This is preparation for sharing this code with layout_2020 and
implementing selective off-the-main-thread animations.
We still look for nodes not in the flow tree in the layout thread.
The image cache returns an `ImageCacheResult::ImageAvailable `the second
time you try getting the placeholder. This means that in some cases, the
loading of an image would fail, then the same image would get fetched
from the cache, the placeholder would be loaded from that but would be
seen as a normal image, firing a load event.
This made the tests in
`fetch/cross-origin-resource-policy/image-loads.html` fail depending on
their order.
This change corrects synchronization issues with animations, by
reworking the animation processing model to do a quick restyle and
incremental layout when ticking animations.
While this change adds overhead to animation ticks, the idea is that
this will be the fallback when synchronous behavior is required to
fulfill specification requirements. In the optimistic case, many
animations could be updated and applied off-the-main-thread and then
resynchronized when style information is queried by script.
Fixes#13865.
This is triggered when an animation finishes. This is a high priority
because it allows us to start rooting nodes with animations in the
script thread.
This doesn't yet cause a lot of tests to pass because they rely on the
existence of `Document.getAnimations()` and the presence of
`animationstart` and animationiteration` events.
Everytime a new LayoutContext was created, it created a new origin which
caused endless stream of image loads to occur in case of reflow. The reason
for this was that the existing image, although cached successfully, was not
used because the entry in hashmap did not match because of different(new)
origin.
This is solved by storing the origin of a window in enum ScriptReflow and
used in creating new LayoutContext in case of reflow.
Implement corner clipping.
Remove PixelFormat from WebrenderImageInfo.
Use WebRender text shadow.
Remove MallocSizeOf and Deserialize for DL items.
Closes#19649, #19680, #19802
Unfortunately, this required quite a bit of changes to the non-test
code. That's because FontContext depends on a FontCacheThread, which in
turn depends on a CoreResourceThread and therefore lots of other data
structures.
It seemed like it would be very difficult to instantiate a FontContext
as it was, and even if we could it seems like overkill to have all these
data structures present for a relatively focused test.
Therefore, I created a FontSource trait which represents the interface
which FontContext uses to talk to FontCacheThread. FontCacheThread then
implements FontSource. Then, in the test, we can create a dummy
implementation of FontSource rather than using FontCacheThread.
This actually has the advantage that we can make our dummy
implementation behave in certain specific way which are useful for
testing, for example it can count the number of times
find_font_template() is called, which helps us verify that
caching/lazy-loading is working as intended.
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`.
Just use WebRender's ClipId directly. This will allow us to create and
use ReferenceFrames in the future, if we need to do that. It will also
make it easier to have Servo responsible for creating the root
scrolling area, which will allow removing some old hacks in the future.
In support of this goal, the layout thread collects information about
CSS images that are missing image data and hands it off to the script
thread after layout completes. The script thread stores a list of
nodes that will need to be reflowed after the associated network
request is complete. The script thread ensures that the nodes are
not GCed while a request is ongoing, which the layout thread is
incapable of guaranteeing.
The image cache's API has also been redesigned in support of this
work. No network requests are made by the new image cache, since it
does not possess the document-specific information necessary to
initiate them. Instead, there is now a single, synchronous
query operation that optionally reserves a slot when a cache
entry for a URL cannot be found. This reserved slot is then
the responsibility of the queryer to populate with the contents
of the network response for the URL once it is complete. Any
subsequent queries for the same URL will be informed that the
response is pending until that occurs.
The changes to layout also remove the synchronous image loading
code path, which means that reftests now test the same code
that non-test binaries execute. The decision to take a screenshot
now considers whether there are any outstanding image
requests for layout in order to avoid intermittent failures in
reftests that use CSS images.
It turns out that it's problematic to embed ThreadLocalStyleContext within
LayoutContext, because parameterizing the former on TElement (which we do
in the next patch) infects all the traversal stuff with the trait parameters,
which we don't really want.
In general, it probably makes sense to use separate scoped TLS types for
the separate DOM and Flow tree passes, so we can add a different ScopedTLS
type for the Flow pass if we ever need it.
We also reorder the |scope| and |shared| parameters in parallel.rs, because
it aligns more with the order in style/parallel.rs. I did this when I was
adding a TLS parameter to all these functions, which I realized we don't need
for now.
This allows us to get rid of a bunch of lifetimes and simplify a lot of code. It
also lets us get rid of that nasty lifetime transmute, which is awesome.
The situation with thread-local contexts is still suboptimal, but we fix that in
subsequent patches.
The primary idea of this patch is to ditch the rigid enum of Previous/Current
styles, and replace it with a series of indicators for the various types of
work that needs to be performed (expanding snapshots, rematching, recascading,
and damage processing). This loses us a little bit of sanity checking (since
the up-to-date-ness of our style is no longer baked into the type system), but
gives us a lot more flexibility that we'll need going forward (especially when
we separate matching from cascading). We also eliminate get_styling_mode in
favor of a method on the traversal.
This patch does a few other things as ridealongs:
* Temporarily eliminates the handling for transfering ownership of styles to the
frame. We'll need this again at some point, but for now it's causing too much
complexity for a half-implemented feature.
* Ditches TRestyleDamage, which is no longer necessary post-crate-merge, and is
a constant source of compilation failures from either needing to be imported
or being unnecessarily imported (which varies between gecko and servo).
* Expands Snapshots for the traversal root, which was missing before.
* Fixes up the skip_root stuff to avoid visiting the skipped root.
* Unifies parallel traversal and avoids spawning for a single work item.
* Adds an explicit pre_traverse step do any pre-processing and determine whether
we need to traverse at all.
MozReview-Commit-ID: IKhLAkAigXE
This patch introduces infrastructure for the rule tree, and constructs it.
We don't use it yet, nor have good heuristics for GC'ing it, but this should not
happen anymore once we store the rule node reference in the node.
I haven't messed up with memory orders because I want to do a try run with it,
then mess with them.
Take down the ApplicableDeclarationsCache, use the rule tree for doing the cascade.
layout/context: Wrap `image_cache_thread` in a `Mutex<>` as well
`SharedLayoutContext.image_cache_thread` didn't get the `Mutex<>`
treatment along with all the other channels in there, because
`ipc-channel::Sender` is presently inherently `Sync`. I believe this to
be an implementation accident though, that should be rectified in the
future -- not something users should actually rely on...
Note that sharing senders (be it `mpsc` or `ipc-channel`) in is probably
not a good idea anyway: just cloning them -- and letting them handle the
sharing internally -- should be both simpler and cheaper. But right now
that's how things are handled here; so let's go with the flow...
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/14048)
<!-- Reviewable:end -->
`SharedLayoutContext.image_cache_thread` didn't get the `Mutex<>`
treatment along with all the other channels in there, because
`ipc-channel::Sender` is presently inherently `Sync`. I believe this to
be an implementation accident though, that should be rectified in the
future -- not something users should actually rely on...
Note that sharing senders (be it `mpsc` or `ipc-channel`) in is probably
not a good idea anyway: just cloning them -- and letting them handle the
sharing internally -- should be both simpler and cheaper. But right now
that's how things are handled here; so let's go with the flow...
Layers were a feature of the legacy drawing path. If we re-add them at
some point, it probably makes more sense to make them a product of
display list inspection.
This change also remove a bunch of dead painting code.
This removes paint threads, rust-layers dependency, and changes
optional webrender types to be required.
The use_webrender option has been removed, however I've left
the "-w" command line option in place so that wpt
runner can continue to pass that. Once it's removed from there
we can also remove the -w option.
Once this stage is complete, it should be fine to change the
display list building code to generate webrender display
lists directly and avoid the conversion step.
As a follow-up, we could move all the data living under a mutex in the
SharedLayoutContext only in order to create the local context to the same place.
This should increase animation performance when there are multiple animations in
one page that happen to be on different threads.
