The style candidate cache had regressed a few times (see #12534), and my
intuition is that being able to disable all style sharing with a single rule in
the page is really unfortunate.
This commit redesigns the style sharing cache in order to be a optimistic cache,
but then reject candidates if they match different sibling-affecting selectors
in the page, for example.
So far the numbers have improved, but not so much as I'd wanted (~10%/20% of
non-incremental restyling time in general). The current implementation is really
dumb though (we recompute and re-match a lot of stuff), so we should be able to
optimise it quite a bit.
I have different ideas for improving it (that may or may not work), apart of the
low-hanging fruit like don't re-matching candidates all the time but I have to
measure the real impact.
Also, I need to verify it against try.
In the Gecko case, this style source would be the style context. In the servo
case, it will be always the computed values.
We could optimise this further in the case of stylo (from three FFI calls to
one) if we use an API of the form CalcAndStore(node, new_cv). But that would
imply borrowing the data twice from Servo (we also have borrow_data_unchecked
fwiw, but...).
If we're restyling a page with animations and layout takes too long, we
might still have the expired animations from the last restyle, without these
being cleared out by layout on `tick_animations`.
Unfortunately it's hard (near to impossible?) to make a reduced test case for
this, since it heavily depends on the speed of the build and conditions that
only happen under heavy loads.
Mainly, it depends on how accurately the `TickAllAnimations` message is sent
from the constellation.
Thus, we can't just re-expire an animation, and we should re-check for it as a
previous animation.
Fixes#12171
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.
There's a bit of flickering when unpausing where the node has the original
state, but I'm not totally sure where it comes from, posibly from
PropertyAnimation returning None due to no styles changing?
So this actually allows some more animations to be triggered. The bad part of
this is that they're actually triggered always when the style is recalculated,
so we're going to have at least some more state into the node, and the
constellation, which would have to keep track of the animations states.
These can't be supported in Servo as of right now, because I'm not
totally sure the accesses that should be done in layout would be
thread-safe.
It can be revisited later though.
This commit refactors the style crate to be completely independent of
the actual implementation and pseudo-elements supported.
This also adds a gecko backend which introduces parsing for the
anonymous box pseudo-elements[1], although there's still no way of
querying them.
https://mxr.mozilla.org/mozilla-central/source/layout/style/nsCSSAnonBoxList.h
