Hashbrown grows a lot sometimes making us waste a lot of memory. Shrink
some of these maps after CascadeData rebuild / stylesheet collection
invalidation.
Differential Revision: https://phabricator.services.mozilla.com/D134716
The only remaining consumers are ::-moz-tree pseudo-elements (we used to
use ThinBoxedSlice for other data structures in the past).
Those are not particularly performance sensitive so I think just
double-boxing is fine. In the future, if we wanted to avoid the double
indirection, we could probably use the "thin" crate
(https://docs.rs/thin) or similar, which stores the length of the slice
along with the allocation, making the pointer thin in all
configurations, much like "ThinArc" does:
https://searchfox.org/mozilla-central/rev/1ce2eea39442190a71a1f8f650d098f286bf4a01/servo/components/servo_arc/lib.rs#891
In practice though, I don't think it's particularly worth it for this
specific case.
Differential Revision: https://phabricator.services.mozilla.com/D134672
By modeling it as a separate layer that behaves somewhat specially.
See https://github.com/w3c/csswg-drafts/issues/6872.
The remaining revert-layer tests that we fail are because either we
don't implement a feature (like @property) or because it's used in
keyframes (where revert is a bit unspecified and we have existing
issues with it).
Differential Revision: https://phabricator.services.mozilla.com/D133373
This patch looks bigger than it is, but it's mostly because
of plumbing.
To implement revert-layer we need not only the cascade origin of the
declaration, but the whole cascade level, plus also the layer order.
In order to do this, encapsulate these two things inside a 32-bit
`CascadePriority` struct and plumb it through the rule tree and so on.
This allows us to remove the packing and unpacking of CascadeLevel,
though I kept the ShadowCascadeOrder limit for now in case we need to
reintroduce it.
Fix `!important` behavior of layers while at it (implementing it in
`CascadeLevel::cmp`, spec quote included since it was tricky to find)
since some revert-layer tests were depending on it.
The style attribute test is failing now, but follow-up commit fixes
it, see spec issue.
In terms of the actual keyword implementation, it's sort of
straight-forward: We implement revert and revert-layer in a shared
way, by storing the cascade priority that reverted it.
Differential Revision: https://phabricator.services.mozilla.com/D133372
This is a "simplified" implementation of 'ic', similar to what Safari Preview
currently supports: it only considers the advance of U+6C34 if found in the
first available font, and otherwise falls back to the default of 1em.
(The spec allows for this "in cases where it is impossible or impractical to
determine the ideographic advance measure".)
Differential Revision: https://phabricator.services.mozilla.com/D132818
Based on our previous patches, we only support default behavior for
source and scroll-offsets:
1. source:auto
2. scroll-offsets: none
3. scroll-offsets: auto, auto, ...
So update the parser for them. We expect to remove whole
@scroll-timeline in Bug 1733260, so now only do a tiny update in parser.
Differential Revision: https://phabricator.services.mozilla.com/D132417
We hook the rule into cascade data, and so we can look it up by timeline
name. Now we only use StyleScrollDirection from @scroll-timeline rule.
`source` and `scroll-offsets` are skipped now and use the default values
instead because I'm pretty sure the syntax will be changed in Bug 1733260,
and `scroll-offsets` may be obsolete because the spec proposal intents to
make it be always 0% ~ 100%.
Also, add some reftests for the default `source` and `scroll-offsets`,
and different `orientation`s.
Besides, we disable at-scroll-timeline-start-end.html in Gecko because
we don't support start/end descriptors, and there are too many
intermittents in it.
Differential Revision: https://phabricator.services.mozilla.com/D126452
To do this, we always draw the native titlebar behind the toolbox, and
then make the toolbox adapt to it by using the titlebar radius. This
makes us preserve the shadow properly.
On Wayland we'd double-draw the shadow (see bug 1509931 comment 4) so
this fixes it by trimming it as well using border-radius.
Differential Revision: https://phabricator.services.mozilla.com/D128681
This is based off work by smurfd. But this patch doesn't support buttons
both at the left and right, which simplifies a lot the implementation.
Also, clean-up the existing env variables while at it.
Co-authored-by: Nicklas Boman <smurfd@gmail.com>
Differential Revision: https://phabricator.services.mozilla.com/D132073
This was a subtle behavior change in bug 1739464. The early-return here:
https://hg.mozilla.org/mozilla-central/rev/2fb74c67b0c5#l11.40
Meant we also bailed out from user font prioritization for system fonts.
It's unclear whether that's really the best behavior but since the
intention of the regressing patch was not to change behavior, preserving
the old behavior seems better.
Differential Revision: https://phabricator.services.mozilla.com/D131899
We allow animating pseudo-elements like ::-moz-progress-bar (and we
treat them like regular elements).
Ideally we should store animations for these in the parent element as
well, so they survive reframes and such. But treating them as regular
elements right now means that we do animate them, but we never update
animations for them correctly because wrapper.rs assumed them to be
non-animatable.
Since it seems reasonable to keep allowing the animations to happen,
let's just correct the update code and add a test.
Differential Revision: https://phabricator.services.mozilla.com/D131794
.fallback is always the default font for the lang group unless we're a
system font (in which case it's "none").
The only reason we need that is because we need to react to
language changes (which affect the initial font).
Simplify the model a bit doing the language lookup in gfxTextRun (this
should avoid allocating a few nsStyleFont structs too.
Depends on D130732
Differential Revision: https://phabricator.services.mozilla.com/D131312
This seems like more sensible behavior. We have another use of
only_generic(), but that affects font sizing and other browsers agree
with us there:
<div style="font-family: monospace">Should be 13px</div>
<div style="font-family: something, monospace">Should be 16px</div>
So not touching that one.
Differential Revision: https://phabricator.services.mozilla.com/D130732
Apparently using transparent borders and outlines is a common
accessibility technique to make those visible in HCM.
Bug 1740924 comment 9 seemed to indicate we were the only browser
rendering those browsers, but I just confirmed that Edge at least does
show them.
Keep respecting system colors as that's per spec.
Differential Revision: https://phabricator.services.mozilla.com/D131412
This causes (among other things) pages to be dark when using regular
windows system colors and forcing colors to "always", which is nice.
Differential Revision: https://phabricator.services.mozilla.com/D131165
This seems like more sensible behavior. We have another use of
only_generic(), but that affects font sizing and other browsers agree
with us there:
<div style="font-family: monospace">Should be 13px</div>
<div style="font-family: something, monospace">Should be 16px</div>
So not touching that one.
Differential Revision: https://phabricator.services.mozilla.com/D130732
In some platforms (like macOS, windows dark mode, android, and some gtk
themes) the foreground selection color might be `currentcolor`, and that
doesn't generally guarantee enough contrast with the attention
background.
Remove HeadlessLookAndFeelGTK's handling of this color since it's
useless (always overridden by prefs in all.js)
Differential Revision: https://phabricator.services.mozilla.com/D130617
This seems like more sensible behavior. We have another use of
only_generic(), but that affects font sizing and other browsers agree
with us there:
<div style="font-family: monospace">Should be 13px</div>
<div style="font-family: something, monospace">Should be 16px</div>
So not touching that one.
Differential Revision: https://phabricator.services.mozilla.com/D130732
Behind a pref for now. Given these selectors do nothing on non-chrome
documents (they just don't match) it seems worth trying.
A cursory search seems to indicate they're not used for UA detection or
something like that (or at least I haven't found such an usage).
Differential Revision: https://phabricator.services.mozilla.com/D130736
Behind a pref for now. Given these selectors do nothing on non-chrome
documents (they just don't match) it seems worth trying.
A cursory search seems to indicate they're not used for UA detection or
something like that (or at least I haven't found such an usage).
Differential Revision: https://phabricator.services.mozilla.com/D130736
Use the same document state mechanism we have for :moz-locale-dir. Also,
simplify the setup of the later to be the same as :dir(), allowing the
matching code to be less repetitive.
This should fix some flakiness in chrome mochitests, but we have no existing
tests for these pseudo-classes more generally and since they're just
chrome-only I'm not super-excited about adding more.
Differential Revision: https://phabricator.services.mozilla.com/D130735
This I noticed while working on the following patches. Shouldn't have
any behavior change: the behavior does in fact match the element state
flag semantics correctly if we do this. We did split the dir flags into
two element bits a while ago.
:not(:dir()) still behaves correctly of course, and we have tests for that.
Differential Revision: https://phabricator.services.mozilla.com/D130734
This seems like more sensible behavior. We have another use of
only_generic(), but that affects font sizing and other browsers agree
with us there:
<div style="font-family: monospace">Should be 13px</div>
<div style="font-family: something, monospace">Should be 16px</div>
So not touching that one.
Differential Revision: https://phabricator.services.mozilla.com/D130732
This should mitigate the code size impact. Also make get_resolved_value
non-generic to avoid monomorphizing it multiple times.
Differential Revision: https://phabricator.services.mozilla.com/D130354
For now, use IntoIterator to figure the right type to add the bound.
If we need this on types that are iterable but don't provide
IntoIterator, we can add another attribute field or something.
Differential Revision: https://phabricator.services.mozilla.com/D129962
These are only used for frameset painting and the non-e10s <select>
dropdown focus codepath. We have other more appropriate standard
colors for this.
Differential Revision: https://phabricator.services.mozilla.com/D129992
The code introduced in the preceding patch deals with currentColor correctly,
so we should be able to do this now.
This uncovers a bug in the existing serialization code when a non-auto
text-decoration-thickness was used, caught by
css/css-text-decor/parsing/text-decoration-computed.html.
Differential Revision: https://phabricator.services.mozilla.com/D130018
This removes the various assumptions that the animation code does.
Code size might be a concern, we can optimize if it is a problem,
but let's do the obvious thing for now.
Differential Revision: https://phabricator.services.mozilla.com/D130017
And remove the autofill.background pref for 95 (or 96, depending on
when this lands) assuming nothing terrible causes us to turn it off on
94.
Differential Revision: https://phabricator.services.mozilla.com/D129988
This doesn't change behavior but will allow us to deduplicate some
logic given we compute the effective color-scheme in C++.
Differential Revision: https://phabricator.services.mozilla.com/D129744
The bitfield approach worked when the layer order was in pre-order, but
the spec was changed to make it work like post-order and I don't think
there's a way to keep it working like that, so keep the layer order in a
separate data structure that we look up when going from Rule to
ApplicableDeclarationBlock.
This is just a vector index operation so hopefully shouldn't be too bad.
This patch intentionally regresses @keyframe handling to some extent,
since we need a bit more complicated approach and it seemed worth
implementing in a separate patch.
Depends on D129380
Differential Revision: https://phabricator.services.mozilla.com/D129381
To more properly support Linux having a different default at runtime.
Expose the resolved value in appinfo for convenience, and use it in the
front-end as needed.
Differential Revision: https://phabricator.services.mozilla.com/D129004
We always use alpha visual for WebRender, and appearance: none is
unnecessary (root element has no intrinsic appearance).
Differential Revision: https://phabricator.services.mozilla.com/D128682
This bit is taken straight from D73454 (I reviewed it but I guess
another pair of eyes is ok, it's really straight-forward).
Co-authored-by: Nicklas Boman <smurfd@gmail.com>
Differential Revision: https://phabricator.services.mozilla.com/D128679
We make it work on macOS by setting pointer-events: none + opacity: 0 rather
than visibility: hidden, and tweaking the caching setup to be Android-like.
Now that the scrollbars sheet is the same across platforms, move it to where
the rest of the UA sheets are. This way we guarantee that the RDM vs. Android
difference is less (just the ifdef at the top of the sheet).
Depends on D128084
Differential Revision: https://phabricator.services.mozilla.com/D128085
Apparently ActiveCaption/CaptionText/InactiveCaption/InactiveCaptionText
are supposed to be used for titlebars, so implement them properly and
remove the GTK-specific colors.
Differential Revision: https://phabricator.services.mozilla.com/D128028
The serialization of rotate should be updated to match the current spec.
1. If a rotation about the z axis (that is, in 2D) is specified, the property
must serialize as just an <angle>.
2. If any other rotation is specified, the property must serialize with an axis
specified. If the axis is parallel with the x or y axes, it must serialize
as the appropriate keyword.
Also, we need to handle the zero length vector separately because it is
parallel to every other vector but different from x axis, y axis, or z axis.
Differential Revision: https://phabricator.services.mozilla.com/D127747
ScrollSensitivity is not used by Gecko. Also remove some remnants
of the old code to combine scroll frames when display lists swap.
Differential Revision: https://phabricator.services.mozilla.com/D127609
For that:
* Tweak the standin system colors to match the non-native theme.
* Use those system colors for button and field backgrounds.
* Rename the "should use system colors" bit to "is high contrast",
which is what it really is (specially now that we use system colors
also in non-high-contrast).
Border colors and other colors like the <input type=range> and such
might need some extra tweaking perhaps, but this is a decent start and
looks good in https://crisal.io/tmp/form-controls.html afaict (dark mode
toggle needs the color-scheme pref enabled of course).
Differential Revision: https://phabricator.services.mozilla.com/D127533
This patch does three things:
* Unifies mac and gtk's buttonactivetext system colors (unships mac's from
content, but it was never meant to be exposed).
* Simplifies the forms.css rules, since <input type=color>'s color property
value doesn't affect its rendering in any meaningful way.
* Adds a buttonactiveface color, which we'll use to provide dark backgrounds
for buttons in Windows dark mode (and is good practice, since generally
every text system color should have a corresponding background).
So as-is it shouldn't change content-exposed behavior (except we stop exposing
the -moz-mac-buttonactivetext to content), but it's a worthy cleanup.
Depends on D127246
Differential Revision: https://phabricator.services.mozilla.com/D127259
There's only one meaningful usage of it, which is to disable native
appearance of the <input type=range> (the windows native theme is no
longer exposed to content).
<input type=range> is inconsistent with every other native widget, which
only disables native appearance if the author specifies backgrounds or
borders. So make it match literally all other widgets and simplify a bit
the code.
We had no tests for this special behavior, let me know if you think it's
worth adding one (but I don't feel very strongly about it).
Differential Revision: https://phabricator.services.mozilla.com/D127082
There's only one meaningful usage of it, which is to disable native
appearance of the <input type=range> (the windows native theme is no
longer exposed to content).
<input type=range> is inconsistent with every other native widget, which
only disables native appearance if the author specifies backgrounds or
borders. So make it match literally all other widgets and simplify a bit
the code.
We had no tests for this special behavior, let me know if you think it's
worth adding one (but I don't feel very strongly about it).
Differential Revision: https://phabricator.services.mozilla.com/D127082
This patch adds the animation-timeline longhand property. For
shorthand, we will do that in the next patch.
This patch includes the aut-generated code in
devtools/shared/css/generated/properties-db.js, by `./mach devtools-css-db`.
Note:
1. we will use this property in Bug 1676791. For now, only make sure
we parse it and serialize it correctly.
2. The syntax of animation-timeline may be updated, based on the spec
issue: https://github.com/w3c/csswg-drafts/issues/6674.
However, it's not a big problem to update it later, so we still can
prototype this property based on the current version of spec.
Differential Revision: https://phabricator.services.mozilla.com/D126450
@counter-style/@font-face/@scroll-timeline need similar fixes, but tests
for those haven't been synced yet so waiting for that before writing
them.
Differential Revision: https://phabricator.services.mozilla.com/D126616
I'm not aware of any reason we shouldn't do this, as it is interoperable
with other browsers, and it causes compat issues from sites that forget
to use the prefixed version.
Note this removes some #[parse(aliases)]. These only affect devtools
autocomplete behavior, and we avoid autocompleting -moz-prefixed
keywords when possible, so even though it's a slight behavior change,
it's worth it.
Differential Revision: https://phabricator.services.mozilla.com/D126718
Note that the crash reason is sanitized so we're not exposing anything
sensitive.
I think my patch just changed the signature of the stack, as it didn't
change anything related to guards or what not. But without knowing why
is failing or a repro it's hard to know what's going on.
Printing the address at list would give us some indication of what might
be going wrong (perhaps we're using a static lock when we don't expect
one or such?).
Differential Revision: https://phabricator.services.mozilla.com/D125948
We add scroll-timeline rule into the stylesheet rule type, and add a new
perference to protect it: layout.css.scroll-linked-animations.enabled.
We will use this perference for animation-timeline property as well.
Differential Revision: https://phabricator.services.mozilla.com/D125765
This was needed before, since we used to optimize empty style rules away
(and thus we could get confused between the important and non-important
block for two different elements).
But now we always keep them, and thus there's no way for the same
declaration block to appear in two similar paths of the rule tree
without different cascade levels.
I found this because if we want to implement the revert-layer CSS-wide
keyword, we need to put the layer order in the rule tree. However the
same thing applies: The same style source will always have the same
layer for a given cascade-data rebuild, so I believe this should be
fine.
Differential Revision: https://phabricator.services.mozilla.com/D124740
We need to compute the anonymous name on the fly while building the
CascadeData, otherwise we may see the same layer rule in two places due
to stylehseet sharing and make them incorrectly share a name.
Differential Revision: https://phabricator.services.mozilla.com/D125175
I'll add some tests before enabling. Also, WebKit folks (who have
implemented cos() / tan() / sin()) said they will upstream their tests
to WPT, so I'll extend those with the inverse functions before landing
as well.
Differential Revision: https://phabricator.services.mozilla.com/D124990
The specifics of how this is going to work are still getting spec'd /
discussed in https://github.com/w3c/csswg-drafts/issues/6576, but this
allows DevTools to work fine and the feature to be complete enough for
Nightly experimentation (with the other in-flight patches).
Otherwise devtools crashes when trying to inspect pages that use them.
Differential Revision: https://phabricator.services.mozilla.com/D124656
When we had:
@layer A.B;
We were registering "A" and "B", not "A" and "A.B", which was the intention.
Fix is trivial.
Depends on D124620
Differential Revision: https://phabricator.services.mozilla.com/D124621
This makes layer order use a fixed set of bits per nesting level, to "reserve"
bits for children before they are registered.
See the comment in LayerOrder for the implementation limits it imposes, and
potential alternatives if these limits are not enough (but I think they should
be).
Enable the tests, as they mostly pass now (commit incoming to fix the remaining
ones).
Differential Revision: https://phabricator.services.mozilla.com/D124620
This works modulo the existing nested layer order bug. Will be covered
by WPT /css/css-cascade/layer-import.html once the feature is enabled (I
can probably enable it right away for those tests, but I'd rather fix
the obvious bugs first).
Differential Revision: https://phabricator.services.mozilla.com/D124538
This works modulo the existing nested layer order bug. Will be covered
by WPT /css/css-cascade/layer-import.html once the feature is enabled (I
can probably enable it right away for those tests, but I'd rather fix
the obvious bugs first).
Differential Revision: https://phabricator.services.mozilla.com/D124538
Same, I want to land this separately to see if it affects
micro-benchmarks. If so, we might want to pack the layer order
_somewhere_ (though in this case I'm not sure where, tbh).
With this, layer rules should have an effect on the page. There are
a few things missing before being able to enable them:
* Fix nested layer order in some cases (when parent layers are declared
out of order, see the previous commit mentioning this).
* Some kind of OM representation, perhaps.
* Tests of course, which are coming in bug 1728722 and bug 1727276.
But this should be enough to allow playing with them.
Depends on D124337
Differential Revision: https://phabricator.services.mozilla.com/D124338
I want to land this separately because we might want to get smarter with
the size of the Rule struct (maybe restricting layer order to a u8 per
scope and packing it with the source order, since 255 layers seem
plenty), but I'd rather do the obvious thing for now.
Depends on D124336
Differential Revision: https://phabricator.services.mozilla.com/D124337
This code is really hot, and we've had perf regressions in the past for
introducing function calls in the hot path.
After the previous patch, add_rule is recursive and thus it can't be
inlined, causing a function call for each CSS rule.
This reduces the overhead by making the function take a rule list
instead, causing a function call per rule _list_, which should be
unnoticeable in practice.
Depends on D124335
Differential Revision: https://phabricator.services.mozilla.com/D124336
For that, deal with children in add_rule recursively, to keep the
current layer name up-to-date in block layer rules.
This is not the final version of this code, as right now something like
this:
@layer A {
...
}
@layer B {
...
}
@layer A.A {
...
}
Would give A.A more priority over B, which is not correct. There are
tests for this incoming in wpt sync and such, but that can be tweaked
later.
Differential Revision: https://phabricator.services.mozilla.com/D124335
This shouldn't have any behavior change, but is necessary because for
cascade layers we are going to need to handle the child rules / sheets
ourselves, in order to handle nested layers properly.
Differential Revision: https://phabricator.services.mozilla.com/D124334
See the discussion here: https://twitter.com/Rich_Harris/status/1433153204678799365
This should make attribute selectors roughly as fast as class selectors.
I think it's worth trying and see if perf bots complain on
micro-benchmarks and stylebench and such.
I made attributes more specific than local names, but less specific than
classes, which I think makes sense. When doing something like
foo[data-bar], filtering by data-bar seems likely to yield less elements
than filtering by foo.
While at it, remove the bloom filter pref since we shipped it in
bug 1704551 for 87 and we haven't heard complaints.
Differential Revision: https://phabricator.services.mozilla.com/D124383
Not hooked anywhere yet, so this doesn't change behavior, but adds the
basic data model etc.
Adding parsing support requires some changes to cssparser to allow the
same at rule to be block and statement-like at the same time, so better
done separately.
Differential Revision: https://phabricator.services.mozilla.com/D124079
See the discussion here: https://twitter.com/Rich_Harris/status/1433153204678799365
This should make attribute selectors roughly as fast as class selectors.
I think it's worth trying and see if perf bots complain on
micro-benchmarks and stylebench and such.
I made attributes more specific than local names, but less specific than
classes, which I think makes sense. When doing something like
foo[data-bar], filtering by data-bar seems likely to yield less elements
than filtering by foo.
While at it, remove the bloom filter pref since we shipped it in
bug 1704551 for 87 and we haven't heard complaints.
Differential Revision: https://phabricator.services.mozilla.com/D124383
Also, more directly go from StyleImageRendering to wr::ImageRendering.
* image-rendering: smooth the non-deprecated version of
OptimizeQuality, which maps to SamplingFilter::LINEAR /
wr::ImageRendering::Auto (which uses gl::LINEAR).
* image-rendering: pixelated maps to wr::ImageRendering::Pixelated /
SamplingFilter::POINT which is the same crisp-edges does.
Note that this uncovers that we were mapping image-rendering:
crisp-edges to wr::ImageRendering::Pixelated.
I'm going to preserve behavior on this patch but we should consider
switching that to map to wr::ImageRendering::CrispEdges on a
follow-up (filed bug 1728831 for this).
Differential Revision: https://phabricator.services.mozilla.com/D124378
Emilio Cobos Álvarez
Sandor Molnar
Jonathan Kew
Marian-Vasile Laza
Boris Chiou
Emily McDonough
Alexandru Michis
Butkovits Atila
lamoure6
Emilio Cobos Alvarez
Ting-Yu Lin
Norisz Fay
Mats Palmgren
Noemi Erli
stransky
Mike Hommey
Glenn Watson
Harry Twyford
Cristian Tuns
Stephen A Pohl
Csoregi Natalia
Chris Peterson
criss