PlatformCompositorWidgetDelegate was meant to be a pure virtual base
class for all the functions that nsWindow could call that would
either go to an in-process compositor widget or an OMTC widget.
By that definition, it does not seem like CompositorWidgetParent should
be a subclass, since nsWindow cannot directly call its methods and
currently CompositorWidgetParent has several "do nothing"
implementations of the interface methods because they don't really
belong.
This changeset remedies this by refactoring CompositorWidgetParent so
it is no longer an implementor of PlatformCompositorWidgetDelegate. Now
the only implementors are CompositorWidgetChild and InProcessWin-
CompositorWidget, which makes sense because they are both directly
called through the nsWindow delegate.
It also eliminates some of the methods that seem like they don't belong
in PlatformCompositorWidgetDelegate.
Differential Revision: https://phabricator.services.mozilla.com/D57429
--HG--
extra : moz-landing-system : lando
Previously we simply ignored skipped tests when performing test chunking in the
taskgraph. This could sometimes result in a task that runs a single manifest
that happens to be skipped on that task's configuration. This results in an error
that no tests were run from the harness.
This patch attempts to guess a mozinfo based on a task's definition. Then it
uses those values to take skipped tests into account when running the chunking
algorithm. That way the chances of scheduling a manifest that would be skipped
anyway are much lower.
Depends on D58987
Differential Revision: https://phabricator.services.mozilla.com/D58988
--HG--
extra : moz-landing-system : lando
The issue here is that the sort order of 2 or more elements with the same key
is indeterminate: One run (say, one test chunk) may order them one way and another
run (another test chunk) may order them another. The chunk_by_manifest() filter
relies on sort order in two places, in both cases using list length as the key
for sorting; as equal list lengths are not uncommon, multiple runs can assign
very different lists of tests to the chunks, even though all the inputs are the
same. Extending keys with unique data provides a simple fix.
Differential Revision: https://phabricator.services.mozilla.com/D59227
--HG--
extra : moz-landing-system : lando
BytecodeEmitter.cpp:
- Drive-by change in `emitObjLiteralValue()`: Call MOZ_CRASH instead of returning
`false` for unhandled parse nodes. And also assert the function is only called
with valid parse nodes in the first place.
NameFunctions.cpp:
- BigInts aren't currently supported for guessed names, so just I've only updated
the assertion condition.
Parser.cpp:
- Update spec reference from `NumericLiteralBase` to `NonDecimalIntegerLiteral`.
- The new `bigIntAtom()` function runs counter to the deferred gc-things work,
but this inevitable as long as function names are computed in the parser.
Differential Revision: https://phabricator.services.mozilla.com/D58757
--HG--
extra : moz-landing-system : lando
The tests are now using mutation observer to check
when the dom is updated instead.
Differential Revision: https://phabricator.services.mozilla.com/D57499
--HG--
extra : moz-landing-system : lando
This test is proving to be a nuisance during the migration.
Temporarily disable it while the reftest-no-accel suite is greened up on ubuntu1804 and ready for migration on mozilla-central.
Differential Revision: https://phabricator.services.mozilla.com/D59384
--HG--
extra : moz-landing-system : lando
BytecodeEmitter.cpp:
- Drive-by change in `emitObjLiteralValue()`: Call MOZ_CRASH instead of returning
`false` for unhandled parse nodes. And also assert the function is only called
with valid parse nodes in the first place.
NameFunctions.cpp:
- BigInts aren't currently supported for guessed names, so just I've only updated
the assertion condition.
Parser.cpp:
- Update spec reference from `NumericLiteralBase` to `NonDecimalIntegerLiteral`.
- The new `bigIntAtom()` function runs counter to the deferred gc-things work,
but this inevitable as long as function names are computed in the parser.
Differential Revision: https://phabricator.services.mozilla.com/D58757
--HG--
extra : moz-landing-system : lando
Also adds tests for checking the search engine lists when falling back to a default region.
Differential Revision: https://phabricator.services.mozilla.com/D59084
--HG--
extra : moz-landing-system : lando
The new code is a superset of the old, so I've left it much the same to avoid conflicts until we come up with a more long term solution.
Differential Revision: https://phabricator.services.mozilla.com/D59181
--HG--
extra : moz-landing-system : lando
Width does not have an effect on a table-row element, but only when it is
displayed in a horizontal writing-mode.
Similarly, height does not have effect when it is displayed in a vertical
writing-mode.
So this patch makes our inactiveCSS validator a little more specific in those
cases.
Differential Revision: https://phabricator.services.mozilla.com/D57768
--HG--
extra : moz-landing-system : lando
This patch only allows sacrificing subpixel anti-aliasing when the
screen size is larger than WUXGA, and when the force disable pref is not
set. In the future, we may also add disable this for high end GPUs.
This also consolidates the WebRender debug flags to use the same
signaling infrastructure to avoid needing to store the debug flag state
and check on each transaction. Instead it now applies the debug flag
updates when the gfxVar changes.
Differential Revision: https://phabricator.services.mozilla.com/D57469
--HG--
extra : moz-landing-system : lando
clearKeptObjects will be called in asyncGCDeref.
This is needed for the following test:
test262/built-ins/WeakRef/prototype/deref/gc-cleanup-not-prevented-with-wr-deref.js
Differential Revision: https://phabricator.services.mozilla.com/D59305
--HG--
extra : moz-landing-system : lando
To fix the failures in:
js/src/tests/test262/built-ins/WeakRef/length.js
test262/built-ins/WeakRef/prototype/deref/length.js
Differential Revision: https://phabricator.services.mozilla.com/D59304
--HG--
extra : moz-landing-system : lando
Matt Brandt
Chris Martin
Brad Werth
Andrew Halberstadt
Geoff Brown
Brian Hackett
André Bargull
Nicolas Chevobbe
Junior Hsu
Brindusan Cristian
Edwin Takahashi
John Lin
Julian Descottes
Erica Wright
Andi-Bogdan Postelnicu
Dão Gottwald
Nika Layzell
Matthew Gaudet
Mihai Alexandru Michis
Mark Banner
Botond Ballo
Michal Novotny
Simon Fraser
Matt Singleton
Noemi Erli
Patrick Brosset
Elliot
ffxbld
shravanrn@gmail.com
Yoshi Cheng-Hao Huang
Andrew Osmond