This is the first part of cleaning up unused WebRender resources.
Currently this only cleans up web font resources, but a more
full-featured implementation in the future could also clean up unused
system fonts.
Fixes#32345.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Co-authored-by: Mukilan Thiyagarajan <mukilan@igalia.com>
The `gfx_traits` crate is gradually become just about text and fonts and
this is one of the few things there that is standing in the way of this.
Eventually `webrender_traits` will be the general cross-process
compositor API, so this sort of makes sense as well.
Instead of bouncing messages from the compositor to script and then to
layout, just have script call methods on Layout. Additionally, we do not
need to send any followup messages to script for these messages. Instead
just execute code after calling the method on Layout.
* Move WebRender related types to `webrender_traits`
This refactor moves several WebRender related types
from `compositing_traits`, `script_traits` and `net_traits`
crates to the `webrender_traits` crate.
This change also moves the `Image` type and associated
function out of `net_traits` and into the `pixels` crate.
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* Move `script_traits::WebrenderIpcSender` to `webrender_traits::WebRenderScriptApi`
---------
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
For a long time, `gfx_traits` has held a lot of things unrelated to graphics
and also unrelated to the `gfx` crate (which is mostly about fonts).
This is a cleanup which does a few things:
1. Move non `gfx` crate things out of `gfx_traits`. This is important in
order to prevent dependency cycles with a different integration between
layout, script, and fonts.
2. Rename the `msg` crate to `base`. It didn't really contain anything
to do with messages and instead mostly holds ids, which are used
across many different crates in Servo. This new crate will hold the
*rare* data types that are widely used.
Details:
- All BackgroundHangMonitor-related things from base to a new
`background_hang_monitor_api` crate.
- Moved `TraversalDirection` to `script_traits`
- Moved `Epoch`-related things from `gfx_traits` to `base`.
- Moved `PrintTree` to base. This should be widely useful in Servo.
- Moved `WebrenderApi` from `base` to `webrender_traits` and renamed it
to `WebRenderFontApi`.
* clippy: Squish warnings and errors in gfx
warning: redundant closure (gfx/font.rs:415:18)
warning: useless conversion to the same type (gfx/font.rs:534:9)
warning: the following explicit lifetimes could be elided: 'a (gfx/font.rs:619:16)
error: this loop never actually loops (gfx/font_cache_thread.rs:112:9)
warning: this expression creates a reference which is immediately dereferenced by the compiler (gfx/font_cache_thread.rs:229:51)
warning: redundant closure (gfx/font_cache_thread.rs:551:18)
3 instances of:
warning: casting integer literal to `f64` is unnecessary (gfx/platform/freetype/font_list.rs:271-273)
* clippy: methods called `from_*` usually take no `self`
It reports that by standard convention, from_* methods should not take any `&self` parameter
* clippy: you should consider adding a `Default` implementation
It reports that public types with a pub fn new() -> Self should have a Default implementation since they can be constructed without arguments
* clippy: casting to the same type is unnecessary (`f32` -> `f32`)
* clippy: use of `unwrap_or_else` to construct default value
* clippy: methods called `is_*` usually take `self` by mutable reference or `self` by reference or no `self`
* clippy: manual `!RangeInclusive::contains` implementation
contains expresses the intent better and has less failure modes (such as fencepost errors or using || instead of &&)
* clippy: this function has an empty `#[must_use]` attribute, but returns a type already marked as `#[must_use]`
* clippy: Fix some new warnings
warning: this `if` statement can be collapsed (gfx/font.rs:468:130)
warning: this lifetime isn't used in the impl (gfx/platform/freetype/font.rs:341:6)
warning: field assignment outside of initializer for an instance created with Default::default() (compositor.rs:881:17)
This is a speculative fix for #32202, which I cannot reproduce
consistently. This prevents handling any embedder events while
shutting down or after shutdown is complete. It doesn't make sense to do
this as the compositor can be in a very inconsistent state during this
time and lead to panics.
Fixes ##32202.
This change reworks the way that platform fonts are created and
descriptor data is on `FontTemplate` is initialized.
The main change here is that platform fonts for local font faces are
always initialized using the font data loaded into memory from disk.
This means that there is now only a single path for creating platform
fonts.
In addition, the font list is now responsible for getting the
`FontTemplateDescriptor` for local `FontTemplate`s. Before the font had
to be loaded into memory to get the weight, style, and width used for
the descriptor. This is what fonts lists are for though, so for every
platform we have that information before needing to load the font. In
the future, hopefully this will allow discarding fonts before needing to
load them into memory. Web fonts still get the descriptor from the
platform handle, but hopefully that can be done with skrifa in the
future.
Thsese two fixes together allow properly loading indexed font variations
on Linux machines. Before only the first variation could be
instantiated.
Fixes https://github.com/servo/servo/issues/13317.
Fixes https://github.com/servo/servo/issues/24554.
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
----
- [x] `./mach build -d` does not report any errors
- [x] `./mach test-tidy` does not report any errors
- [x] These changes fix#13317 and #24554
- [x] There are tests for these changes
---------
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
This change contains three semi-related clean ups:
1. the `to_webrender()` and `from_webrender()` functions on Pipeline are
turned into more-idiomatic `From` and `Into` implementations.
2. `combine_id_with_fragment_type` now returns a `u64` as that is what is
expected for all callers and not a `usize`.
3. The `query_scroll_id` query is removed entirely. The
`ExternalScrollId` that this queries is easily generated directly
from the node's opaque id. Querying into layout isn't necessary at
all.
* Add multiple concurrent top-level browsing contexts
Co-authored-by: Delan Azabani <dazabani@igalia.com>
* Rename variables and comments
There are some variable and comments still use browser as names.
This commit renames them to webview.
* Update log message from web view to webview
* Revert offscreen_framebuffer_id rename
* Rename all web view to webview
* Cargo fmt
* Fix viewport/event/clear coordinates when multiview is disabled
* Only deprecate things when multiview is enabled
* Update WebViewManger with shown and invisible sets
Replace visible_webviews and native_window_is_visible with shown_webviews
and invisible_webviews. Add 4 more methods to set them accordingly. The
behavior of is_effectively_visible will return true if the wbview is in
shown_webviews set but not in invisible_webviews.
* Update variant behaviors
* Rename WebViewVisibilityChanged to MarkWebViewInvisible
* Fix unit test by marking id 3 visible again
* Update MarkWebViewInvisible and add UnmarkWebViewInvisible
* Update format and doc comments
* Clean up doc comments
* Address style and naming changes
* Rename UpdateWebView to UpdateFrameTreeForWebView
* constellation: send frame tree unconditionally over focus and feature
* Clarify shown and invisible sets in constellation WebViewManager
* Eliminate forward_to_constellation!()
* Actually remove the unused macro
* Don’t gate compositor changes on multiview feature flag
* Update todo in mouse event dispatch
* Pass all visible webview ids in a single ReadyToPresent message
* Fix compile and lint errors
* servoshell: fix gap between minibrowser toolbar and webview
* Fix failure in /_mozilla/mozilla/window_resizeTo.html
* Fix compile warnings
* Remove stray dbg!()
* Remove confusing “effectively visible” logic (see #31815, #31816)
* Allow embedder to show/hide/raise webviews without ipc
* Update root pipeline only when painting order actually changes
* Stop gating old focus and SetFrameTree behaviour behind Cargo feature
* Use webview_id and WebViewId in webview-related code
* Improve logging of webview-related embedder events
* Allow webview Show and Raise events to optionally hide all others
* Don’t do anything in response to WebViewPaintingOrder
* Remove WebViewPaintingOrder, since its payload is unreliable
* On MoveResizeWebView, only update root pipeline if rect changed
* Rename IOCompositor methods for clarity
* compositor: add event tracing; log webview ops even without ipc
* Add temporary debug logging
* Add more temporary debug logging
* Remove temporary logging in compositor
* Remove temporary debug logging
* Add temporary debug logging, but defer I/O until panic
* Capture a backtrace with each crash log entry
* Proper error handling without panicking in WebViewManager
* Clean up imports in constellation
---------
Co-authored-by: Delan Azabani <dazabani@igalia.com>
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
WebRender does not preserve spatial tree offsets when updating the
spatial tree. Updating the spatial tree of a pipeline can also
update the spatial tree of child pipelines. This change ensures that
WebRender always gets the scroll offsets of the entire scene when
modifying display lists in a way that may rebuild the spatial tree.
Fixes#31807.
During the shutdown process, various threads (such as the
font cache thread) may be finishing up their work. If those threads make
synchronous requests to the compositor, answer them -- even if the
results will be unused. This is at least enough processing for them to
finish their work and exit cleanly.
This addresses crashes that are sometimes seen at exit, particuarly when
the font cache thread tries to register a font during shutdown.
In addition, this change also removes an unused compositor message.
This brings the version of WebRender used in Servo up-to-date with Gecko
upstream. The big change here is that HiDPI is no longer handled via
WebRender. Instead this happens via a scale applied to the root layer in
the compositor. In addition to this change, various changes are made to
Servo to adapt to the new WebRender API.
Co-authored-by: Mukilan Thiyagarajan <mukilan@igalia.com>
These tests don't seem to time out any longer and this mechanism is
probably better served by some sort of adjustment to `RUST_LOG` in the
WPT harness.
When exiting after load or dumping a screenshot, limit the speed that
the compositor sends tick events to the constellation (and on to the
script thread). Previously, these tick messages were sent as fast as the
winit event loop (that's really fast!), until the document was ready to
for its screenshot. rAf is meant to run at more or less the same
frequency as the refresh rate of the screen.
On my system, the massive amount of ticks was overwhelming the router
thread, preventing fonts from loading properly. This change is an attempt
to make tests with requestAnimationFrame less flaky.
* compositor: Improve the way we wait for frames
In the newest version of WebRender it will be harder to make the
distinction between frame queued for scrolling and other kinds of
pending frames. This change makes it so that we queue frames for both
kinds of changes the same way and keeps a counting of pending frames.
This is conceptually a lot simpler.
In addition, do queue a composite even when recomposite isn't necessary
for a WebRender frame when there are active requestAnimationFrame
callbacks. Doing a composite is what triggers the callbacks to actually
run in the script thread! I believe this was a bug, but the WebRender
upgrade made it much more obvious.
These changes are in preparation for the WebRender upgrade.
* Remove spurious println
* script: Do not run layout in a thread
Instead of spawning a thread for layout that almost always runs
synchronously with script, simply run layout in the script thread.
This is a resurrection of #28708, taking just the bits that remove the
layout thread. It's a complex change and thus is just a first step
toward cleaning up the interface between script and layout. Messages are
still passed from script to layout via a `process()` method and script
proxies some messages to layout from other threads as well.
Big changes:
1. Layout is created in the script thread on Document load, thus every
live document is guaranteed to have a layout. This isn't completely
hidden in the interface, but we can safely `unwrap()` on a Document's
layout.
2. Layout configuration is abstracted away into a LayoutConfig struct
and the LayoutFactory is a struct passed around by the Constellation.
This is to avoid having to monomorphize the entire script thread
for each layout.
3. Instead of having the Constellation block on the layout thread to
figure out the current epoch and whether there are pending web fonts
loading, updates are sent synchronously to the Constellation when
rendering to a screenshot. This practically only used by the WPT.
A couple tests start to fail, which is probably inevitable since removing
the layout thread has introduced timing changes in "exit after load" and
screenshot behavior.
Co-authored-by: Josh Matthews <josh@joshmatthews.net>
* Update test expectations
* Fix some issues found during review
* Clarify some comments
* Address review comments
---------
Co-authored-by: Josh Matthews <josh@joshmatthews.net>
This is a small cleanup that moves and renames this class. The rename is
simply because we are exposing a lot about the details of Servo's
rendering in the API and it makes sense to start thinking about
abstracting that away a bit.
This also moves the struct to `gfx`, which does have an effect on
Servo's dependency graph. This adds a new dependency on gfx to
`compositing`, but `compositing` had a transitive dependency on
gfx before through `canvas`.
* Android build
* Fixes
* More fixes
- Still failing in the linking step
* More work on getting linking working
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* android: use mozjs with ndk r25c. loads servo.org
more android build fixes.
* fix ./mach run for android and make it follow logs
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* add experimental logic for compositor pause/resume
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* pass DPI from android to simpleservo
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* ci: add android workflow
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* switch to ANDROID_SDK_ROOT and ANDROID_NDK_ROOT vars
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* upgrade gradle to 4.10.1
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* upgrade to gradle 5.1.1
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* upgrade to gradle 8 and agp 8
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* make compositing work again with external present
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* android: improve mach support for non-NixOS and CI
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* fix sampler compilation bug introduced in #30490
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* ci: add android build to main workflow
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* gradle: set MinSdk = targetSdk = 30
NDK requires we compile against the minSdk API level
which is 30 in our case.
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* add instructions for android in README.md
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* apk: move servosurface to servoview
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* apk: uncomment the mediasession callbacks on MainActivity
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* apk: fix crash on MainAtivity.onDestroy
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* apk: drop VR, arm 5 and unused code
This commit drops:
* support for google, oculusvr
* support for arm5 architecture
and also removes
* fakeld scripts
* unused java code
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* cleanup shell.nix
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* android: add FIXMEs for gstreamer code
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* apk: remove commented code and debug logs
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* cleanup ServoView.java
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* mach: comment call to download gstreamer deps for android
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* disable bluetooth for jniapi as blurdroid is broken
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* fixup! README.md
* fixup! remove change in Cargo.toml
* fixup! move shell variables together
* fixup! cleanup jniapi/Cargo.toml comments
* delete commented gstreamer related android code
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* remove unused config variable in servbuild
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* android: more cleanup
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* force no_static_freetype only for android
* use actions to manage sdk, ndk and java
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* rename embedder event names to be more clear.
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* link to startup crash issue
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* fix lint issues
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* upgrade env_logger to 0.10 with duplicate exception
libservo and android_logger can use env_logger 0.10
but quickcheck is still stuck on 0.8 and has not seen
any activity in the last 2 years. This commit adds
a duplicate exception until the quickcheck dependency
can be upgraded (or replaced)
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* android: fix comments
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* disable jemalloc on android
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
* fixup! replace linux with android in cfg
---------
Signed-off-by: Mukilan Thiyagarajan <mukilan@igalia.com>
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
* Offscreen rendering
* shared memory case never actually rendered to backbuffer
* fix compile errors (in theory) when gl crate feature disabled
* update doc comments
* remove dark CentralPanel border covering edges of viewport
* clear to transparent, to avoid pink artifacts
* fix mouse input for browser being consumed by egui
* avoid destroying OpenGL resources unless resizing window
* clean up compositing::gl
* fix flickering around edges after resizing window
* unset invalidate_last_render_target after invalidating
* fix incorrect DRAW_FRAMEBUFFER name when blitting
* bind the widget surface fbo before painting egui
* make composite_specific_target take CompositeTarget, not Option
* compositing: remove cargo feature “gl”
* capitalise FBO in bind log message
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
* capitalise FBO in drop log message
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
* rename RenderTargetInfo fields and use OnceCell for next field
* rename RenderTargetInfo.read to read_back_from_gpu
* document servo_framebuffer_id in Minibrowser::update
* rename needs_fbo to use_offscreen_framebuffer
* capitalise FBO in unbind log message
* clarify the purpose of Minibrowser::on_event
* fix unused_must_use warning
* reduce nesting in Minibrowser::update
* use implicit format argument in panic
* store Minibrowser.widget_surface_fbo as glow type
* explain why servo_framebuffer_id is None in first call site
* rename output_framebuffer_id to offscreen_framebuffer_id
---------
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
* Replace the time crate with std::time in components/compositing
Signed-off-by: Reid Swan <reidswan@outlook.com>
* Remove elapsed_since_epoch function
---------
Signed-off-by: Reid Swan <reidswan@outlook.com>
The constellation forwards messages from other tasks to the compositor.
Mainly, these are passed to WebRender. This change updates the names of
these messages so it is clearer where they are coming from and where
they are going.
When display lists update quickly, a hit test result might be returned
for a previous display list / list of hit test items. When that happens,
ignore the hit test result.
This fixes a crash, but there might be situations where we can do
something better, such as wait for display list processing to finish
before performing the hit test. A future change might do this for events
like mouse clicks and touch events that should never be thrown away.
Ultimately, the best thing is likely moving hit testing back to layout
or script so a valid hit test can always be performed against the
current DOM.
Fixes#29796.
Pinch zoom was broken because pinch zoom events were triggered at -1, -1
in the compositor. This change:
1. Differentiates between magnify and scroll events in the compositor to
remove the question of where to trigger them.
2. Converts winit TouchpadMagnify events into pinch zoom events so that
this works properly on MacOS.
In the most recent version of WebRender, the pinch zoom APIs have been
removed. We can recreate this functionality by creating a root display
list with a transform.
Main frame scrolling after the WebRender upgrade was broken because the
main frame no longer has content size content bounds and quirks mode
leads to a body with zero height.
Fixes#30368.
* servoshell: fix lockups while animating
* move comment to external_present declaration
* disable needs_recomposite optimisation for now due to breakage
* fix compile error that only happens on ci
* fix more compile errors
* Upgrade vendored version of WebRender
* Patch WebRender: upgrade version of gleam
* Restore hit testing implementation
* Fix WebRender warnings
* Adapt Servo to new WebRender
* Update results
* Add a workaround for #30313
This slightly expands text boundaries in order to take into account the
fact that layout isn't measuring glyph boundaries.
* winit: add minibrowser feature that depends on egui{,-winit}
* winit: carve out some space at the top of headed windows
* winit: minimal toolbar and egui/winit integration (but no painting)
* winit: try to paint with egui_glow (doesn’t work yet)
* winit: add comment about toolbar size
* Add framebuffer object, set it as glow's target
* compositing: clear only the viewport, not the whole framebuffer
* plumb the actual size of the egui toolbar to webrender
* fix formatting
* winit: fix crash when fbo is zero
* winit: don’t bother binding the framebuffer object
* winit: remove unsafe and get toolbar_height
* winit: location field should reflect the current top-level url
* [NFC] winit: move Minibrowser out of App::run
* winit: clean up toolbar height code
* winit: make App own the Minibrowser if any
* winit: make the go button work
* winit:make the location field reflect the current top-level url
* winit: allow enabling minibrowser from command line
* winit: tell compositor to repaint WR and flush when we repaint
* winit: fix bug where location field edits would get overridden
* winit: borrow the minibrowser once in App::handle_events
* winit: address todo about viewport origin coordinates
* winit: fix some minor problems with comments and errors
* winit: update location field once per HistoryChanged event
* winit: rename Window::set_toolbar_size to set_toolbar_height
* winit: take toolbar height into account in hit testing
* winit: pass egui only relevant CursorMoved events
* winit: scratch that, coalesce minibrowser updates instead
* ensure both minibrowser and WR are repainted on every frame
* compositing: only skip framebuffer clear in external present mode
* winit: destroy egui glow Painter when shutting down
* winit: clean up and fix license lint
* fix duplicate versions lint by downgrading bytemuck_derive
was egui_glow ^0.22.0 (0.22.0)
→ egui/bytemuck ^0.22.0 (0.22.0)
→ epaint/bytemuck ^0.22.0 (0.22.0)
→ bytemuck ^1.7.2 (1.13.1)
→ bytemuck_derive ^1.4 (1.4.1)
→ syn ^2.0.1 (2.0.28)
now lock has bytemuck_derive 1.4.0
→ syn ^1.0.99 (1.0.103)
* fix duplicate versions lint by disabling egui-winit/links
(we don’t need support for hyperlinks in our use of egui)
* squelch duplicate versions lint by excluding clipboard-win
* winit: fix compile warnings
* winit: make gleam an optional dependency under /minibrowser
* winit: remove cargo feature, since it’s not really optional
* winit: extract Minibrowser and related code to separate module
* winit: remove unnecessary trailing comma
* winit: simplify the ServoUrl serialisation optimisation
---------
Co-authored-by: atbrakhi <atbrakhi@igalia.com>
This API has been removed in the latest version of WebRender and we can
simply get this information from the compositor-side scroll tree. In
addition, this change limits the amount of data sent to the pipeline
that actually changed and gives the function in the compositor a better
name.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
This also removes the meta viewport support (which was implemented on top), but that also had a single test and is disabled everywhere, so I'm not too concerned, it can be implemented again if / when needed.
This moves more members to the CompositorDisplayListInfo struct, which
now holds all miscellaneous, non-WebRender data when sending display
lists. It also documents what each things sent with a display list does.
This will allow the compositor to properly chain scrolling requests up
when a node has reached the extent of the scroll area. This fixes
scrolling on servo.org.
eca0acf459 uncovered a bug in the way that
the scrolling area of `window` was calculated and broke scrolling on the
root element. This change does two things in order to fix that:
1. Does a partial revert of eca0acf459 in
order to get scrolling from script working again on the window
object.
2. Has the compositor always generate a frame for scrolls starting from
script and waits for them. This is speculative fix for flakiness in
root scrolling tests on CI.
Store hit testing information in a data structure that sits alongside
the display list in the compositor. This will allow the compositor to
store more information per-node. The data structure also takes care of
de-duplicating information between successive display list entries. In
the future, the data structure can be even more aggressive in producing
smaller side hit testing lists, if necessary.
