Change LogIncomingRtcpPacket and LogOutgoingRtcpPacket to take ArrayView<uint8_t>.
Split LogSessionAndReadBack into three functions and create class to share state between them.
Split VerifyRtpEvent into one incoming and one outgoing version.
Originally uploaded as https://codereview.webrtc.org/2997973002/
Bug: webrtc:8111
Change-Id: I22bdc35163bef60bc8293679226b19e41e8f49b3
Reviewed-on: https://webrtc-review.googlesource.com/5020
Reviewed-by: Fredrik Solenberg <solenberg@webrtc.org>
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Reviewed-by: Elad Alon <eladalon@webrtc.org>
Commit-Queue: Björn Terelius <terelius@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20063}
This is a reland of 57fb3154b5411934b80051ad827db4e54d00f381
Original change's description:
> Clean up libjingle API dependencies.
>
> This CL moves candidate.h into the public API, since it has
> been implicitly included before.
>
> This is a straightforward way of solving the circular
> dependencies involving that file. For instance,
> libjingle_peerconnection_api includes candidate.h from
> jsepicecandidate.h, but _api can't depend on rtc_p2p, which
> depends on _api. In fact, _api can't depend on much at all
> since it's a very high level abstraction; instead, things
> should depend on it.
>
> Furthermore, we have the case where deprecated headers
> include headers in internal modules. I just have to turn
> off include checking for those, but that's not a big deal.
>
> This CL punts the problem of callfactoryinterface.h being
> implicitly included, and pulling in most of the call
> module with it. This should be addressed in a follow-up
> CL.
>
> Bug: webrtc:7504
> Change-Id: I1b1729408158418333ccdf702bf529386090f0d7
> Reviewed-on: https://webrtc-review.googlesource.com/2020
> Commit-Queue: Patrik Höglund <phoglund@webrtc.org>
> Reviewed-by: Fredrik Solenberg <solenberg@webrtc.org>
> Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#20034}
Bug: webrtc:7504
Change-Id: I74aeeff678a4ce6482d2f402493ae13e698f1390
Reviewed-on: https://webrtc-review.googlesource.com/4703
Commit-Queue: Patrik Höglund <phoglund@webrtc.org>
Reviewed-by: Fredrik Solenberg <solenberg@webrtc.org>
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Reviewed-by: Patrik Höglund <phoglund@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20062}
WebRTC is currently using the SSL_CTX_set_verify callback. This
configures a callback for use with X509_STORE_CTX_set_verify_cb. See
https://www.openssl.org/docs/man1.0.2/crypto/X509_STORE_CTX_set_verify_cb.html
This callback does not override certificate verification. Rather, it
allows EACH failure in OpenSSL's built-in certificate verification, as
well as the final success, to be overridden (that's why there's an ok
parameter). It still runs the usual OpenSSL certificate verification
(which will never succeed).
The upshot is that the callback is called multiple times and
OpenSSLStreamAdapter does a ton of redundant work and checks the hash at
least twice, or more for certificates with other errors.
Instead, use SSL_CTX_set_cert_verify_callback. This short-circuits the
OpenSSL behavior entirely and uses a caller-supplied one.
https://commondatastorage.googleapis.com/chromium-boringssl-docs/ssl.h.html#SSL_CTX_set_cert_verify_callbackhttps://wiki.openssl.org/index.php/Manual:SSL_CTX_set_cert_verify_callback(3)
(This also removes the SSL_CTX_set_verify_depth call which is ignored
with SSL_CTX_set_cert_verify_callback. It didn't do anything before
either---it tells OpenSSL to reject chains that are too short, but the
rejection was overwritten by the callback anyway.)
(Later on, we'll need to switch this to the BoringSSL-only
SSL_CTX_set_custom_verify and CRYPTO_BUFFER APIs to fix WebRTC's
contribution to Chrome's binary size, but I've left that alone for the
time being.)
Bug: none
Change-Id: I9320a367d0961935836df63dc6f0868b069f0af0
Reviewed-on: https://webrtc-review.googlesource.com/4581
Commit-Queue: David Benjamin <davidben@webrtc.org>
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20053}
I420Buffer implementations (I420BufferImpl and WrappedNativeI420Buffer) rely on
the 'position' of the underlying ByteBuffers to indicate the start of Y, U, and
V channels. Returning slices prevents callers from altering the state of the
I420Buffer by changing the position.
ByteBuffers are especially prone to accidentally moving the position: relative
read operations (such as get()) increment the position by the size of data read.
BUG=webrtc:8303
Change-Id: I52edce8a3bf46a6c41980ff5110a9480f021f22f
Reviewed-on: https://webrtc-review.googlesource.com/4521
Commit-Queue: Bjorn Mellem <mellem@webrtc.org>
Reviewed-by: Sami Kalliomäki <sakal@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20050}
The TransportController was in p2p/base before and it cannot depend on
pc/ or media/ level targets because of the circular dependency. To make the
TransportController be responsible for creating and managing
the RtpTransport related objects which are pc/ level targets, the
TransportController is moved from p2p/base to pc/.
The TransportController makes more sense in pc/ anyway, since its main
responsibility is processing the "transport" parts of SDP which is
PeerConnection-specific.
This is also easier than moving RtpTransport related objects to p2p/base
because those objects also depend on other media/ and pc/ level targets
such as srtpfilter, cryptoparams etc.
Bug: webrtc:7013
Change-Id: Ic48dd5c454046ff3c81331f4b459f96a3255f328
Reviewed-on: https://webrtc-review.googlesource.com/4560
Commit-Queue: Zhi Huang <zhihuang@webrtc.org>
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20049}
We will have to convert from webrtc::SdpVideoFormat to
cricket::VideoCodec in a couple of places until
cricket::WebRtcVideoEncoderFactory is gone. It will be convenient to
have the conversion logic in a common place.
Bug: webrtc:7925
Change-Id: Ie5e88599f28aeea647e936300c04f9071daffd53
Reviewed-on: https://webrtc-review.googlesource.com/4840
Reviewed-by: Sami Kalliomäki <sakal@webrtc.org>
Commit-Queue: Magnus Jedvert <magjed@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20048}
The new class wraps any CoreFoundation Type ref and
automatically releases it when it goes out of scope.
Conceptually similar to std::unique_ptr.
Bug: webrtc:7825
Change-Id: Ie49572b9215fcb5b92b2c0c3e3d52b0b3cf01752
Reviewed-on: https://webrtc-review.googlesource.com/3380
Commit-Queue: Kári Helgason <kthelgason@webrtc.org>
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20045}
It is now used only by FileRotatingStream.
Bug: webrtc:6424
Change-Id: I216b20baadae836d24c39899efe4cb45c2935f41
Reviewed-on: https://webrtc-review.googlesource.com/4720
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Commit-Queue: Niels Moller <nisse@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20040}
We currently suppress warnings for bad constructs in media/. Still, the
warnings are causing problems when trying to include header files from
this directory. This CL cleans up some of the bad constructs.
Bug: None
Change-Id: I808ad39eb23870d20fa5bb05429b50c9078543ae
Reviewed-on: https://webrtc-review.googlesource.com/4541
Reviewed-by: Sami Kalliomäki <sakal@webrtc.org>
Commit-Queue: Magnus Jedvert <magjed@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20039}
Previously VideoFrame.Buffers would be converted to I420 if
apply_rotation() is true. With this change the operation is skipped if
the rotation is 0.
Bug: webrtc:7749
Change-Id: I24a1a8801e41d8f415b33fe57fec953b74df7459
Reviewed-on: https://webrtc-review.googlesource.com/4665
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Commit-Queue: Sami Kalliomäki <sakal@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20038}
All previous initialize methods are deprecated and a new initialize
that uses a builder pattern is added. This gives us full control over
the order of initialization.
Bug: webrtc:7474
Change-Id: I006190e50f2e75c5015f0be75b86d367676db2cc
Reviewed-on: https://webrtc-review.googlesource.com/4160
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Commit-Queue: Sami Kalliomäki <sakal@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20037}
This reverts commit 57fb3154b5411934b80051ad827db4e54d00f381.
Reason for revert: Breaks jingle_glue in chromium; need to leave candidate.h in place and include the new location until it's fixed.
Original change's description:
> Clean up libjingle API dependencies.
>
> This CL moves candidate.h into the public API, since it has
> been implicitly included before.
>
> This is a straightforward way of solving the circular
> dependencies involving that file. For instance,
> libjingle_peerconnection_api includes candidate.h from
> jsepicecandidate.h, but _api can't depend on rtc_p2p, which
> depends on _api. In fact, _api can't depend on much at all
> since it's a very high level abstraction; instead, things
> should depend on it.
>
> Furthermore, we have the case where deprecated headers
> include headers in internal modules. I just have to turn
> off include checking for those, but that's not a big deal.
>
> This CL punts the problem of callfactoryinterface.h being
> implicitly included, and pulling in most of the call
> module with it. This should be addressed in a follow-up
> CL.
>
> Bug: webrtc:7504
> Change-Id: I1b1729408158418333ccdf702bf529386090f0d7
> Reviewed-on: https://webrtc-review.googlesource.com/2020
> Commit-Queue: Patrik Höglund <phoglund@webrtc.org>
> Reviewed-by: Fredrik Solenberg <solenberg@webrtc.org>
> Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#20034}
TBR=phoglund@webrtc.org,deadbeef@webrtc.org,solenberg@webrtc.org,perkj@webrtc.org
Change-Id: Ic5c3d0cf0b8c4d48ecbc49efdb76b373e3c950a5
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: webrtc:7504
Reviewed-on: https://webrtc-review.googlesource.com/4702
Reviewed-by: Patrik Höglund <phoglund@webrtc.org>
Commit-Queue: Patrik Höglund <phoglund@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20036}
This reverts commit 3dc4d4a21f80cdf44c508412d784b254957696eb.
Reason for revert: breaks internal project
Original change's description:
> Move clients of WebRtcSession to use PeerConnection
>
> This change is part of the work to merge WebRtcSession into
> PeerConnection. To make that work easier, this moves all clients
> of WebRtcSession to use shims added to PeerConnection. That way
> when the classes are merged they won't need to be modified.
>
> Bug: webrtc:8183
> Change-Id: I43de7acf7e38c9fcf2dbf55d50eb05e73767c251
> Reviewed-on: https://webrtc-review.googlesource.com/4320
> Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
> Commit-Queue: Steve Anton <steveanton@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#20030}
TBR=steveanton@webrtc.org,deadbeef@webrtc.org
Change-Id: I13f335b24c26753429cd08a4ca3e295eed5660ff
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: webrtc:8183
Reviewed-on: https://webrtc-review.googlesource.com/4700
Reviewed-by: Alex Loiko <aleloi@webrtc.org>
Commit-Queue: Alex Loiko <aleloi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20035}
This CL moves candidate.h into the public API, since it has
been implicitly included before.
This is a straightforward way of solving the circular
dependencies involving that file. For instance,
libjingle_peerconnection_api includes candidate.h from
jsepicecandidate.h, but _api can't depend on rtc_p2p, which
depends on _api. In fact, _api can't depend on much at all
since it's a very high level abstraction; instead, things
should depend on it.
Furthermore, we have the case where deprecated headers
include headers in internal modules. I just have to turn
off include checking for those, but that's not a big deal.
This CL punts the problem of callfactoryinterface.h being
implicitly included, and pulling in most of the call
module with it. This should be addressed in a follow-up
CL.
Bug: webrtc:7504
Change-Id: I1b1729408158418333ccdf702bf529386090f0d7
Reviewed-on: https://webrtc-review.googlesource.com/2020
Commit-Queue: Patrik Höglund <phoglund@webrtc.org>
Reviewed-by: Fredrik Solenberg <solenberg@webrtc.org>
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20034}
Allow a custom version of audioproc_f in APM-QA.
Bug: webrtc:7494
Change-Id: Id9adffd63927202d868bc2fc8b6a54c8e6b07039
Reviewed-on: https://webrtc-review.googlesource.com/4060
Reviewed-by: Alex Loiko <aleloi@webrtc.org>
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20033}
This change is part of the work to merge WebRtcSession into
PeerConnection. To make that work easier, this moves all clients
of WebRtcSession to use shims added to PeerConnection. That way
when the classes are merged they won't need to be modified.
Bug: webrtc:8183
Change-Id: I43de7acf7e38c9fcf2dbf55d50eb05e73767c251
Reviewed-on: https://webrtc-review.googlesource.com/4320
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Commit-Queue: Steve Anton <steveanton@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20030}
Function pointer tables require relocations, so this goes into
.data.rel.ro, not .rodata, but this will at least mark the pages
read-only after relocations are resolved.
Bug: None
Change-Id: I8625e7466b2dcadafc4e4e5f9c6eccbd87af7109
Reviewed-on: https://webrtc-review.googlesource.com/4580
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Commit-Queue: Taylor Brandstetter <deadbeef@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20029}
SSL_CIPHER_standard_name is a bit easier to use. BoringSSL has the
strings in the library statically these days. (Turns out that's more
size-efficient than the code to build it up anyway!)
Bug: None
Change-Id: I91ffa725fa716791cdf75d944cf8d9a3e2cb9021
Reviewed-on: https://webrtc-review.googlesource.com/4362
Reviewed-by: Taylor Brandstetter <deadbeef@webrtc.org>
Commit-Queue: Taylor Brandstetter <deadbeef@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20028}
Stop using PayloadUnion's public member variables, since a future CL
will make them private.
BUG=webrtc:8159
Change-Id: Ia3dada56be7ef00ed80f3733209b18c178a36561
Reviewed-on: https://webrtc-review.googlesource.com/4380
Commit-Queue: Karl Wiberg <kwiberg@webrtc.org>
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20027}
I need to replace the audio part of PayloadUnion with SdpAudioFormat,
but that's a non-trivially-deletable class and those just don't work
well in unions, especially unions that don't have a discriminator that
says which member is currently active.
This CL converts the union to a class, adds a discriminator, and
provides accessor functions. CL #2 in the series will change all
outsiders to use the accessors instead of the public member variables
directly, and CL #3 will remove the public member variables. (It needs
to be done in separate steps like this because PayloadUnion is
unfortunately part of the API, and just changing it all in one go
would break users.)
BUG=webrtc:8159
Change-Id: I38c44bbb21a2d38600cff59bf37d8d47dfdbce21
Reviewed-on: https://webrtc-review.googlesource.com/4340
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20025}
Test enables single-nalu mode, sets limit for nalu lenght and verifies
that encoder follows that limit.
I found that QP jumps significantly when the mode is enabled. In result
encoder might produce 4kbyte and 0.4kbyte frames back-to-back. But it
seems that happens only to couple of frames in the beginning. This
caused test to fail with default RC thresholds. To bypass this I
increased frame size mismatch threshold from 20 to 30%. This should be
Ok considering single-nalu mode is rare.
BUG=webrtc:8070
Review-Url: https://codereview.webrtc.org/3014623002
Cr-Commit-Position: refs/heads/master@{#20023}
Also include a small fix for getting the encoder queue.
Bug: webrtc:7760
Change-Id: I96dc8ffb363b90382276d88148f81d5f89dca5f2
Reviewed-on: https://webrtc-review.googlesource.com/2683
Commit-Queue: Sami Kalliomäki <sakal@webrtc.org>
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20022}
If this is not done, the RTC_DCHECK_CALLED_SEQUENTIALLY might fire
if the encoder is used on a new VideoStreamEncoder. This happens
after VideoSendStream recreations due to changes in the SDP.
BUG=b/66590444
Change-Id: I086370526afbbe2ba629805f97f89e512ba3f472
Reviewed-on: https://webrtc-review.googlesource.com/4360
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Sami Kalliomäki <sakal@webrtc.org>
Reviewed-by: Magnus Jedvert <magjed@webrtc.org>
Commit-Queue: Rasmus Brandt <brandtr@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20020}
There is bug in AQ in v1.6.0 which causes raising QP to maximum value
and results in very poor quality of video no matter of allocated bitrate.
To trigger this someone needs to set packetization_mode=0 (or just do
not transmit this flag at all) in SDP. In this mode the encoder
enables multi-slice and disables AQ partially such that some part of
AQ algo still works and leads QP to maximum. The issue is fixed in v1.7.0.
BUG=webrtc:8070
Review-Url: https://codereview.webrtc.org/3011373002
Cr-Commit-Position: refs/heads/master@{#20015}
ReportBlock is the the real receiver report.
Triggering rtt update on ReportBlock support clients that send receiver
report blocks attached to SenderReport rather than ReceiverReport.
Bug: webrtc:7996
Change-Id: Ie826fa09fd1bf0e5256e995649f66811b5192761
Reviewed-on: https://webrtc-review.googlesource.com/4040
Reviewed-by: Åsa Persson <asapersson@webrtc.org>
Commit-Queue: Danil Chapovalov <danilchap@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#20014}
