the priority of
fetch requests.
Fixes#29834.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
* CanGc fixes and checks in multiple files
Signed-off-by: L Ashwin B <lashwinib@gmail.com>
* CanGc fixes in storageevent.rs
Signed-off-by: L Ashwin B <lashwinib@gmail.com>
---------
Signed-off-by: L Ashwin B <lashwinib@gmail.com>
* Implement encoding-parse-a-url algorithm
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
* Parse url with encoding when setting href on HtmlAnchorElement
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
* Update WPT expectations
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
---------
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
