Pointer was already protected by a critical section in two places but
not the third. Added thread annotations to prevent this from happening
in the future.
BUG=None
Review-Url: https://codereview.webrtc.org/2726263004
Cr-Commit-Position: refs/heads/master@{#17017}
This isn't used any more so there's no point in maintaining it.
BUG=None
Review-Url: https://codereview.webrtc.org/2731673002
Cr-Commit-Position: refs/heads/master@{#17016}
Hopefully this will reduce the flakiness of PostDelayedTask.
BUG=none
Review-Url: https://codereview.webrtc.org/2728663008
Cr-Commit-Position: refs/heads/master@{#17001}
This adds a simple mechanism that provides protection against
implementations that use the legacy callback type in PlatformThread
and are prone to entering a busy loop.
Enabled only in DCHECK enabled builds.
BUG=webrtc:7187
Review-Url: https://codereview.webrtc.org/2720223003
Cr-Commit-Position: refs/heads/master@{#16973}
It's the faster, less strict cousin of checked_cast.
BUG=none
Review-Url: https://codereview.webrtc.org/2714063002
Cr-Commit-Position: refs/heads/master@{#16958}
Reason for revert:
breaks linux_ubsan bots.
Original issue's description:
> Reland of Use sched_yield on all POSIX platforms in PlatformThread. (patchset #1 id:1 of https://codereview.webrtc.org/2712133003/ )
>
> Reason for revert:
> Relanding - using sched_yield() in PlatformThread on all posix platforms.
>
> Original issue's description:
> > Revert of Use sched_yield on all POSIX platforms in PlatformThread. (patchset #1 id:1 of https://codereview.webrtc.org/2716683002/ )
> >
> > Reason for revert:
> > Reverting this change since it didn't affect the perf regressions we were seeing and actually seems to have caused more regressions as per comment in the bug.
> >
> > Original issue's description:
> > > Use sched_yield on all POSIX platforms in PlatformThread.
> > > (not only MacOS)
> > >
> > > This is a test to see if perf regressions we're seeing may be related to the use of nanosleep().
> > >
> > > BUG=695438
> > > TBR=solenberg@webrtc.org
> > >
> > > Review-Url: https://codereview.webrtc.org/2716683002 .
> > > Cr-Commit-Position: refs/heads/master@{#16807}
> > > Committed: 384498abb5
> >
> > TBR=solenberg@webrtc.org
> > # Skipping CQ checks because original CL landed less than 1 days ago.
> > NOPRESUBMIT=true
> > NOTREECHECKS=true
> > NOTRY=true
> > BUG=695438
> >
> > Review-Url: https://codereview.webrtc.org/2712133003
> > Cr-Commit-Position: refs/heads/master@{#16833}
> > Committed: 3ba1a8cd1b
>
> TBR=solenberg@webrtc.org
> # Not skipping CQ checks because original CL landed more than 1 days ago.
> BUG=695438
>
> Review-Url: https://codereview.webrtc.org/2725573002
> Cr-Commit-Position: refs/heads/master@{#16899}
> Committed: 4974df4183TBR=solenberg@webrtc.org,tommi@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=695438
Review-Url: https://codereview.webrtc.org/2721893002
Cr-Commit-Position: refs/heads/master@{#16903}
Reason for revert:
Relanding - using sched_yield() in PlatformThread on all posix platforms.
Original issue's description:
> Revert of Use sched_yield on all POSIX platforms in PlatformThread. (patchset #1 id:1 of https://codereview.webrtc.org/2716683002/ )
>
> Reason for revert:
> Reverting this change since it didn't affect the perf regressions we were seeing and actually seems to have caused more regressions as per comment in the bug.
>
> Original issue's description:
> > Use sched_yield on all POSIX platforms in PlatformThread.
> > (not only MacOS)
> >
> > This is a test to see if perf regressions we're seeing may be related to the use of nanosleep().
> >
> > BUG=695438
> > TBR=solenberg@webrtc.org
> >
> > Review-Url: https://codereview.webrtc.org/2716683002 .
> > Cr-Commit-Position: refs/heads/master@{#16807}
> > Committed: 384498abb5
>
> TBR=solenberg@webrtc.org
> # Skipping CQ checks because original CL landed less than 1 days ago.
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=695438
>
> Review-Url: https://codereview.webrtc.org/2712133003
> Cr-Commit-Position: refs/heads/master@{#16833}
> Committed: 3ba1a8cd1bTBR=solenberg@webrtc.org
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG=695438
Review-Url: https://codereview.webrtc.org/2725573002
Cr-Commit-Position: refs/heads/master@{#16899}
Use TaskQueue in IncomingVideoStream instead of the PlatformThread + event timer approach.
TBR=mflodman@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=webrtc:7219, webrtc:7253
Reland of
686aa37382 (revert)
e2d1d64295 (original)
Review-Url: https://codereview.webrtc.org/2720773002
Cr-Commit-Position: refs/heads/master@{#16872}
Reason for revert:
Reverting while fixing build issue in Chromium.
Original issue's description:
> Use TaskQueue in IncomingVideoStream instead of the PlatformThread + event timer approach.
>
> BUG=webrtc:7219, webrtc:7253
>
> Review-Url: https://codereview.webrtc.org/2716473002
> Cr-Commit-Position: refs/heads/master@{#16860}
> Committed: e2d1d64295TBR=mflodman@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=webrtc:7219, webrtc:7253
Review-Url: https://codereview.webrtc.org/2714393003
Cr-Commit-Position: refs/heads/master@{#16863}
In some rare circumstances, capturing a reference may be undesired. For
example, when creating an asynchronous task owned by the object itself,
the object may not need or want this task to keep itself alive.
BUG=None
Review-Url: https://codereview.webrtc.org/2711113008
Cr-Commit-Position: refs/heads/master@{#16840}
I'd like to make a change to rtc::Bind in another CL, and that will
be easier if there are fewer lines of code to modify.
BUG=None
Review-Url: https://codereview.webrtc.org/2719683002
Cr-Commit-Position: refs/heads/master@{#16838}
Reason for revert:
Reverting this change since it didn't affect the perf regressions we were seeing and actually seems to have caused more regressions as per comment in the bug.
Original issue's description:
> Use sched_yield on all POSIX platforms in PlatformThread.
> (not only MacOS)
>
> This is a test to see if perf regressions we're seeing may be related to the use of nanosleep().
>
> BUG=695438
> TBR=solenberg@webrtc.org
>
> Review-Url: https://codereview.webrtc.org/2716683002 .
> Cr-Commit-Position: refs/heads/master@{#16807}
> Committed: 384498abb5TBR=solenberg@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=695438
Review-Url: https://codereview.webrtc.org/2712133003
Cr-Commit-Position: refs/heads/master@{#16833}
The documentation for AsyncInvoker states that it owns the lifetime of
calls, and when its destructor is called, all in-flight calls are
cancelled or finish executing. The "cancelled" part is working, but if
a call is in the middle of executing, the destructor does *not* wait.
This is fixed by keeping a count of pending invocations, which is
decremented when a call is either cleared from a message queue or
finishes executing.
BUG=webrtc:3914, webrtc:3911
Review-Url: https://codereview.webrtc.org/2694723004
Cr-Commit-Position: refs/heads/master@{#16811}
The new constructor introduces two new changes:
* Support specifying thread priority at construction time.
- Moving forward, the SetPriority() method will be removed.
* New thread function type.
- The new type has 'void' as a return type and a polling loop
inside PlatformThread, is not used.
The old function type is still supported until all places have been moved over.
In this CL, the first steps towards deprecating the old mechanism are taken
by moving parts of the code that were simple to move, over to the new callback
type.
BUG=webrtc:7187
Review-Url: https://codereview.webrtc.org/2708723003
Cr-Commit-Position: refs/heads/master@{#16779}
These tests involve interactions with the file system, so to avoid
flakiness they shouldn't be run in parallel.
BUG=webrtc:7195
NOTRY=True
Review-Url: https://codereview.webrtc.org/2710433003
Cr-Commit-Position: refs/heads/master@{#16727}
gtest can print objects if they have an operator<< or a PrintTo
function in the same namespace as the object's class. Since
std::optional does not seem to have an operator<<, it'd be preferable
not to rely on rtc::Optional being printable through operator<<.
Currently, gtest errors will just dump the raw bytes of
rtc::Optionals, which make them really annoying to work with in tests.
BUG=webrtc:7196
Review-Url: https://codereview.webrtc.org/2704483002
Cr-Commit-Position: refs/heads/master@{#16717}
This is step 1 in the following process to move the task runner
abstraction over to Chrome, without gettings link errors on duplicate
symbols.
1. Move files from the rtc_base target to a new target
rtc_task_runner, and let rtc_base publicly depend on it.
2. In Chrome, add an explicit dependency on rtc_task_runner where it
depends on rtc_base.
3. Drop the webrtc dependency rtc_base --> rtc_task_runner.
4. Copy task runner code to Chrome (cl
https://codereview.chromium.org/2694903005/), and drop its
dependency on webrtc's rtc_task_runner target.
5. Delete the rtc_task_runner target and corresponding source files
from webrtc. Mission accomplished!
BUG=webrtc:6424
Review-Url: https://codereview.webrtc.org/2696703009
Cr-Commit-Position: refs/heads/master@{#16710}
This reduces binary size considerably and solves some other problems.
Also rewrote using variadic templates.
Initial patch contributed by andrey.semashev@gmail.com.
BUG=webrtc:2305
Review-Url: https://codereview.webrtc.org/2509733003
Cr-Commit-Position: refs/heads/master@{#16703}
The AsyncClosures only ever have one thing referencing them, so they
should be using std::unique_ptr to manage ownership. Maybe this code was
written before std::unique_ptr was available.
Originally reverted because it made a change to ScopedMessageData
that wasn't backwards compatible, and applications using the rtc::Thread
infrastructure may be using it.
BUG=None
NOTRY=True
Review-Url: https://codereview.webrtc.org/2689233003
Cr-Commit-Position: refs/heads/master@{#16684}
Reason for revert:
The change to messagequeue.h isn't backwards compatible. Will reland after making it backwards compatible.
Original issue's description:
> Use std::unique_ptr instead of rtc::scoped_refptr in AsyncInvoker.
>
> The AsyncClosures only ever have one thing referencing them, so they
> should be using std::unique_ptr to manage ownership. Maybe this code was
> written before std::unique_ptr was available.
>
> BUG=None
>
> Review-Url: https://codereview.webrtc.org/2689233003
> Cr-Commit-Position: refs/heads/master@{#16680}
> Committed: a5a472927bTBR=pthatcher@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=None
Review-Url: https://codereview.webrtc.org/2703613006
Cr-Commit-Position: refs/heads/master@{#16683}
The AsyncClosures only ever have one thing referencing them, so they
should be using std::unique_ptr to manage ownership. Maybe this code was
written before std::unique_ptr was available.
BUG=None
Review-Url: https://codereview.webrtc.org/2689233003
Cr-Commit-Position: refs/heads/master@{#16680}
This does not fix the myriad of other problems here, but at least
removes the dependency on CONF_VALUE.
BUG=526270
Review-Url: https://codereview.webrtc.org/2705603003
Cr-Commit-Position: refs/heads/master@{#16676}
Multimedia timers are higher precision than WM_TIMER, but they're also
a limited resource and more costly. So this implementation is a best
effort implementation that falls back on WM_TIMER when multimedia
timers aren't available.
A possible future change could be to make high precision timers in a
TaskQueue, optional. The reason for doing so would be for TaskQueues
that don't need high precision timers, won't eat up timers from TQ
instances that really need it.
BUG=webrtc:7151
Review-Url: https://codereview.webrtc.org/2691973002
Cr-Commit-Position: refs/heads/master@{#16661}
The files socketpool.h and diskcache.h also become unused, and are
deleted together with their sources.
BUG=webrtc:6424
Review-Url: https://codereview.webrtc.org/2694753002
Cr-Commit-Position: refs/heads/master@{#16601}
If android_setsocknetwork() is available, and it fails, then bind()
should *not* be called, and an error should be returned.
If it succeeds, then bind should be called, but with an "any" address.
This is to prevent cases where sockets are sent with a source address
that doesn't match the network interface they're sent on. See bug below.
This CL also changes "NetworkBinderResults" to an enum class, and
renames it to "NetworkBinderResult".
BUG=webrtc:7026
Review-Url: https://codereview.webrtc.org/2646863005
Cr-Commit-Position: refs/heads/master@{#16597}
