As the exposure of power efficient stats to JavaScript are limited as
to reduce the fingerprinting surface to getStats, a new RTCStatsMember
derivation, RTCLimitedStatsMember, was added in this change. This sets
the exposure criteria of the stat on the type, which keeps the size of
the RTCStatsMember class the same and allows for extension in the future
for new types of stat restrictions.
Bug: webrtc:14483
Change-Id: Ib0303050a112441ba2416fd5f004dd8be26b47ca
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/279021
Commit-Queue: Evan Shrubsole <eshr@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38576}
These metrics were not only non-standard, but residing in the
non-standard "track" stats object that we want to delete. As per
https://github.com/w3c/webrtc-stats/issues/695#issuecomment-1259611462
these metrics are no longer needed because we already have
inbound-rtp.totalInterFrameDelay/totalSquaredInterFrameDelay which is
basically the same thing.
// mac_rel infra failures are unrelated
NOTRY=True
Bug: webrtc:14522
Change-Id: I565da42514a93f15532ba8357dd006547a5296ee
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/278100
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38509}
This information is now readily available. Let's expose it.
In practise we don't pace audio by default and the delay is ~0, however
we can tell that this metric is working as intended by setting
PacingController's pace_audio_ to true via the "WebRTC-Pacer-BlockAudio"
field trial. In this case chrome://webrtc-internals/ plots neats graphs
for audio send delay.
Bug: webrtc:10635
Change-Id: Iecfd93bb84ec61e5d54232769a9e7a500601b199
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/280523
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38483}
This metric was always supposed to be the spec's answer to
googBucketDelay, and is defined as "The total number of seconds that
packets have spent buffered locally before being transmitted onto the
network." But our implementation measured the time between capture and
send, including encode time. This is incorrect and yields a much larger
value than expected.
This CL updated the metric to do what the spec says. Implementation-wise
we measure the time between pushing and popping each packet from the
queue (in modules/pacing/prioritized_packet_queue.cc).
The spec says to increment the delay counter at the same time as we
increment the packet counter in order for the app to be able to do
"delta totalPacketSendDelay / delta packetSent". For this reason,
`total_packet_delay` is added to RtpPacketCounter. (Previously, the
two counters were incremented on different threads and observers.)
Running Google Meet on a good network, I could observe a 2-3 ms average
send delay per packet with this implementation compared to 20-30 ms
with the old implementation. See b/137014977#comment170 for comparison
with googBucketDelay which is a little bit different by design -
totalPacketSendDelay is clearly better than googBucketDelay.
Since none of this depend on the media kind, we can wire up this metric
for audio as well in a follow-up:
https://webrtc-review.googlesource.com/c/src/+/280523
Bug: webrtc:14593
Change-Id: If8fcd82fee74030d0923ee5df2c2aea2264600d4
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/280443
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38480}
This is exposing something that is already exposed in the legacy
getStats() API and is only available if the "video-timing" header
extension is used. Adding this metric here should unblock legacy
getStats() API deprecation. The follow-up to unship or standardize this
metric is tracked by https://crbug.com/webrtc/14586.
Bug: webrtc:14587
Change-Id: Ic3d45b0558d7caf4be2856a4cd95b88db312f85e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/279860
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38444}
This includes the stats dictionaries that have been made obsolete in
the spec and whose IDs are prefixed "DEPRECATED_":
- RTCMediaStreamTrackStats
- RTCMediaStreamStats
There is an ongoing experiment to unship these stats dictionaries in
Chrome (https://crbug.com/1374215). Marking then as [[deprecated]] helps
alert other dependencies that these classes are deprecated.
In the meantime, the "DEPRECATED_RTCMediaStreamTrackStats" prefix makes
it possible to use the deprecated classes.
# Unrelated infra failures
NOTRY=True
Bug: webrtc:14175, webrtc:14419
Change-Id: I498d370294058a628278e1e5b027cd12e24ad31a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/279700
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38439}
They may be non-standard, but they shouldn't be on a stats dictionary
that is deprecated (track is going away soon-ish). By moving them to
inbound-rtp they can continue to exist beyond track deprecation and
live in the right place in case we decide to standardize them later.
To help downstream projects transitions, the metrics are temporarily
available in both old and new locations. Delete of old location will
happen in a follow-up CL. TODOs added.
Bug: webrtc:14524
Change-Id: I2008060fa4ba76cde859d9144d2bb9648c7ff9af
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/278200
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38315}
These metrics were recently standardized. Part of the standardization
effort was to move them from obsolete "track" stats (on track for
deprecation and removal: https://crbug.com/webrtc/14175) into the
"inbound-rtp" stats which are not deprecated.
To ease transition for downstream projects, the metrics are temporarily
duplicated in both the old and new locations. In a follow-up CL, they
will be deleted from "track".
Bug: webrtc:14521
Change-Id: I0d9036472607a8c717ec823a458a79a49ddb80c7
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/278080
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38308}
According to a pprof, creating RTCCodecStats is one of the places where
we spend the most CPU time in the event of creating hundreds of them:
https://screenshot.googleplex.com/B6QNDvvoX8dK5vk
The lifetime was recently updated so that we no longer have to risk
creating hundreds of them, here is the relevant section:
https://w3c.github.io/webrtc-stats/#codec-dict*
This allows code simplifications and the deletion of
ProduceCodecStats_n since we can now do a lazy instantiation of codec
stats at the point of being referenced.
Bug: webrtc:14444
Change-Id: I342c5bfebe6a4be0359da3ea106692c7a217779e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/275763
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38209}
Unlike the cache of the entire stats report which is time limited, this
certificate cache is valid for an unlimited amount of time, but is
cleared at ClearCachedStatsReport() which is already called on each
SLD/SRD call. Since certificates can only change by negotiation, this
cache is ensured to always be invalidated when certificates change.
Since ClearCachedStatsReport() can happen for other reasons than
certificates changing we may clear the cache more often then is
necessary, but arguably this is seldom enough that we don't have to
create a separate "ClearCertificateStats()" method. Keep it simple?
The cache specifically avoids rtc::SSLCertChain::GetStats which
trigger rtc::SSLCertificate::GetStats and rtc::Base64::EncodeFromArray.
Bug: webrtc:14458
Change-Id: I5f95a4a5eb51cc4462147270fdae7bb9fb7bc822
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/276602
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38205}
In non-BUNDLE use cases, it is possible for multiple RTP streams to have
the same SSRC (as long as the SSRC is unique within the same transport).
This CL adds support for "outbound-rtp" and "inbound-rtp" stream stats
to have the same SSRC on different transports by adding the transport to
the stats ID. This avoids multiple RTP stream stats having the same
stats ID and fixes the problem. It's a stupid use case, but it should
work.
There could still be a stats ID collision in the event of multiple
"remote-inbound-rtp" or "remote-outbound-rtp" reference the same SSRC
but on separate transports for the same reason, and would require the
same fix... but one bug at a time. Not addressed in this CL.
Bug: webrtc:14443
Change-Id: I1a2ffd79fc67c2765e6dbd1ccc6828d4e91c4589
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/275769
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38201}
To properly handle SSRC collisions in non-BUNDLE we need to change how
RTP stats IDs are generated, but that is a riskier change to be dealt
with in a separate CL.
For now, we just make sure that crashing is not a possibility during
SSRC collisions as a mitigation for https://crbug.com/1361612. This is
achieved by adding a TryAddStats() method to RTCStatsReport returning
whether successful.
Bug: chromium:1361612
Change-Id: I8577ae4c84a7c1eb3c7527e9efd8d1b0254269a3
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/275766
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38197}
The spec says: "Represents the total number of connectivity check
requests sent (not including retransmissions)."
I was surprised to find candidate-pair.requestsSent wired up to
`sent_ping_requests_before_first_response`, which is the subset of
`sent_ping_requests_total` that happened when `recv_ping_responses`
was 0. This is not what the spec says.
By wiring it up to `sent_ping_requests_total` instead, the modern
getStats implementation of "requestsSent" will match the legacy
getStats implementation which is already wired up to this value.
// Unrelated bot issues
NOTRY=True
Bug: webrtc:14425
Change-Id: Ia53c9711ee7a13e596ae0eacf6066b97d9a1face
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/274174
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38025}
Ultimately, IDs should be random according to spec[1], so we shouldn't
rely on the ID to convey easily readable information. By making the IDs
shorter we reduce the overhead of string copies and make report dumps a
little bit smaller.
Drive-by: Add "DEPRECATED_" prefic to the RTCMediaStreamStats ID.
[1] https://w3c.github.io/webrtc-pc/#dom-rtcstats-id
# Examples of IDs before and after this CL #
RTCDataChannel_3
-> D3
RTCPeerConnection
-> P
RTCTransport_0_1
-> T01
RTCCodec_RTCTransport_0_1_100_minptime=10;useinbandfec=1
-> CIT01_100_minptime=10;useinbandfec=1
RTCInboundRTPAudioStream_6666
-> IA6666
RTCAudioSource_1
-> SA1
RTCOutboundRTPAudioStream_2943129392
-> OA2943129392
RTCRemoteInboundRtpAudioStream_3541280085
-> RIA3541280085
RTCIceCandidate_6cWRqicY
-> I6cWRqicY
RTCIceCandidatePair_6cWRqicY_haEcM2xD
-> CP6cWRqicY_haEcM2xD
RTCCertificate_FD1:BC:58:90:DF:E8:40:58:8D:04:91:44:93:4E:6C:52:9E:F0:14:98:AA:67:7B:8B:C8:30:C8:31:D0:84:1B:BF
-> CFD1:BC:58:90:DF:E8:40:58:8D:04:91:44:93:4E:6C:52:9E:F0:14:98:AA:67:7B:8B:C8:30:C8:31:D0:84:1B:BF
DEPRECATED_RTCMediaStreamTrack_receiver_3
-> DEPRECATED_TI3
RTCMediaStream_45a6e766-5d1a-40f9-a55c-ea8fdefcde49
-> DEPRECATED_S45a6e766-5d1a-40f9-a55c-ea8fdefcde49
Bug: webrtc:14416, webrtc:14419
Change-Id: I11f0a8b8354203fea1df1093d8864a6d47ee71e6
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/273709
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37992}
The code incorrectly assumed that codecs exist on a per-mid/transceiver
basis, but codec payload types are unique on a per-transport basis and
in practise most applications use BUNDLE (single transport for the
entire PC).
This CL makes the codecs per-transport instead of per-transceiver. We
still need to iterate transceivers because codecs are exposed on a
per-transceiver basis and as shown in
https://jsfiddle.net/henbos/7kqxgnr8/ it is possible for FMTP lines to
be different on different m= sections despite BUNDLE.
Manual testing shows that this CL brings down the number of "codec"
stats in Google Meet 50p from 872 objects to 43 objects.
Bug: webrtc:14414
Change-Id: Ic854b31bd595799554b99fff22cbd48264ebd141
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/273707
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Reviewed-by: Philipp Hancke <phancke@microsoft.com>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37989}
There's no way to add a deprecation warning unique to using
RTCMediaStreamTrackStats, but we could signal to users that it is
deprecated by adding "DEPRECATED_" to its ID.
This could break apps with hardcoded assumptions about what the stats
IDs are, but apps doing this are using the API incorrectly anyway, so
if anyone is affected by this change that would be a good time to
remove any dependency on this (see https://crbug.com/webrtc/10656
regading the fact that IDs should be unpredictable).
Bug: webrtc:14175
Change-Id: I6242c4efc08e9570420c00af5aaf491b1af819f1
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/269004
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37595}
increasing precision since summing up rounded values leads to
a rounding error, in particular for small frames which take very
little time to decode.
BUG=webrtc:12526,webrtc:13756
Change-Id: I647c702808856a002c746ed9f115aa9bcaddc1f3
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/262810
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Philipp Hancke <philipp.hancke@googlemail.com>
Cr-Commit-Position: refs/heads/main@{#37249}
This should allow standard stats users not to have to rely on the
obsolete "track" stats.
Bug: webrtc:14174
Change-Id: I24e5e1478ee47c73c12fcdecf7314f41fcc76bc7
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/266020
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37246}
implements a total frame assembly time statistic that measures the
cumulative time between the arrival of the first packet of a frame
(the lowest reception time) and the time all packets of the frame have
been received (i.e. the highest reception time)
This is similar to totalProcessingDelay
https://w3c.github.io/webrtc-stats/#dom-rtcinboundrtpstreamstats-totalprocessingdelay
in particular with respect to only being incremented for frames that are being decoded but does not include the amount of time spent decoding the frame.
This statistic is useful for evaluating mechanisms like NACK and FEC
and gives some insight into the behavior of the pacer sending the
packets.
Note that for frames with just a single packet the assembly time will be zero. In order to calculate an average assembly time an additional frames_assembled_from_multiple_packets counter for frames with more than a single packet is added.
Currently this is a nonstandard stat so will only show up in webrtc-internals and not in getStats. Formally it can be defined as
totalAssemblyTime of type double
Only exists for video. The sum of the time, in seconds, each video frame takes from the time the first RTP packet is received (reception timestamp) and to the time the last RTP packet of a frame is received.
Given the complexities involved, the time of arrival or the reception timestamp is measured as close to the network layer as possible.
This metric is not incremented for frames that are not decoded, i.e., framesDropped, partialFramesLost or frames that fail decoding for other reasons (if any). Only incremented for frames consisting of more than one RTP packet. The average frame assembly time can be calculated by dividing the totalAssemblyTime with framesAssembledFromMultiplePacket.
framesAssembledFromMultiplePacket of type unsigned long
Only exists for video. It represents the total number of frames correctly decoded for this RTP stream that consist of more than one RTP packet.
For such frames the totalAssemblyTime is incremented.
BUG=webrtc:13986
Change-Id: Ie0ae431d72a57a0001c3240daba8eda35955f04e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/260920
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36922}
Since the lifetime of an SctpDataChannel is not strictly controlled
by its controller, the controller might go away before the channel
does. This CL guards against this.
Bug: webrtc:13931
Change-Id: I07046fe896d1a66bf89287429beb0587382a13a9
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/261940
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Florent Castelli <orphis@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36852}
This better reflects the ownership passing of AddTrack, and is more
consistent for RemoveTrack.
Bug: webrtc:13980
Change-Id: Ide5baccf15fc687a4e092f8831ce8c0fea46604e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/259740
Reviewed-by: Niels Moller <nisse@webrtc.org>
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36603}
This cl/ extends the RTCIceCandidateStats object with
network_adapter_type and vpn, so that it maps the underlying
WebRTC objects completly.
Bug: webrtc:13773
Change-Id: I5cf79972c60ca6bf2a127dc96fa90811263ba6fd
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/253241
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36110}
This cl/ changes so that the RTCTransportStats bytes/packets
sent/recevied is computed in P2PTransportChannel. Previously
they were computed by aggregating over the Connections, but that
does not work when Connections are created and destroyed.
Bug: webrtc:13769
Change-Id: Ia97dfae70b5aced897d4813ec007ba61bc032f87
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/253100
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36103}
Also apply IWYU to all .cc files in pc/, and correct BUILD file to match.
Note: Some files came out wrong when iwyu was applied. These are not included.
Bug: none
Change-Id: Ib5ea46b8fcc505414d0447cca7218ad3afc2e321
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/252280
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36064}
This makes SetChannel() consistently make 2 invokes instead of a
multiple of senders+receivers (previous minimum was 4 but could be
larger).
* Stop() doesn't hop to the worker thread.
* SetMediaChannel(), an already-required step on the worker thread for
senders and *sometimes* for receivers[1], is now consistently required
for both. This simplifies transceiver teardown and enables the next
bullet.
* Transceiver stops all senders and receivers in one go rather than
ping ponging between threads.
[1] When not required, it was done implicitly inside of Stop().
See changes in `RtpTransceiver::SetChannel`
Bug: webrtc:13540
Change-Id: Ied61636c8ef09d782bf519524fff2a31e15219a8
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/249797
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Tomas Gunnarsson <tommi@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36057}
This is a straightforward replacement; simplifications due to the ability
to inline functions in the lambdas are for a later CL.
Bug: webrtc:11943
Change-Id: I7274cedde507b954f1d8aa8bc560861102eeb264
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/250540
Reviewed-by: Niels Moller <nisse@webrtc.org>
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35936}
Move deletion of channel objects over to the RtpTransceiver instead
of having it done by SdpOfferAnswer.
The deletion is now also done via PostTask rather than Invoke.
Bug: webrtc:11992, webrtc:13540
Change-Id: I5aff14956d5e572ca8816bbfef8739bb609b4484
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/248170
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35798}
