It's easy to construct examples where not inserting in those cases causes performance
to get worse (for example, any long list of siblings that match the same selectors
while having some non-effectual differences in LocalName/Class/Id/etc). And the LRU
nature of the cache already does the right thing of pruning non-useful entries.
Fixing this causes several hundred more sharing hits on wikipedia.
MozReview-Commit-ID: L7W8vcMnHaq
One less hack, a few more to go.
Bug: 1374235
Reviewed-by: bholley
MozReview-Commit-ID: 6katL1EGn2U
Signed-off-by: Emilio Cobos Álvarez <emilio@crisal.io>
The buggy animation handling isn't a regression, since currently we pass
UnstyledChildrenOnly in those cases, which blocks the animation traversal
in Servo_TraverseSubtree.
In general I really wanted to handle these two paths together. But there's
enough broken with the NewChildren path that I wanted to scope the buginess
as tightly as possible. And I really need to separate the handling here from
StyleDocument() to make the restyle root stuff work.
MozReview-Commit-ID: 9F0mcQl7AAX
Doing anything else is non-sensical, since we're not guaranteed to reach all of
the bits from traversal Y when doing traversal X.
MozReview-Commit-ID: FQliRxBan70
After bug 1356141, the setup of animation-only dirty bit should have matched
to normal dirty bit's one (Though they don't match in post traversal due to
throttled animation flush). An unset_animation_only_dirty_descendants call
removed in this patch cleared dirty bits which are needed for post traversal if
there is a second animation-only traversal and if there is no need to restyle
for the second animation-only traversal.
The reftest in this patch fails without either this fix or the fix for bug
1367975.
See [Gecko bug 1384435 comment 12](https://bugzilla.mozilla.org/show_bug.cgi?id=1384435#c12)
for more detail what's going on at that time.
We already have a more-specific check further down in the file, which was added in the same revision.
I think this one was erroneous.
MozReview-Commit-ID: CnP0zCpBtnp
In animation-only restyle, we just need to check the element
has animation restyle hints or has recascade self which is a result
of animation-only restyle for ancestors.
has_current_styles() in Servo_ResolveStyle() is intentionally left there,
it will be changed in a subsequent patch.
In the next patch, we'll move logic to identify the children for traversal into
preprocess_children (which will be renamed), and the set_dirty_descendants logic
will move along with it. So left as-is, the code here will clobber the flags.
MozReview-Commit-ID: 7ZskKWD4QC3
This allows us to have ensure_data() and clear_data() functions on the TElement
trait, instead of hacking around it adding methods in random traits.
This also allows us to do some further cleanup, which I'd rather do in a
followup.
