As documented in webrtc:11908 this cleanup is fairly invasive and
when a part of a frequently executed code path, can be quite costly
in terms of performance overhead. This is currently the case with
synchronous calls between threads (Thread) as well with our proxy
api classes.
With this CL, all code in WebRTC should now either be using MessageHandlerAutoCleanup
or calling MessageHandler(false) explicitly.
Next steps will be to update external code to either depend on the
AutoCleanup variant, or call MessageHandler(false).
Changing the proxy classes to use TaskQueue set of concepts instead of
MessageHandler. This avoids the perf overhead related to the cleanup
above as well as incompatibility with the thread policy checks in
Thread that some current external users of the proxies would otherwise
run into (if we were to use Thread::Send() for synchronous call).
Following this we'll move the cleanup step into the AutoCleanup class
and an RTC_DCHECK that all calls to the MessageHandler are setting
the flag to false, before eventually removing the flag and make
MessageHandler pure virtual.
Bug: webrtc:11908
Change-Id: Idf4ff9bcc8438cb8c583777e282005e0bc511c8f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/183442
Reviewed-by: Artem Titov <titovartem@webrtc.org>
Commit-Queue: Tommi <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32049}
This is part of a CL series merging rtc::MessageQueue into rtc::Thread.
Bug: webrtc:9883
Change-Id: I3cb857cc707d5e897759366d1478cc1ec19bce9a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/165344
Reviewed-by: Stefan Holmer <stefan@webrtc.org>
Commit-Queue: Sebastian Jansson <srte@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30180}
Previously, each RTCP packet was handled several times in a row, once
per m-section. This caused various weirdness and log warning spam, in
particular when using unified plan.
The cause was that the packets were wired trough each BaseChannel
instance up to the Call class. With this fix, the RTCP packets are wired
once per RtpTransportInternal via the common peer connection class.
Bug: chromium:1002875
Change-Id: I41c4eb3b68e215ebe0f2c6fb93ae0ee73335b89a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/152668
Commit-Queue: Sebastian Jansson <srte@webrtc.org>
Reviewed-by: Steve Anton <steveanton@webrtc.org>
Reviewed-by: Niels Moller <nisse@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29226}
MediaChannel accepted the RtpPacket buffers through non-const pointer.
This is both unclear and introduces questions regarding if the buffer is
actually copied or not.
This change modifies the method to accept by value to reduce ambiguity.
Usage of the non-const data() method which could potentially copy the
buffer contents is also reduced in favor of cdata() which never copies.
Bug: None
Change-Id: I3b2daef0d31cb6aacceb46c86da3a40ce836242b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/127340
Reviewed-by: Steve Anton <steveanton@webrtc.org>
Reviewed-by: Seth Hampson <shampson@webrtc.org>
Commit-Queue: Amit Hilbuch <amithi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#27090}
