StripHandlerFromOBJREF shortens the OBJREF by sizeof(CLSID), so it needs to seek the stream back after tweaking the OBJREF.
Previously, this was done using a relative seek.
Unfortunately, for some reason I can't fathom on Windows 7, this doesn't work when marshaling for VT_DISPATCH.
The Seek call succeeds, but either does nothing or sets the stream position to a garbage value.
Instead, we now use an absolute seek, which seems to behave.
This was breaking IAccessible::accNavigate and AccessibleChildren on Windows 7.
MozReview-Commit-ID: FEH93oiyP5R
--HG--
extra : rebase_source : b15db60da888b49cbd371bc5c8311577a2c7ece4
This was used to support cross-architecture NPAPI plugins on OS X, but
we stopped supporting that in 54 (bug 1339182).
MozReview-Commit-ID: 2BcWYD6mguY
--HG--
extra : rebase_source : 6e509a3cc1f356ccd24f1459c43bc8fb66d7b0f4
This patch adds two additional fields to each mscom log entry: The first is
the duration, in microseconds, of time spent in mscom overhead when executing
a call from the MTA on behalf of a remote client.
The second field is the duration, in microseconds, of time spent actually
executing the method within Gecko itself.
(In other words, the sum of the two fields will equal the total duration of
time spent executing the call.)
MozReview-Commit-ID: EhFieEPrhE5
As its original comments indicate, SetAllFDsToCloseOnExec has an
unavoidable race condition if another thread creates file descriptors
during launch. Instead, use POSIX_SPAWN_CLOEXEC_DEFAULT, which is an
Apple-specific extension to posix_spawn that accomplished the desired
effect atomically.
This patch also introduces some RAII to simplify cleanup in error cases.
MozReview-Commit-ID: 6oHggs77AiY
--HG--
extra : rebase_source : a9391031a95fee4977af800ca993871277db51ce
The current API makes the life time and ownership of the result array unclear
without careful reading. The result array is always owned by the principal,
and its lifetime tied to the lifetime of the principal itself. Returning a
const array reference makes this clear, and should prevent callers from
accidentally modifying the returned array.
MozReview-Commit-ID: 3f8mhynkKAj
--HG--
extra : source : 237acf2879f6222bc4b076c377bf026d18a6ebef
extra : amend_source : dfaf6e88e3c4758f7fdcf7fb422d457edafab1b7
The current API makes the life time and ownership of the result array unclear
without careful reading. The result array is always owned by the principal,
and its lifetime tied to the lifetime of the principal itself. Returning a
const array reference makes this clear, and should prevent callers from
accidentally modifying the returned array.
MozReview-Commit-ID: 3f8mhynkKAj
--HG--
extra : rebase_source : d2a5e0862f8c964fb5a3e46b50c2e9629b218699
extra : amend_source : 27d7a7ef5da6fe2aa1104009b6ee067465db73e1
ChildPrivileges is a leftover from the B2G process model; it's now
mostly unused, except for the Windows sandbox using it to carry whether
a content process has file:/// access.
In general, when sandboxing needs to interact with process launch, the
inputs are some subset of: the GeckoProcessType, the subtype if content,
various prefs and even GPU configuration; and the resulting launch
adjustments are platform-specific. And on some platforms (e.g., OS X)
it's all done after launch. So a simple enum used cross-platform isn't
a good fit.
MozReview-Commit-ID: K31OHOpJzla
--HG--
extra : rebase_source : 3928b44eb86cd076bcac7897536590555237b76b
This is mostly based on the BSD version, which in turn is more or less
the Mac version minus some race conditions. The Linux version does
something similar, but more verbosely and (at least in my opinion) is
harder to follow. Some changes have been made, mainly to use C++11
features like UniquePtr.
MozReview-Commit-ID: 3Gv4DKCqWvu
--HG--
extra : rebase_source : 972264a778b9361d1259851554b5b7ae8f3dcdc6
Currently the Gecko Profiler defines a moderate amount of stuff when
MOZ_GECKO_PROFILER is undefined. It also #includes various headers, including
JS ones. This is making it difficult to separate Gecko's media stack for
inclusion in Servo.
This patch greatly simplifies how things are exposed. The starting point is:
- GeckoProfiler.h can be #included unconditionally;
- everything else from the profiler must be guarded by MOZ_GECKO_PROFILER.
In practice this introduces way too many #ifdefs, so the patch loosens it by
adding no-op macros for a number of the most common operations.
The net result is that #ifdefs and macros are used a bit more, but almost
nothing is exposed in non-MOZ_GECKO_PROFILER builds (including
ProfilerMarkerPayload.h and GeckoProfiler.h), and understanding what is exposed
is much simpler than before.
Note also that in BHR, ThreadStackHelper is now entirely absent in
non-MOZ_GECKO_PROFILER builds.
Avoid going through GeckoAppShell and move the start child process JNI
call directly to GeckoProcessManager.
MozReview-Commit-ID: KU62TiHVQJX
--HG--
extra : rebase_source : 0e8546da502257e1c59bc00b79f50c79a314f3e6
After the previous patch, all that is needed to pass eslint is some
whitespace fixes I generated using:
./mach eslint --fix ipc
The .eslintrc.js file makes eslint expect XPCShell global variables.
With those two changes, eslint can be enabled for the ipc/ directory.
MozReview-Commit-ID: BqSICp2iV6O
--HG--
extra : rebase_source : ccab8be9fc1ae1116ae55631d3978c39d35cad05
I don't fully understand what this test is trying to do, but as far as
I can see, test_ipcshell_child.js does nothing when loaded from
test_ipcshell.js besides define a few variables, so just define them
directly in test_ipcshell.js. I am doing this because eslint does not
like the way the test is currently written: the definition of runtime
in another file confuses it, and it does not like the assignment to
run_test.
MozReview-Commit-ID: Ek0HY9n49eZ
--HG--
extra : rebase_source : 5c99b20afe93c085a198e7824e82207be347f26b
Never store names in Message. One can get string names from
Message::name() or use IPC::StringFromIPCMessageType() when only
message id is available.
MozReview-Commit-ID: 15ksx6SE90c
--HG--
extra : rebase_source : 1a041dc365b7f42edd540d8c7a4dfd8912e48921
xpcshell can sometimes create and then almost immediately kill the GPU process. This can
cause the connection state of GPUParent's message channel to be in error or closed
when it attempts to send its build ID. In this case we should ignore the request
because the process will soon close.
MozReview-Commit-ID: EAupU844iFo
--HG--
extra : histedit_source : df05a01df5ad095b4e3c91e0e038d4e145da5f9d
The NS_LITERAL_CSTRING macro creates a temporary nsLiteralCString to encapsulate the string literal and its length, but AssignLiteral() can determine the string literal's length at compile-time without nsLiteralCString.
MozReview-Commit-ID: B5Y8KyExPQ8
--HG--
extra : rebase_source : e27b266c145daa5acd887e998c6d5b408101e1db
extra : source : 33f49977a33cbdb1c7127871b940eefccc018f65
Each protocol in IPDL has a bunch of autogenerated functions that
instantiate IPC::Message with various parameters. Each of these
functions, then:
1) Pays the cost of calling malloc()
2) Setting up various parameters
3) Calling IPC::Message()
There's no reason that we should be duplicating 1) across all of these
autogenerated functions. In step 2), several of the parameters we're
setting up are common across all or nearly all calls: the message
segment size is almost always zero, and we're always indicating that
IPDL-generated messages should be recorded in telemetry.
Instead of duplicating that code several thousand times, we can add a
small helper function that takes the only interesting parameters for an
IPDL message. This helper function can then deal with calling malloc in
a single place and setting up the common parameters. For messages that
require a custom segment size, we'll have to use the old scheme, but
such messages are uncommon.
The previous changes are not required for this scheme to work, but they
do help significantly, as the helper function (Message::IPDLMessage) can
now take four parameters, which ensures that its arguments are passed
solely in registers on Win64 and ARM. The wins from this change are
also larger than they would be without the previous parts: ~100K on
x86-64 Linux (!) and ~80K on ARM Android.
The current IPC::Message constructor takes a large number of arguments,
three of which--the nesting level, the priority, and the
compression--are almost always constant by virtue of the vast majority
of Message construction being done by auto-generated IPDL code. But
then we take these constant values into the Message constructor, we
check them for various values, and then based on those values, we
perform a bunch of bitfield operations to store flags based on those
values. This is wasted work.
Furthermore, for replies to IPDL messages, we'll construct a Message
object, and then call mutating setters on the Message object that will
perform even more bitfield manipulations. Again, these operations are
performing tasks at runtime that are the same every single time, and use
information we already have at compile time.
The impact of these extra operations is not large, maybe 15-30K of extra
code, depending on platform. Nonetheless, we can easily make them go
away, and make everything cleaner to boot.
This patch adds a HeaderFlags class that encapsulates all the knowledge
about the various kinds of flags Message needs to know about. We can
construct HeaderFlags objects with strongly-typed enum arguments for the
various kinds of flags, and the compiler can take care of folding all of
those flags together into a constant when possible (and it is possible
for all the IPDL-generated code that instantiates Messages). The upshot
is that we do no unnecessary work in the Message constructor itself. We
can also remove various mutating operations on Message, as those
operations were only there to support post-constructor flag twiddling,
which is no longer necessary.
There's no need to be repeating 'IPC::Message::' prefixes or spreading
around more ExprVar calls than we need here. Let's try to improve the
signal-to-noise ratio of this code by introducing a helper function to
inject some of the boilerplate for us.
_generateMessageConstructor takes a lot of `md.FOO`-style parameters,
which could be derived inside the function by simply passing `md`.
Especially with the upcoming changes to calculate things like reply-ness
of messages, sync-ness, etc, we'd be wanting to pass even more
parameters like `md.FOO`. So let's just pass `md` in, and then we can
make all the necessary future changes in a single place.
The GPU process doesn't have the directory service enabled, so it can't find
a tmp dir to put its .extra files for crash reports. Even if we do enable the
directory service, we still don't get the correct "content process tmp dir" in
the GPU process, because the UUID baked into that folder is passed via the
preferences service, and that isn't initialized in the GPU process either.
Rather than unneccessarily initialize all this stuff in the GPU process just
to get one folder name, we can pass that folder name directly in the argv list.
See comments 12-19 on the bug for further discussion of the various solutions
attempted/explored.
MozReview-Commit-ID: 1sFg27hIe7S
--HG--
extra : rebase_source : 78eb600a58fed45800b9df8303cc4d6898d96ae9
This function is very large and has many nested ifdefs. Without this cleanup
it was really hard to figure out where all I needed to add code.
MozReview-Commit-ID: IeA0AWW62Mn
--HG--
extra : rebase_source : a8447df42c60da79f1c982111c8f65bd0175f1de
The GPU process doesn't have the directory service enabled, so it can't find
a tmp dir to put its .extra files for crash reports. Even if we do enable the
directory service, we still don't get the correct "content process tmp dir" in
the GPU process, because the UUID baked into that folder is passed via the
preferences service, and that isn't initialized in the GPU process either.
Rather than unneccessarily initialize all this stuff in the GPU process just
to get one folder name, we can pass that folder name directly in the argv list.
See comments 12-19 on the bug for further discussion of the various solutions
attempted/explored.
MozReview-Commit-ID: 1sFg27hIe7S
--HG--
extra : rebase_source : 62ff819c5b03f642cd4b9af7c89c84b790397372
This function is very large and has many nested ifdefs. Without this cleanup
it was really hard to figure out where all I needed to add code.
MozReview-Commit-ID: IeA0AWW62Mn
--HG--
extra : rebase_source : a8447df42c60da79f1c982111c8f65bd0175f1de
The nsIU2FToken and its implementors are no longer needed; the soft token was
re-implemented into dom/webauthn/U2FSoftTokenManager.cpp during the WebAuthn
implementation. When the dom/u2f/ code changed to the implementation from
WebAuthn, the old synchronous version became dead code.
This patch removes the dead code.
MozReview-Commit-ID: 2yDD0tccgZr
--HG--
extra : rebase_source : 0f14d8de8f62599a41c13aa4d8fc9cdbc1fd79c7
Since LinearHistogram and its descendants inherit ranges_ from
Histogram, and we wanted to replace the copying into a std::vec
for Histogram, the simplest approach seemed to just be to
precompute ranges for all histograms, exponential or otherwise.
This should have the added benefit of reducing the memory
footprint for those histograms, since they will benefit from the
deduplication work that the precomputing script already does.
MozReview-Commit-ID: JTV5Dej5ZIb
--HG--
extra : rebase_source : de942d54b3475be54c70d43d2fa8e772ee2e18c4
Since NoFocusState is am empty struct used in the |mData| variant in
FocusTarget, we need to add a Reader and a Writer for IPC for NoFocusState so we
can properly read and write the |mData| variant. The NoFocusState Read and Write
methods do not read or write anything, since NoFocusState does not contain any
data. This is done by creating a helper class EmptyStructSerliazer and
inheritting from EmptyStructSerializer for the NoFocusState specialization.
The |Read| and |Write| methods for FocusTarget are updated by removing the read
and write code for the individual types of |mData| and instead makes use of the
IPC read and write methods for Variant.
MozReview-Commit-ID: 3159sp6FLek
--HG--
extra : rebase_source : ff82797c26abffbd8bbcc465dc40c621be3257d0
* VariantWriter construction switched to use aggregate initialization
* Call to AsVariant was inappropriately called via |paramType| when it should
have been called via |mozilla|
* |Next::Read| call in VariantReader specialization was missing the |result|
argument
MozReview-Commit-ID: Izany7iDX0k
--HG--
extra : rebase_source : 7387e72100c7d2ba8fcfd1e5a3b6d0ce6be6c740
It's important that shmem creation/destruction messages be ordered
correctly with respect to other messages that use shmems. For example,
if we create a shmem with ID 10 and then send a message that
references shmem 10, then the creation message must be handled before
the referencing message. If shmem creation/destruction messages go in
a separate queue from other messages, this ordering may not be
preserved.
Leaving shmem creation/destruction unlabeled will give us the correct
ordering. Eventually, though, we'll need to provide a solution that
doesn't bottleneck the event queue.
MozReview-Commit-ID: 88MrslRrfnh
Sebastian Hengst
Gian-Carlo Pascutto
Andrew McCreight
Andrea Marchesini
James Teh
Aaron Klotz
Chris Manchester
sotaro
Attila Craciun
Sylvestre Ledru
Jed Davis
Kris Maglione
Dan Banner
evilpies@gmail.com
Nicholas Nethercote
Tom Ritter
Jim Chen
Mike Hommey
Wes Kocher
Benjamin Smedberg
Arthur Edelstein
Petr Sumbera
Ryan VanderMeulen
Kan-Ru Chen
Ryan Hunt
Bevis Tseng
Bill McCloskey
Dan Minor
Chris Peterson
Nathan Froyd
Markus Stange
Kartikaya Gupta
J.C. Jones
Bob Owen
Doug Thayer
Jeff Hajewski