admin/webrtc_m130
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Ignore "b=AS:-1" instead of treating as a hard error.

Browse Source 
Follow up to https://codereview.webrtc.org/2989243002/.

It turns out that "b=AS:-1" was being used to mean "no bandwidth limit",
even though just omitting "b=AS" completely will do that. So we should
treat this as a soft error for now, and give applications time to
transition to doing the standard thing.

BUG=chromium:675361

Review-Url: https://codereview.webrtc.org/2995463002
Cr-Commit-Position: refs/heads/master@{#19244}
This commit is contained in:
deadbeef 2017-08-03 17:49:30 -07:00 committed by Commit Bot
parent d89b0bcc8a
commit 3e8016e1d5
2 changed files with 33 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
webrtc/pc/webrtcsdp.cc
View File

@ -2719,6 +2719,16 @@ bool ParseContent(const std::string& message,
if (!GetValueFromString(line, bandwidth, &b, error)) {
return false;
}
// TODO(deadbeef): Historically, applications may be setting a value
// of -1 to mean "unset any previously set bandwidth limit", even
// though ommitting the "b=AS" entirely will do just that. Once we've
// transitioned applications to doing the right thing, it would be
// better to treat this as a hard error instead of just ignoring it.
if (b == -1) {
LOG(LS_WARNING) << "Ignoring \"b=AS:-1\"; will be treated as \"no "
"bandwidth limit\".";
continue;
}
if (b < 0) {
return ParseFailed(line, "b=AS value can't be negative.", error);
}

23
webrtc/pc/webrtcsdp_unittest.cc
View File

@ -3358,6 +3358,29 @@ TEST_F(WebRtcSdpTest, DeserializingNegativeBandwidthLimitFails) {
ExpectParseFailure(std::string(kSdpWithNegativeBandwidth), "b=AS:-1000");
}
// An exception to the above rule: a value of -1 for b=AS should just be
// ignored, resulting in "kAutoBandwidth" in the deserialized object.
// Applications historically may be using "b=AS:-1" to mean "no bandwidth
// limit", but this is now what ommitting the attribute entirely will do, so
// ignoring it will have the intended effect.
TEST_F(WebRtcSdpTest, BandwidthLimitOfNegativeOneIgnored) {
static const char kSdpWithBandwidthOfNegativeOne[] =
"v=0\r\n"
"o=- 18446744069414584320 18446462598732840960 IN IP4 127.0.0.1\r\n"
"s=-\r\n"
"t=0 0\r\n"
"m=video 3457 RTP/SAVPF 120\r\n"
"b=AS:-1\r\n";
JsepSessionDescription jdesc_output(kDummyString);
EXPECT_TRUE(SdpDeserialize(kSdpWithBandwidthOfNegativeOne, &jdesc_output));
const ContentInfo* vc = GetFirstVideoContent(jdesc_output.description());
ASSERT_NE(nullptr, vc);
const VideoContentDescription* vcd =
static_cast<const VideoContentDescription*>(vc->description);
EXPECT_EQ(cricket::kAutoBandwidth, vcd->bandwidth());
}
// Test that "ufrag"/"pwd" in the candidate line itself are ignored, and only
// the "a=ice-ufrag"/"a=ice-pwd" attributes are used.
// Regression test for: