For some reason the current code is resetting it twice - once explicitly and
once with the AutoSetOnScopeExit. To make matters worse, we have a monitor drop
between the two. So when DecodeSeek runs on the decode task queue but SeekCompleted
runs on the state machine thread, we can start another DecodeSeek during the monitor
drop, and then clobber it with the AutoSeetOnScopeExit, causing us to hang.
This is a non-issue with the patches in bug 1135170, but necessary to make the
patches in this bug independently green.
This already gets incoded in the DECODER_STATE_DORMANT case of RunStateMachine,
which will run momentarily on the state machine thread. Doing this allows us to
avoid calling StopPlayback on the main thread.
This is necessary because we're going to want to start disconnecting sample
and seek requests directly from the state machine thread, and the machinery
asserts that disconnection happens on the same thread as resolution.
More generally, this is the right thing to do architecturally, and will help
wean us off the monitor.
The current situation is really dangerous because it compiles on release builds,
but just lies. This bit me when I tried to use it for non-assertion purposes.
My reading of the reasoning for the current setup in bug 968016 is that we didn't
trust nsIEventTarget::IsCurrentThreadOn or thought it might be slow. But the
implementation of MediaTaskQueue::IsCurrentThreadIn doesn't actually use that, and
indeed currently does all of the work for this feature in release builds anyway.
Now that we have a mechanism for defining file-based metadata, let's add
a mach command to interface with it.
Currently, we limit ourselves to simple Bugzilla data dumping. Features
will be added over time.
The Files sub-context allows us to attach metadata to files based on
pattern matching rules.
Patterns are matched against files in a last-write-wins fashion.
The sub-context defines the BUG_COMPONENT variable, which is a 2-tuple
(actually a named tuple) defining the Bugzilla product and component for
files. There are no consumers yet. But an eventual use case will be to
suggest a bug component for a patch/commit. Another will be to
automatically suggest a bug component for a failing test.
We want the ability to read data from any moz.build file without needing
a full build configuration (running configure). This will enable tools
to consume metadata by merely having a copy of the source code and
nothing more.
This commit creates the EmptyConfig object. It is a config object that -
as its name implies - is empty. It will be used for reading moz.build
files in "no config" mode.
Many moz.build files make assumptions that variables in CONFIG are
defined and that they are strings. We create the EmptyValue type that
behaves like an empty unicode string. Since moz.build files also do some
type checking, we carve an exemption for EmptyValue, just like we do for
None.
We add a test to verify that reading moz.build files in "no config" mode
works. This required some minor changes to existing moz.build files to
make them work in the new execution mode.
moz.build files should execute in filesystem traversal mode. Add a test
that verifies this is true.
This test performs a brute force filesystem scan to find relevant
moz.build files. This can be a little slow. That's unfortunate. But it's
a price we need to pay in order to ensure metadata extraction mode
continues to work.
Building on top of the API to retrieve relevant moz.build files for a
given path, we introduce a moz.build reading API that reads all
moz.build files relevant to a given set of paths. We plan to use this
new API to read metadata from moz.build files relevant to a set of
files.
This patch changes the generator behavior of read_mozbuild to emit the
main context before any processing occurs. This allows downstream
consumers to manipulate state of the context before things like
directory processing occurs. We utilize this capability in the new
reading API to forcefully declare the directory traversal order for
processed moz.build files, overriding DIRS and similar variables.
Since variable exporting doesn't work reliably in this new traversal
mode, variable exporting no-ops when this mode is activated.
Currently, MozSandbox assumes that the FUNCTIONS, SPECIAL_VARIABLES, and
SUBCONTEXTS data structures are the instances that should be associated
with the sandbox. As we introduce new moz.build processing modes that
wish to change processing behavior, it is necessary for them to have
control over these special symbols.
This patch moves the declaration of these types to the special metadata
dictionary which is inherited during recursion. The "read_topsrcdir" API
now explicitly passes the initial metadata into "read_mozbuild".
We have an eventual goal to store file-level metadata in moz.build files
and to have this metadata "cascade" down directory hierarchies. e.g.
metadata in the root directory will apply to all children directories.
A prerequisite for this feature is a way to query which moz.build files
are relevant to a given file. In this patch, we implement an API that
returns this information.
With some GPUs (such as Intel HD-x000), Apple VideoTool box provides poor
decoding speed, causing us to drop frames for most HD videos.
VDA is around 50 times faster on those machines (31ms average to decode a 4K
frame with VT, while 0.6ms average with VDA)
Cairo has a number of nifty features predicated on support for atomic
operations on integer types. Normally, such support would be determined
by cairo's configure script. But since we don't run cairo's configure
script, we need to manually define HAVE_INTEL_ATOMIC_PRIMITIVES during
cairo's build. That macro enables codepaths that depend on certain
SIZEOF_* variables being defined by autoconf, so we also need to add the
necessary code in moz.build to set those.
