Altivec code was previously uncompilable, and the altivec accelerations
also could not be disabled with a build flag or env variable, leading
to pale moon not being buildable on any altivec supported CPU ( all
PPC CPUs ever almost ). Why? I don't know, 'inline' in C is cursed
and making it static makes it significantly less so.
Closes: https://forum.palemoon.org/viewtopic.php?f=5&t=31108&p=252462#p252462
Signed-off-by: Richard Rogalski <rrogalski@firemail.cc>
We null-initialize all of the other pointer members in NewPerSpanData; we should
do the same for these ones, for consistency & robustness.
(In practice, the callers end up initializing these members before reading them
anyway, so it's been benign that we weren't initializing them. But better for
safety & futureproofing to have them reliably initialized.)
We already suppressed GC for part of this, but we may sweep before the pre-write
barrier is established.
The simplest and safest thing is to suppress GC for the whole method.
Remove early return in nsExternalProtocolHandler::NewChannel when handler doesn't exist.
Fix devtools expectation that newChannel throws for unsupported external handlers
According to the original SVG spec, the SVG element in documents had to follow
special rules as to what was allowed for width/height attributes. This is no
longer the case, so we can decouple it from whatever internal units SVG/XML
language uses and allow all units like every other element.
Loosely based on https://phabricator.services.mozilla.com/D25101
Includes changes to `GetEmojiPresentation`, added in e38b57ca.
The `EmojiComponent` response includes hair color, skin tone, ZWJ, VS16, Combining Enclosing Keycap, and the Tag unicode characters used for subdivision flag emojis.
This means that `IsClusterExtender` should support all these characters, and be forward compatible with any future Unicode changes.
This also lets us restructure `ClusterIterator::Next` to check for EmojiComponents after a TextDefault character, without any stupid constants.
This implementation is different than Mozilla, not indirectly queueing cancel
tasks to be arbitrarily ordered, but rather synchronously close in reverse
top layer order.
This adds a pseudo-class `:-moz-modal-dialog` to be able to track whether the
dialog is in the top layer in CSS.
This implements default styling for modal dialogs following the HTML standard.
Slightly off-extreme color scheme chosen for unstyled <dialog> elements.
Fullscreen stack isn't part of the spec anymore; it's been changed to a more
generic version called Top Layer stack, which is being used by both full-screen
APIs and dialog elements.
This patch refactors it to Top Layer stack so that it can be reused for dialog
elements.
Top Layer stack spec: https://fullscreen.spec.whatwg.org/#new-stacking-layer
Prerequisite for top layer refactor.
FunctionRef is a generic callable type that can be initialized from any
compatible callable, suitable for use as a function argument for the duration of
the function call (and no longer).
- Don't check STAT designAxisSize if the designAxisCount is zero.
- Avoid potential arithmetic overflow during Buffer read operations.
- Use more careful range checks in STAT parsing.
