This also reduces the configuration to just the -qr subset of the
configuration that runs for mochitest -fis.
Differential Revision: https://phabricator.services.mozilla.com/D56197
--HG--
extra : moz-landing-system : lando
Previously this wouldn't drain any jobs queued by FinalizationGroup callbacks.
Differential Revision: https://phabricator.services.mozilla.com/D56538
--HG--
extra : moz-landing-system : lando
Since bug 1598548 scripts could start with a JSOP_LOOPHEAD for while-loops
(do-while loops always have a JSOP_NOP for the debugger) but this confuses
the JITs in various places.
Fixing this edge case properly isn't trivial (see comments on bug 1602681 for
example) so this patch inserts a JSOP_NOP if needed to prevent this for now.
Differential Revision: https://phabricator.services.mozilla.com/D56364
--HG--
extra : moz-landing-system : lando
If the user types a full url, confirms it, and then focuses the urlbar, we should
not
When focusing the address bar after typing and confirming a page, we end up reopening
the view because the search string is the same. This causes unexpected noise and
selection problems (the url is not selected anymore).
This patch changes the behavior so we don't reopen on valid pageproxystate, that
seems to make sense considering the feature scope of restoring an abandoned search.
Additionally, make the keyboard shortcut not trying to reopen the view if it's
already open, to avoid messing up with the selection.
Finally, restore the appropriate allowAutofill status from the last context.
This may still cause us to not autofill something that was previously autofilled,
if 2 tabs contain the same exact search, but that's an edge case, so acceptable.
Differential Revision: https://phabricator.services.mozilla.com/D56635
--HG--
extra : moz-landing-system : lando
The test starts timed out after applying the changes in P1-P3. The main reason is that P1-P3 ensure SWM and the QuotaManger clear their storage in any condition in clear data service. However, since the SWM adds a shutdown blocker during the initialization and it's initialized during the profile-change-teardown because of the changes and the test scenario.
To fix that, ideally, SWM should differentiate if it's initialized before or during the profile-change-teardown and that requires a non-small change. Since we haven't got this signature in the real world and similar cases (SWM gets initialized during profile-change-teardown) have been taken care of, this patch only adds a workaround to avoid the issue.
Differential Revision: https://phabricator.services.mozilla.com/D56526
--HG--
extra : moz-landing-system : lando
The idea of this patch is to try to not use oberver mechanism as possible. To
achieve that, it introduces deleteByOriginAttributes() to cleaners. Different
from other methods, it would only be executed if it's implemented from a
cleaner.
It doesn't remove oberver mechanism entirely since some cleaners are still using
that for other deleteByXXX() functions. So, it only applies removing stuff to
PushService, QuotaManagerService, ServiceWorkerManager, nsPermissionManager,
nsApplicationCacheService, and nsCookieService.
Since the original issue is related to QuotaManagerService, it adds xpcshell
test under the dom/quota/test/unit/ to ensure the behavior won't be changed
accidentally in the future.
Differential Revision: https://phabricator.services.mozilla.com/D33758
--HG--
extra : moz-landing-system : lando
We now have a shared code path to compile these ops.
JSOP_INITELEM_ARRAY gains IC support and JSOP_INITELEM_INC can potentially use the
fast path now.
Depends on D56533
Differential Revision: https://phabricator.services.mozilla.com/D56534
--HG--
extra : moz-landing-system : lando
Just move the MConstant-creation code into the callers to simplify later changes.
Depends on D56529
Differential Revision: https://phabricator.services.mozilla.com/D56531
--HG--
extra : moz-landing-system : lando
This removes the confusing isNewArray() check in that function. It's also
incorrect with other changes because it doesn't account for loop phis for
example.
Differential Revision: https://phabricator.services.mozilla.com/D56529
--HG--
extra : moz-landing-system : lando
This was accidentally regressed to match existing line endings in the switch to prettier.
Differential Revision: https://phabricator.services.mozilla.com/D55991
--HG--
extra : moz-landing-system : lando
The test results in a phi with 3 operands: (double, double, float32). The second
operand is not convertible to float32, but we didn't set convertibleToFloat32 to
false in this case. We then incorrectly converted it to float32.
This patch also adds an assertion to help catch similar bugs.
Differential Revision: https://phabricator.services.mozilla.com/D56352
--HG--
extra : moz-landing-system : lando
* storage-mozStorage was used by Fenenc but that is no longer built from m-c.
* Existing Android tests need to be disabled since they rely on being able to manipulate storage but this commit only stubs searching for logins.
* A later commit will actually filter the logins returned by a GV delegate.
Differential Revision: https://phabricator.services.mozilla.com/D54139
--HG--
rename : toolkit/components/passwordmgr/storage-json.js => toolkit/components/passwordmgr/storage-geckoview.js
extra : moz-landing-system : lando
Note that `TextControlState::BindToFrame()` may also run script if it's
called without script blocker. However, it shouldn't occur. Therefore,
we don't need to mark it `MOZ_CAN_RUN_SCRIPT`, but we should check
script blocker existence with `MOZ_ASSERT`.
Differential Revision: https://phabricator.services.mozilla.com/D55776
--HG--
extra : moz-landing-system : lando
If `PreareEditor()` returns `NS_OK`, but if the instance is deleted at returning
from it, `GetTextEditor()` will access `mTextEditor`. Therefore, this patch
makes `PrepareEditor()` return error if the instance is destroyed at its end.
Differential Revision: https://phabricator.services.mozilla.com/D55775
--HG--
extra : moz-landing-system : lando
Similar to the previous patch, it may notify selection listeners of native
anonymous tree in `<input>` and `<textarea>` element.
Differential Revision: https://phabricator.services.mozilla.com/D55774
--HG--
extra : moz-landing-system : lando
They are really complicated and can run script while the instance is being
destroyed. Therefore, they should have their own `AutoTextControlHandlingState`
as "kungFuDeathGrip".
Differential Revision: https://phabricator.services.mozilla.com/D55773
--HG--
extra : moz-landing-system : lando
The destructor of `AutoTextControlHandlingState`is marked as
`MOZ_CAN_RUN_SCRIPT`, but it's not checked at build time. And this does not
need to be created when the method does not run script. Therefore, we should
mark its constructor as `MOZ_CAN_RUN_SCRIPT` instead since it's stack only
class so that its constructor and destructor are always called from same
method.
Differential Revision: https://phabricator.services.mozilla.com/D55772
--HG--
extra : moz-landing-system : lando
NOTIFY_IME_OF_FOCUS performs lazy initialization for text dispatcher. When
GeckoViewSupport attaches GeckoEditable during initialization,
GeckoEditableSupport destroys myself.
If GeckoEditable already has focus, don't destroy GeckoEditable child.
Also, I want to add debug warning to debug like this issue.
Differential Revision: https://phabricator.services.mozilla.com/D55914
--HG--
extra : moz-landing-system : lando
James Graham
Jon Coppeard
Johann Hofmann
Jan de Mooij
Brian Grinstead
Martin Stransky
Dão Gottwald
Dana Keeler
Marco Bonardo
Marian Raiciof
Tom Tung
Tom Tung
Megha
Mark Banner
Jan Varga
Sylvestre Ledru
sotaro
Glenn Watson
Alastor Wu
Jeff Walden
Matthew Noorenberghe
Masayuki Nakano
Makoto Kato
Brindusan Cristian
Paul Bone
Emilio Cobos Álvarez
Daosheng Mu