Sending it back via the parent process ensures that it will take the same path
that regular touch events do, and so guarantees that the Tap event won't overtake
the touch events and get dispatched to content first.
MozReview-Commit-ID: 8TiHY2PFPvE
We currently make the initial browser in a window remote by default. If early
on in the session, that one remote browser goes away (and the content process
was still booting), there's about 5 seconds before the shutdown kill timer
will take that content process out for not quitting fast enough.
There are some cases during startup where the content process is waiting
on information from the parent, so it cannot respond to the request to
quit in time. The parents shutdown kill timer goes off, and the shutdown
kill occurs.
In this bug, what's happening is that the initial browser flips remoteness
from remote to non-remote when it goes to about:sessionrestore. This starts
the shutdown kill timer. The content process runs out of time, and the
shutdown kill timer fires, killing the content process. The TabParent::ActorDestroy
method (which still exists, even though the browser is no longer remote),
interprets this as an abnormal shutdown, and bubbles the oop-browser-crashed
event to the associated <xul:browser>, which causes the page to browser to
about:tabcrashed, when it had already loaded about:sessionrestore.
This patch makes it so that the TabParent::ActorDestroy method first checks
to ensure that the associated remote frameloader is still the one that the
frameloader owner cares about. If not (because, say, the remoteness has
flipped and a new non-remote frameloader has been created), then the
event is not fired, since the user has moved on.
MozReview-Commit-ID: G4jmR6lMMFl
We currently make the initial browser in a window remote by default. If early
on in the session, that one remote browser goes away (and the content process
was still booting), there's about 5 seconds before the shutdown kill timer
will take that content process out for not quitting fast enough.
There are some cases during startup where the content process is waiting
on information from the parent, so it cannot respond to the request to
quit in time. The parents shutdown kill timer goes off, and the shutdown
kill occurs.
In this bug, what's happening is that the initial browser flips remoteness
from remote to non-remote when it goes to about:sessionrestore. This starts
the shutdown kill timer. The content process runs out of time, and the
shutdown kill timer fires, killing the content process. The TabParent::ActorDestroy
method (which still exists, even though the browser is no longer remote),
interprets this as an abnormal shutdown, and bubbles the oop-browser-crashed
event to the associated <xul:browser>, which causes the page to browser to
about:tabcrashed, when it had already loaded about:sessionrestore.
This patch makes it so that the TabParent::ActorDestroy method first checks
to ensure that the associated remote frameloader is still the one that the
frameloader owner cares about. If not (because, say, the remoteness has
flipped and a new non-remote frameloader has been created), then the
event is not fired, since the user has moved on.
MozReview-Commit-ID: G4jmR6lMMFl
This change avoids lots of false positives for Coverity's CHECKED_RETURN
warning, caused by NS_WARN_IF's current use in both statement-style and
expression-style.
In the case where the code within the NS_WARN_IF has side-effects, I made the
following change.
> NS_WARN_IF(NS_FAILED(FunctionWithSideEffects()));
> -->
> Unused << NS_WARN_IF(NS_FAILED(FunctionWithSideEffects()));
In the case where the code within the NS_WARN_IF lacks side-effects, I made the
following change.
> NS_WARN_IF(!condWithoutSideEffects);
> -->
> NS_WARNING_ASSERTION(condWithoutSideEffects, "msg");
This has two improvements.
- The condition is not evaluated in non-debug builds.
- The sense of the condition is inverted to the familiar "this condition should
be true" sense used in assertions.
A common variation on the side-effect-free case is the following.
> nsresult rv = Fn();
> NS_WARN_IF_(NS_FAILED(rv));
> -->
> DebugOnly<nsresult rv> = Fn();
> NS_WARNING_ASSERTION(NS_SUCCEEDED(rv), "Fn failed");
This makes a lot of code more compact, and also avoids some redundant nsresult
checks.
The patch also removes a handful of redundant checks on infallible setters.
This patch makes the following changes on many in-class methods.
- NS_IMETHODIMP F() override; --> NS_IMETHOD F() override;
- NS_IMETHODIMP F() override {...} --> NS_IMETHOD F() override {...}
- NS_IMETHODIMP F() final; --> NS_IMETHOD F() final;
- NS_IMETHODIMP F() final {...} --> NS_IMETHOD F() final {...}
Using NS_IMETHOD is the preferred way of marking in-class virtual methods.
Although these transformations add an explicit |virtual|, they are safe --
there's an implicit |virtual| anyway because |override| and |final| only work
with virtual methods.
This patch makes most Run() declarations in subclasses of nsIRunnable have the
same form: |NS_IMETHOD Run() override|.
As a result of these changes, I had to add |override| to a couple of other
functions to satisfy clang's -Winconsistent-missing-override warning.
There were a couple of problems when delivering tap gestures to content with
full zoom applied. One was that the ConverToGecko function converted the coords
into "CSS pixel" space by using the web content's CSS-to-LD scale, but also
applied that on the translation from the chrome area. Moving that conversion
to later in the process (after the coords got passed through TabParent::
AdjustTapToChildWidget) corrected that issue.
The other problem was that bits of code in APZEventState and APZCCallbackHelper
were using the widget->GetDefaultScale() value as the CSS-to-LD scale, but that
omitted the full zoom value. Getting the CSS-to-LD scale from the presShell and
propagating that through corrected that issue.
MozReview-Commit-ID: KdrkdEZslHo
I think that we can drop nsIMEUpdatePreference::DontNotifyChangesCausedByComposition(), i.e., nsIMEUpdatePreference::NOTIFY_CHANGES_CAUSED_BY_COMPOSITION because it's now used only by TSFTextStore but TSFTextStore ignores if SelectionChangeDataBase::mCausedByComposition or TextChangeDataBase::mCausedOnlyByComposition is true (for supporting async changes in e10s mode). So, only issue is, dropping the flag might cause increasing computing TextChangeData cost during composition in TSF mode. However, now, it's already enough fast and even if it'd cause performance regression, we could add a hack with TextComposition's offset information. Therefore, we don't need to worry about the performance regression so seriously.
MozReview-Commit-ID: HNT3G4isONj
