To prevent addons or other code to rapidly create and destroy content processes
under various conditions, we let the preallocated process manager to reuse short
lived content processes.
ContentCacheInParent::mPendingCompositionCount is now managed with composition events which TabParent received. However, TextComposition doesn't dispatch composition events after coming request to commit active composition. Therefore, composition is committed forcibly in a remote process over 255 times, the main process crashes.
It's the safest way to use TextComposition to manage ContentCacheInParent::mPendingCompositionCount.
MozReview-Commit-ID: DEhzYcK1zcW
--HG--
extra : rebase_source : a47891b1d620bbe4e380e73134ec6da5d21f4ea9
This environment variable works on both Windows and Linux for force-disabling
the content sandbox, and now does so on macOS as well.
The xpcshell tests force disable the sandbox because they do things like bind()
sockets, which is not compatible with the content sandbox. This is needed now
because bug 1358223 was force upgrading the sandbox from level 0 (disabled) to
level 1 on beta channel, which caused breakage.
MozReview-Commit-ID: 5DGxtoDLp0C
--HG--
extra : rebase_source : 87edd138e8b711eace5cb8103c67feae0361e148
MozReview-Commit-ID: A1lCqvbQYAF
There is no clean API-based solution to this, so instead I went grovelling
through the DCOM wire protocol and was able to write a function that converts
handler OBJREFs into standard OBJREFs.
See also:
https://msdn.microsoft.com/en-us/library/cc226801
--HG--
extra : rebase_source : a650055c4adda3a1d99262e47f2b463074c6b935
This permits NSS to load libsoftokn3.dylib, even when the rest of the file
system's access is removed (as is the goal for the content sandbox level 3).
This is needed for WebCrypto.
MozReview-Commit-ID: Bh54b87zIjY
--HG--
extra : rebase_source : e1fa59648d683e97a3bc1310ac1c362009f657f8
This adds the RemoteType annotation to a content crash report so that we can
distinguish between content processes that crashed while running remote, local
or extension code. The annotation is passed along the others to Socorro by the
crashreporter and is also whitelisted for inclusion in the crash ping.
MozReview-Commit-ID: 4avo0IWfMGf
--HG--
extra : rebase_source : 8d03f7e166b5762a5ce7cab13c2101302b4f1d2f
Every JS plugin is assigned a unique ID. When an instance of a JS plugin is created the
frame loader that we use to load the plugin's handler URI will create a special
TabContext. This TabContext causes the ContentParent to use the process for this specific
JS plugin (creating one if it hasn't already) when it creates the PBrowser actors.
This causes the iframes for all the instances of a specific JS plugin to be grouped in the
same process.
--HG--
extra : rebase_source : c39560bdf66cda1a005c7b823b3a46e4734878a4
extra : source : 9cba1db527c7eed4371c9f4caf96fd942608cab6
This patch removes the C++ code used to run the minidump analyzer when a
content process crashes, and replaces it with JS code within the CrashService
object. This removes the need for a separate shutdown blocker in C++ code and
allows end-to-end testing of the crash service functionality. Additionally
the exception handler code can be simplified since it's now only used to run
the crash reporter client.
The test added to test_crash_service.js covers computing the minidump SHA256
hash (bug 1322611) and of the minidump analyzer itself (bug 1280477).
MozReview-Commit-ID: LO5w839NHev
This event is no longer necessary, since checking nsITabParent.hasPresented is enough to know
if we need to blank out the tab or not.
MozReview-Commit-ID: 445XMqhorxC
--HG--
extra : rebase_source : 1a05de827502c409d979a621471978b08ce39fb2
If the "security.sandbox.content.level" preference is set to a value less than
1, all consumers will automatically treat it as if it were level 1. On Linux and
Nightly builds, setting the sandbox level to 0 is still allowed, for now.
MozReview-Commit-ID: 9QNTCkdbTfm
--HG--
extra : rebase_source : cd5a853c46a5cd334504b339bef8df30a3cabe51
If the "security.sandbox.content.level" preference is set to a value less than
1, all consumers will automatically treat it as if it were level 1. On Linux and
Nightly builds, setting the sandbox level to 0 is still allowed, for now.
MozReview-Commit-ID: 9QNTCkdbTfm
--HG--
extra : rebase_source : 1a26ffc5b9f80e6df4c37c23f506e907ba44053a
Originally the pref is called security.data_uri.inherit_security_context, so the pref is default on.
And when we finish this feature, we turn it off.
To avoid causing confusion in the future, perhaps we should rename it and make the pref default off.
Switching to Telemetry::ProcessID allows us to break out extension process data from the content process data.
--HG--
extra : rebase_source : ab334ad33b879dce373f810b452b73747ba82109
extra : source : 9cf5185e6d6fd15b1dd35b1552317ebd5ff258ed
Changes:
- remove code addressed by reviewer
- remove PContent.ipdl, PBrowser.ipdl, and ProcessPriorityManager code
that relates only to removed AudioChannelService methods
- correct test case listening to event from removed code
- remove useless test case files
MozReview-Commit-ID: I96nR8zTXJt
--HG--
extra : rebase_source : 127876c672744811c025ca55839ff2e8a06b1fce
Removing support for this preference means that mReuseLoaderGlobal
will always be false, which lets us eliminate that field, and remove a
lot of code.
This also means that false is always passed to
PrepareObjectForLocation for aReuseLoaderGlobal.
ReadCachedFunction is no longer used, so it is deleted.
MozReview-Commit-ID: 5JD24EYVcQf
--HG--
extra : rebase_source : 3b23b4b8d2b1a2f6a53223477afb4cb13b8a671c
PBrowser::RequestNativeKeyBindings() is used only when somebody tries to
execute native key bindings for synthesized keyboard events. Therefore, it
doesn't need to retrieve edit commands for all editor types. Instead, it
should take the editor type and just return the edit commands for it.
MozReview-Commit-ID: HF4Gz99SBQP
--HG--
extra : rebase_source : e332d2ef5ae36949a9a362d8f95c5d427c9812e7
Now, nsIWidget::ExecuteNativeKeyBinding() isn't used by anybody for executing
edit commands. Instead, they need array of edit commands for the key
combination. So, the method should be renamed to GetEditCommands() and just
return edit commands as an array.
MozReview-Commit-ID: 4G0B1lJ8Lbe
--HG--
extra : rebase_source : 4a5829281d59677c773950c34d91a1cd8074d243
When eKeyPress event is dispatched from TabParent to a remote process, it
should store edit command for all editor types. Then, copied
WidgetKeyboardEvent in the remote process doesn't need to request the edit
commands when its ExecuteEditCommands() is called.
Note that this patch also changes a automated test, browser_bug1316330.js,
that uses nsIDOMWindowUtils.dispatchDOMEventViaPresShell() to dispatch
repeated keyboard events in the tab. However, it should use synthesizeKey()
to emulate everything of native keyboard events and the API can dispatch
repeated keyboard events too. (And the test has a bug. It tries to wait 0.5
sec when every keydown or keypress event. However, it fails since startTime
is never initialized. This patch fixes this bug too.)
MozReview-Commit-ID: IYhyxqH3Ch8
--HG--
extra : rebase_source : 0db911e8e6ea7fc537bda76f4cc0f8952cc13dd2
This function is arguably nicer than calling NS_ProcessNextEvent
manually, is slightly more efficient, and will enable better auditing
for NS_ProcessNextEvent when we do Quantum DOM scheduling changes.
Currently, we only correctly support remote layer trees for frameloaders that
use the same layer manager as their document. Since we need to be able to host
remote <browser> content in popup widgets for remote WebExtensions, we need to
tie the frameloaders to the layer manager of their host element, rather than
the root layer manager for the document.
MozReview-Commit-ID: 4RCsamFBiQw
PLayerTransaction's constructor was previously synchronous so we could
return a TextureFactoryIdentifier. This is quite reliably available
already in the case of opening a tab, due to RenderFrameParent knowing
which compositor it is attached to, so we can make the constructor
asynchronous.
In the top-level widget case, we add a new synchronous message to find
the TextureFactoryIdentifier.
Currently the profiler mostly uses an array of strings to represent which
features are available and in use. This patch changes the profiler core to use
a uint32_t bitfield, which is a much simpler and faster representation.
(nsProfiler and the profiler add-on still use the array of strings, alas.) The
new ProfilerFeature type defines the values in the bitfield.
One side-effect of this change is that profiler_feature_active() now can be
used to query all features. Previously it was just a subset.
Another side-effect is that profiler_get_available_features() no longer incorrectly
indicates support for Java and stack-walking when they aren't supported. (The
handling of task tracer support is unchanged, because the old code handled it
correctly.)
This permission was needed for the memory bloat logging, which is used for
leaktest, including logging intentionally crashing processes. Now we restrict
ourselves to only allowing writes to the location needed for this logging,
rather than all of /private/var.
MozReview-Commit-ID: 5AbJEZlDHNV
--HG--
extra : rebase_source : 26936b8d8bca53f2c37a195b5e7c69c151ec18d2
Since NotifyCurrentTopLevelOuterContentWindowId message has high priority and could jump ahead of PNecko constructor message, we have to also give high priority to PNecko.
Currently, we only correctly support remote layer trees for frameloaders that
use the same layer manager as their document. Since we need to be able to host
remote <browser> content in popup widgets for remote WebExtensions, we need to
tie the frameloaders to the layer manager of their host element, rather than
the root layer manager for the document.
MozReview-Commit-ID: 4RCsamFBiQw
--HG--
extra : rebase_source : 86bca4ae1c012ff1bb84a9ad796be311cfe580f6
extra : histedit_source : 19577d69430adc8cb38c195f13db2c6de6605c4c
To mitigate the delay that a new content process startup might cause, when the
browser reaches a non-busy state we attempt to prelaunch an empty content process
in the background that can be grabbed and used the next time we need one. This patch
enables the preallocated process manager by default and attempts to fix all the
issues that prevented us doing this sooner.
We sometimes briefly have more than one root DocAccessible associated with a TabChild, for example, while navigating links in a page. This patch makes sure that we use the correct accessible when delaying messages that we forward to the parent process.
PLayerTransaction's constructor was previously synchronous so we could
return a TextureFactoryIdentifier. This is quite reliably available
already in the case of opening a tab, due to RenderFrameParent knowing
which compositor it is attached to, so we can make the constructor
asynchronous.
In the top-level widget case, we add a new synchronous message to find
the TextureFactoryIdentifier.
PLayerTransaction's constructor was previously synchronous so we could
return a TextureFactoryIdentifier. This is quite reliably available
already in the case of opening a tab, due to RenderFrameParent knowing
which compositor it is attached to, so we can make the constructor
asynchronous.
In the top-level widget case, we add a new synchronous message to find
the TextureFactoryIdentifier.
--HG--
extra : rebase_source : 4b29b859aa5745fabe3db0fe68742328fc0af175
nsIFilePicker.displaySpecialDirectory is a string that can be set to TmpD,
Desk, or any other special directory value. The real value of this directory
will be read in the parent process.
Opening pages in a new tab might suffer an extra delay from e10s-multi because
the new process has to start up and then run all the process / frame scripts
before it can react on the request from the parent to load the first page.
There are two code paths. Either we start the tab with a remote browser and
then the RemoteWebNavigation will send the request. Or we start with a non-remote
browser and have to change the remoteness flag on it, and then the SessionStore
will send the request.
In each cases we start the timer on the parent side, send it with the message,
and when the child receives it it stops the timer and reports the measured delay.
LocaleService serves two main functions. It is a central place for all code in the
engine to learn about locales, but it also does the language negotiation and selection.
The former is relevant in all processes, but the latter should only be performed
by the "main" process. In case of current Desktop Firefox, the parent process
is the one performing all the language negotiation, and content processes should
operate in the "client" mode.
In Fennec, there's a Java app on top of Gecko which should work as a "server"
and then all processes, including parent process of Gecko is merely a "client" for that.
This refactor finalizes this duality making it easily configurable to define in
which mode a given LocaleService operates.
The server-client model allows all clients to stay in sync with the server,
but operate transparently for all callers just returning the right values.
In order to initialize LocaleService in the client mode in child process with the
right locales I'm adding the list of app locales to the XPCOMInitData,
and then fire LocaleService::SetAppLocales in the child process initialization.
In order to keep the list up to date, I'm adding intl:app-locales-changed to
the list of observed topics, and when triggered, I send the updated list
to the child process, which updates LocaleService::SetAppLocales with the new
list.
MozReview-Commit-ID: K9X6berF3IO
--HG--
extra : rebase_source : ca5e502d064023fddfd63fe6fe5eccefce8dee52
This patch replaces the usage of sNextTabParent pointer to store the next
PBrowser parent actor to be used by the next frame loader with the
following information:
* In the case where the content JS has requested a new tab, the ID of the
next TabParent will be stored on the <xul:browser> element.
* In the case where the content JS has requested a new window, the ID of
the next TabParent will be stored on the created nsXULWindow.
This patch will go away when I'll finishing the removing of PBlob. Currently,
when a PBlob is sent from child to parent, we use PMemoryStream in order to
recreate the inputStream on the parent side. PMemoryStream sends the data in
chunks.
But if PBlob is dealing with a IPCBlobInputStream, it doesn't have access to
the real data. In this case, we must send data using IPCStream. In this way,
Note that thisIPCBlobInputStream will send its ID, and the parent will take the
real inputStream from the IPCBlobInputStreamStorage. Note that I check the
size to be 1mb instead 0. No particular reasons, but better to avoid the use of
PMemoryStream for nothing.
IPCBlobInputStream is a new type of nsIInputStream that is used only in content
process when a Blob is sent from parent to child. This inputStream is for now,
just cloneable.
When the parent process sends a Blob to a content process, it has the Blob and
its inputStream. With its inputStream it creates a IPCBlobInputStreamParent
actor. This actor keeps the inputStream alive for following uses (not part of
this patch).
On the child side we will have, of course, a IPCBlobInputStreamChild actor.
This actor is able to create a IPCBlobInputStream when CreateStream() is
called. This means that 1 IPCBlobInputStreamChild can manage multiple
IPCBlobInputStreams each time one of them is cloned. When the last one of this
stream is released, the child actor sends a __delete__ request to the parent
side; the parent will be deleted, and the original inputStream, on the parent
side, will be released as well.
IPCBlobInputStream is a special inputStream because each method, except for
Available() fails. Basically, this inputStream cannot be used on the content
process for nothing else than knowing the size of the original stream.
In the following patches, I'll introduce an async way to use it.
This is the first use of IPCBlob: ClonedMessageData.
ClonedMessageData is used for BroadcastChannel, MessagePort and any
postMessage() communication. This patch changes StructuredCloneData in order to
use IPCBlob instead of PBlob.
BroadcastChannel has a custom way to manage Blobs because when the parent
receives them from a content process, it must send them to any other
BroadcastChild actor duplicating the serialization.
The content process stores the incoming initial gfxVars updates, which are
lazily used when the gfxVars are first initialized.
MozReview-Commit-ID: ExUVdr5xGLb
--HG--
extra : rebase_source : fd6f3e1bc4eabdd85447eff0c0fa22537747431f
Remove sync protocol AllocateTabId. Instead we generate tabId in
each process with nsContentUtils::GenerateTabId, and register
RemoteFrameInfo in parent process. If the tab id was generated from
a content process, it's sent parent through either PBrowserConstructor
or PContent::CreateChildProcess.
MozReview-Commit-ID: D3W2fK9eCNH
--HG--
extra : rebase_source : 1913f8f586537be1c82a70a19cc8c6351671d0df
This requires running code which checks whether or not the permissions have
arrived, potentially delaying a fetch request for a very short period of time if
the permissions are still in-flight.
MozReview-Commit-ID: E6OTY6IDThb
Full Firefox on Linux can now be run with a --headless flag.
This includes seven parts:
1) Running all marionette tests in headless mode.
2) Prevents crashes where Firefox calls into GTK.
3) Adds a headless screen helper which supports changing the headless
screen size with the environment variables MOZ_HEADLESS_WIDTH and
MOZ_HEADLESS_HEIGHT.
4) Supports simulating moving a headless window.
5) Adds a stubbed out nsSound implementation.
6) Supports simulating size mode changes of headless windows.
7) Adds the --headless flag for Firefox.
This will return true if any of the frames loaded in the associated
TabChild have set at least one onbeforeunload event handler. If those
handlers are all removed, or all of the documents with onbeforeunload
event handlers are unloaded, this becomes false again.
Note that subframes that are sandboxed without the allow-modals
permission will not affect the hasBeforeUnload attribute, since
those iframes should never cause the beforeunload confirmation
dialog to display.
MozReview-Commit-ID: 8b0gBYWwMDn
--HG--
extra : rebase_source : 69f3b692d6e73f6277e6982aad02bcd1ebdd8acf
IMEContentObserver may need to change notifications to send when TextInputProcessor begins input transaction. In current design, IMEContentObserver needs to retrieve IMENotificationRequests at every change. However, if nsIWidget returns a reference to its IMENotificationRequests, IMEContentObserver can call it only once.
For that purpose, this patch changes nsIWidget::GetIMENotificationRequests() to nsIWidget::IMENotificationRequestsRef() and make it return |const IMENotificationRequests&|. However, if the lifetime of the instance of IMENotificationRequest is shorter than the widget instance's, it's dangerous. Therefore, it always returns TextEventDispatcher::mIMENotificationRequests. TextEventDispatcher's lifetime is longer than the widget. Therefore, this guarantees the lifetime.
On the other hand, widget needs to update TextEventDispatcher::mIMENotificationRequests before calls of nsIWidget::IMENotificationRequestsRef(). Therefore, this patch makes TextEventDispatcher update proper IMENotificationRequests when it gets focus or starts new input transaction and clear mIMENotificationRequests when it loses focus.
Note that TextEventDispatcher gets proper requests both from native text event dispatcher listener (typically, implemented by native IME handler class) and TextInputProcessor when TextInputProcessor has input transaction because even if TextInputProcessor overrides native IME, native IME still needs to know the content changes since they may get new input transaction after that.
However, there may not be native IME handler in content process. If it runs in Android, PuppetWidget may have native IME handler because widget directly handles IME in e10s mode for Android. Otherwise, native IME handler is in its parent process. So, if TextInputHandler has input transaction in content process, PuppetWidget needs to behave as native event handler. Therefore, this patch makes PuppetWidget inherit TextEventDispatcherListener and implements PuppetWidget::IMENotificationRequestsRef().
MozReview-Commit-ID: 2SW3moONTOX
--HG--
extra : rebase_source : d2634ada6c33dbf7a966fadb68608411ee24bfab
This patch will go away when I'll finishing the removing of PBlob. Currently,
when a PBlob is sent from child to parent, we use PMemoryStream in order to
recreate the inputStream on the parent side. PMemoryStream sends the data in
chunks.
But if PBlob is dealing with a IPCBlobInputStream, it doesn't have access to
the real data. In this case, we must send data using IPCStream. In this way,
Note that thisIPCBlobInputStream will send its ID, and the parent will take the
real inputStream from the IPCBlobInputStreamStorage. Note that I check the
size to be 1mb instead 0. No particular reasons, but better to avoid the use of
PMemoryStream for nothing.
IPCBlobInputStream is a new type of nsIInputStream that is used only in content
process when a Blob is sent from parent to child. This inputStream is for now,
just cloneable.
When the parent process sends a Blob to a content process, it has the Blob and
its inputStream. With its inputStream it creates a IPCBlobInputStreamParent
actor. This actor keeps the inputStream alive for following uses (not part of
this patch).
On the child side we will have, of course, a IPCBlobInputStreamChild actor.
This actor is able to create a IPCBlobInputStream when CreateStream() is
called. This means that 1 IPCBlobInputStreamChild can manage multiple
IPCBlobInputStreams each time one of them is cloned. When the last one of this
stream is released, the child actor sends a __delete__ request to the parent
side; the parent will be deleted, and the original inputStream, on the parent
side, will be released as well.
IPCBlobInputStream is a special inputStream because each method, except for
Available() fails. Basically, this inputStream cannot be used on the content
process for nothing else than knowing the size of the original stream.
In the following patches, I'll introduce an async way to use it.
This is the first use of IPCBlob: ClonedMessageData.
ClonedMessageData is used for BroadcastChannel, MessagePort and any
postMessage() communication. This patch changes StructuredCloneData in order to
use IPCBlob instead of PBlob.
BroadcastChannel has a custom way to manage Blobs because when the parent
receives them from a content process, it must send them to any other
BroadcastChild actor duplicating the serialization.
Everything depending on the widget being gonk can go away, as well as
everything depending on MOZ_AUDIO_CHANNEL_MANAGER, which was only
defined on gonk builds under b2g/ (which goes away in bug 1357326).
--HG--
extra : rebase_source : 9f0aeeb7eea8417fa4e06d662d566d67ecaf2a24
This patch centralizes all of the pref-checking code for e10s-multi in a
single function. It is intended to be used throughout the codebase to see if
e10s-multi is "on". It also introduces dom.ipc.multiOptOut, which can be set
by the user to indicate that they do not want to participate in the e10s-multi
experiment.
MozReview-Commit-ID: Kyq1fqNzwue
--HG--
extra : rebase_source : 321a44fb5909c067a20dbb3b739175ba08569a5f
This patch does a few things:
a) Adds the resources location from the .app directory to the read whitelist
b) When it's a non-packaged build, mach run (and various mach tests) set an environment variable for the repo location which we allow reads from.
r=haik,froydnj
MozReview-Commit-ID: KNvAoUs5Ati
--HG--
extra : rebase_source : 81ba8bfee0ca96979cf8e30d75cdd47f06bc10ea
Our caller is C++ code, and the implementations are all also written in C++,
so there is no reason to go through SpiderMonkey here. This patch also makes
nsILoadContext builtinclass to ensure that the implementation is always native.
This piggybacks the sync message on the pre-existing
EnsureLayersConnected sync message pathway to the compositor.
MozReview-Commit-ID: DfYTlJrr3Gu
--HG--
extra : rebase_source : c2bd29e655db65e3016a79bf3f6068ffb1c8b7c7
The goal of this patch is to remove the call to the sync IPC
GetCompositorOptions message from TabChild::InitRenderingState. In order
to this, we have InitRenderingState take the CompositorOptions as an
argument instead, and propagate that backwards through the call sites.
Eventually we can propagate it back to a set of already-sync IPC
messages in PCompositorBridge that are used during layers id
registration (NotifyChildCreated, NotifyChildRecreated, etc.). Therefore
this patch effectively piggybacks the CompositorOptions sync IPC onto
these pre-existing sync IPC messages.
The one exception is when we propagate it back to the AdoptChild call.
If this message were sync we could just use it like the others and have
it return a CompositorOptions. However, it is async, so instead we add
another call to GetCompositorOptions here temporarily. This will be
removed in the next patch.
MozReview-Commit-ID: AtdYOuXmHu4
--HG--
extra : rebase_source : 5b80831cf84d3a4b57b2214a12ccf8a896cfa3a7
We add a new "on-off" protocol PURLClassifierLocal which calls
nsIURIClassifier.asyncClassifyLocalWithTables on construction and
calls back on destruction. Pretty much the same design as PURLClassifier.
In order to avoid code duplication, the actor implementation is templatized
and |MaybeInfo| in PURLClassifier.ipdl is moved around.
Test case is included and the custom event target is not in place for labelling.
The custom event target will be done in Bug 1353701.
MozReview-Commit-ID: IdHYgdnBV7S
--HG--
extra : rebase_source : ab1c896305b9f76cab13a92c9bd88c2d356aacb7
When a subprocess is launched, gfxVars updates (for non-default values) are
serialized and passed on the command line, up to a limit of 1023 characters,
and ensuring it should not overflow the command line size.
When the child starts, the command line parameter is given to gfxVars, so the
updates can be used during gfxVars::Initialize(), instead of doing a sync
request to the parent.
In case the updates are not sent, or in the unlikely case the child cannot
parse them, we fallback to the sync request -- The former case should be rare
enough that a slow sync request is acceptable: It should only happen if D3D
block-list is *modified* (most people would either use the default, or just
overwrite these prefs with short strings.)
MozReview-Commit-ID: 6MoJC0fe59Q
--HG--
extra : rebase_source : cdc2e451783160c579b8fc84050e8457c600523e
When we're animating, we tick the refresh driver. If that occurs in the parent process
when e10s is enabled, then we currently run TabParent::DidRefresh which does some
dimensions calculations and might send a message to the content process if the
dimensions have changed.
This was originally added to fix a B2G bug in bug 1153023. We don't need to do it
anymore, since we don't set CSS transforms on content browser windows.
MozReview-Commit-ID: JJ7AJHlSyWn
--HG--
extra : rebase_source : b45c9f02c3db8b7ecf0beb40fa7540db39473e8d
MozReview-Commit-ID: GTQF3x1pBtX
A general outline of the COM handler (a.k.a. the "smart proxy"):
COM handlers are pieces of code that are loaded by the COM runtime along with
a proxy and are layered above that proxy. This enables the COM handler to
interpose itself between the caller and the proxy, thus providing the
opportunity for the handler to manipulate an interface's method calls before
those calls reach the proxy.
Handlers are regular COM components that live in DLLs and are declared in the
Windows registry. In order to allow for the specifying of a handler (and an
optional payload to be sent with the proxy), the mscom library allows its
clients to specify an implementation of the IHandlerProvider interface.
IHandlerProvider consists of 5 functions:
* GetHandler returns the CLSID of the component that should be loaded into
the COM client's process. If GetHandler returns a failure code, then no
handler is loaded.
* GetHandlerPayloadSize and WriteHandlerPayload are for obtaining the payload
data. These calls are made on a background thread but need to do their work
on the main thread. We declare the payload struct in IDL. MIDL generates two
functions, IA2Payload_Encode and IA2Payload_Decode, which are used by
mscom::StructToStream to read and write that struct to and from buffers.
* The a11y payload struct also includes an interface, IGeckoBackChannel, that
allows the handler to communicate directly with Gecko. IGeckoBackChannel
currently provides two methods: one to allow the handler to request fresh
cache information, and the other to provide Gecko with its IHandlerControl
interface.
* MarshalAs accepts an IID that specifies the interface that is about to be
proxied. We may want to send a more sophisticated proxy than the one that
is requested. The desired IID is returned by this function. In the case of
a11y interfaces, we should always return IAccessible2_3 if we are asked for
one of its parent interfaces. This allows us to eliminate round trips to
resolve more sophisticated interfaces later on.
* NewInstance, which is needed to ensure that all descendent proxies are also
imbued with the same handler code.
The main focus of this patch is as follows:
1. Provide an implementation of the IHandlerProvider interface;
2. Populate the handler payload (ie, the cache) with data;
3. Modify CreateHolderFromAccessible to specify the HandlerPayload object;
4. Receive the IHandlerControl interface from the handler DLL and move it
into the chrome process.
Some more information about IHandlerControl:
There is one IHandlerControl per handler DLL instance. It is the interface that
we call in Gecko when we need to dispatch an event to the handler. In order to
ensure that events are dispatched in the correct order, we need to dispatch
those events from the chrome main thread so that they occur in sequential order
with calls to NotifyWinEvent.
--HG--
extra : rebase_source : acb44dead7cc5488424720e1bf58862b7b30374f
There are scenarios where we have a TabParent in the UI process hooked up to
a PuppetWidget with a BasicLayerManager. Webextensions fall into this category.
In this scenario, the parent-side layer manager is not hooked up to
the compositor (that is, there is no entry in the CompositorBridge layer tree
state map for the layers id). However, the content-side still ends up creating
a ClientLayerManager or a WebRenderLayerManager, which expects the layers id to
be registered in the compositor. This results in brokenness (in the case of the
ClientLayerManager/PLayerTransaction) or crashes (in the case of WebRenderLayerManager/
PWebRenderBridge). Instead, this patch changes this scenario to have the content
process use a BasicLayerManager which seems safer.
MozReview-Commit-ID: 3f80aZrRrmD
--HG--
extra : rebase_source : 10ec78dd7daf1c1c889929f0d79e0b75675b4b05
Instead we can use the otherwise-unused function to check if initialization worked.
Technically, because of the way RenderFrame construction works, whether or not
initialization succeeds is exactly dependent on whether or not the frameloader is
non-null, so we could even just use that to check success. But this feels slightly
more readable/cleaner.
MozReview-Commit-ID: CGiIAA1h6V7
--HG--
extra : rebase_source : 2d4504f7d5fde3c394edca14fe01840c1738d197
This avoids loading 7 jsms until they are actually needed, reducing
content process memory usage.
MozReview-Commit-ID: 8ukNr7PL80A
--HG--
extra : rebase_source : a926a19645ffcacadec93ecc79927d356d603114
I suspect that the PuppetWidget is trying to create the layer manager after
it has been connected to a TabChild but before the TabChild has populated the
CompositorOptions. This results in the PuppetWidget effectively getting an
uninitialized value for the CompositorOptions, and so it sometimes randomly
creates a WebRenderLayerManager, later resulting in a crash.
It seems like exposing the potentially-uninitialized CompositorOptions from
TabChild like this is a bad idea, so I'm removing that API and using the more
reliable gfxVars in PuppetWidget. This is fine for WebRender purposes because
we no longer care to allow having WR compositors co-exist with non-WR
compositors.
We may eventually want to remove the CompositorOptions entirely, but for now
the rest of the usage of it seems fine.
MozReview-Commit-ID: 6ekG8j1PskK
--HG--
extra : rebase_source : 0099e847ac356ca235969bcd81f47d65f49de2eb
This is the most important part of the patch series. It removes the
PScreenManager protocol and use ScreenManager directly in the content
processes.
Initial and subsequent updates are sent via PContent::RefreshScreens.
struct ScreenDetails are kept to serialize Screen over IPC.
nsIScreenManager::ScreenForNativeWidget is removed because
nsIWidget::GetWidgetScreen can replace it. nsIScreen::GetId is removed
because it's not useful for the more general Screen class.
MozReview-Commit-ID: 5dJO3isgBuQ
--HG--
extra : rebase_source : 06aa4e4fd56e2b2af1e7483aee7c0cc7f35bdb97
It's not used anywhere. Remove it will make removing PScreenManager
easier.
MozReview-Commit-ID: 5dn8kDhTZVl
--HG--
extra : rebase_source : 96b8ddb18deee94ca256bfa118b60ceacfd2d677
The pref cache added in the previous commit initializes at startup.
MozReview-Commit-ID: IvfBALLdcbe
--HG--
extra : rebase_source : b3c63c76f9eda144e6ae2c8801ac8637444f6c48
Our caller is C++ code, and the implementations are all also written in C++,
so there is no reason to go through SpiderMonkey here. This patch also makes
nsILoadContext builtinclass to ensure that the implementation is always native.
Our caller is C++ code, and the implementations are all also written in C++,
so there is no reason to go through SpiderMonkey here. This patch also makes
nsILoadContext builtinclass to ensure that the implementation is always native.
These APIs are intended to use the mechanism defined in Part 1.
Part 3 implements the usage of these APIs to synchronize permissions.
MozReview-Commit-ID: HNKyDPtoaHl
Our caller is C++ code, and the implementations are all also written in C++,
so there is no reason to go through SpiderMonkey here. This patch also makes
nsILoadContext builtinclass to ensure that the implementation is always native.
We're now asserting that we never check these before the END_ALL_PREFS phase,
which means they don't need to be sent to the content process synchronously.
MozReview-Commit-ID: 4BGbvVCjDWz
This adds 'media.cubeb.backend' to ContentPrefs, which is necessary because
`cubeb_init` is called _very_ early in the lifetime of a content process,
because it needs to be called before enabling seccomp.
the << operator of std:stringstream is occationally hanging on Windows 7 PGO builds, this uses nsPrintfCString to do formatting rather than stringstreams and replaces the std::string with nsFixedCSting for string building
CLOSED TREE
Backed out changeset d24fa1b4553b (bug 1346987)
Backed out changeset 34701b9ed4ba (bug 1346987)
Backed out changeset f24f4fdc5cc8 (bug 1346987)
These APIs are intended to use the mechanism defined in Part 1.
Part 3 implements the usage of these APIs to synchronize permissions.
MozReview-Commit-ID: HNKyDPtoaHl
If Firefox is updated while it is running, the content process can end
up being a different version than the parent process. This can cause
odd crashes, that will happen repeatedly until the user restarts
Firefox. To handle this better, this patch adds a special build ID
message that is sent early in content process startup. The parent
process intentionally crashes if the build ID for the child process
does not match that of the parent process.
MozReview-Commit-ID: 7D3ggkaLxNS
--HG--
extra : rebase_source : 1f8d917ce01919524f949dd5bedfbbbd557f7ed3
Instead of initializing DataStorage objects on demand in the content
process, we initialize them at content process startup by getting the
parent to send down the information about the existing DataStorages at
child process startup. After that point, the dynamic change
notifications added in bug 1215723 will take care of keeping the
information in sync.
This change includes the FIDO "App ID" as part of the function used to generate
the wrapping key used in the NSS-based U2F soft token, cryptographically binding
the "Key Handle" to the site that Key Handle is intended for.
This is a breaking change with existing registered U2F keys, but since our soft
token is hidden behind a pref, it does not attempt to be backward-compatible.
- Updated for rbarnes' and qdot's reviews comments. Thanks!
- Made more strict in size restrictions, and added a version field
to help us be this strict.
- Bugfix for an early unprotected buffer use (Thanks again rbarnes!)
- Fix a sneaky memory leak re: CryptoBuffer.ToSECItem
MozReview-Commit-ID: Jf6gNPauT4Y
--HG--
extra : rebase_source : 4ff5898e93e4a0a75576e5e54035a1cb6dd952d7
The MozTabChildNotReady event is queued from the ProcessHangMonitor thread, and
that event might try to fire _after_ the main thread has received notification
that a layer tree was made ready. If that occurs, firing the MozTabChildNotReady
event makes no sense, as clearly the TabChild _became_ ready by the time the
event was about to be dispatched.
MozReview-Commit-ID: Iigtc0SqzeU
--HG--
extra : rebase_source : c01eedb27133d37f348a3b697358d0bfa5420c41
This patch is generated by the following sed script:
find . ! -wholename '*/.hg*' -type f \( -iname '*.html' -o -iname '*.xhtml' -o -iname '*.xul' -o -iname '*.js' \) -exec sed -i -e 's/\(\(text\|application\)\/javascript\);version=1.[0-9]/\1/g' {} \;
MozReview-Commit-ID: AzhtdwJwVNg
--HG--
extra : rebase_source : e8f90249454c0779d926f87777f457352961748d
profiler_get_gatherer() exposes ProfileGatherer to the outside world in a way
that makes future changes difficult.
This patch:
- Removes ProfileGatherer.h from the list of headers exported from the
profiler.
- Removes nsIProfiler.profileGatherer and nsProfiler::GetProfileGatherer().
- Replaces profiler_get_gatherer() with three new functions that provide
minimal but sufficient access to ProfileGatherer:
profiler_will_gather_OOP_profile(), profiler_gathered_OOP_profile(), and
profiler_OOP_exit_profile().
These functions provide access to the ProfileGatherer in a similar fashion to
the pre-existing functions profiler_get_profile_jsobject_async() and
profiler_save_profile_to_file_async()
This significantly reduces the size of the profiler's API surface.
--HG--
rename : tools/profiler/public/ProfileGatherer.h => tools/profiler/gecko/ProfileGatherer.h
extra : rebase_source : d8e06a1133d4098c3a214858d3ff2c4bdcd9f1f2
It looks like it used to be called by multiple constructors, but that's no
longer the case.
--HG--
extra : rebase_source : 2aba63c0786ddd0ce355ed80569cad62810393ad
Andrew Osmond
Gabor Krizsanits
Marco Castelluccio
Mats Palmgren
Wes Kocher
Ryan Hunt
Masayuki Nakano
Alex Gaynor
Stephen A Pohl
Aaron Klotz
Michael Layzell
Christoph Kerschbaumer
Andrea Marchesini
Kyle Machulis
Ryan VanderMeulen
Gabriele Svelto
Markus Stange
Kris Maglione
Peter Van der Beken
Carsten "Tomcat" Book
sotaro
Eden Chuang
Bill McCloskey
Ehsan Akhgari
Mike Conley
Georg Fritzsche
Sebastian Hengst
Phil Ringnalda
Yoshi Huang
btian
Andrew McCreight
Nathan Froyd
Olli Pettay
David Anderson
Kartikaya Gupta
Gerald Squelart
Nicholas Nethercote
Kershaw Chang
David Parks
Ehsan Akhgari
Iris Hsiao
David Anderson
Zibi Braniecki
Samael Wang
Brendan Dahl
Edgar Chen
Valentin Gosu
Mike Hommey
Blake Kaplan
Farmer Tseng
Henry Chang
Sotaro Ikeda
Florian Queze
Geoff Brown
Kevin Chen
Kan-Ru Chen
Chris Pearce
Jessica Jong
Boris Zbarsky
Damien Zammit
Brad Lassey
Evelyn Hung
Stone Shih
J.C. Jones
Thomas Nguyen
Daniel Holbert
Trevor Saunders
Joel Maher
Masatoshi Kimura
Benjamin Smedberg
Eugen Sawin
Jean-Yves Avenard
Nicolas Silva
Cameron McCormack
Andrew Sutherland
Eitan Isaacson
Sylvestre Ledru
Jonathan Hao
Tom Tromey