VPAAMP-193 - iframe tracks can be selected as rampdown target in legacy and new abr

[benmesander]

FragmentfailureRampdown in abr.cpp and HybridABRManager.cpp don't filter out iframe tracks (unlike other abr-related methods). An iframe track bandwidth could be selected as the desired rampdown target, which doesn't correspond to any regular video profile. The caller then passes this bandwidth to GetProfileIndexForBandwidth which may not find a match. I think this currently works by accident for most Comcast live manifests because the iframe track bandwidths are the same as the track from which the iframes are derived.

[Copilot]

Pull request overview

Fixes ABR rampdown selection so iframe-only tracks aren’t considered as rampdown targets (preventing invalid bandwidths being passed to GetProfileIndexForBandwidth), and adds unit coverage for related ABR math edge-cases.

Changes:

• Filter iframe tracks out of FragmentfailureRampdown() candidate profiles in both abr/abr.cpp and support/aampabr/HybridABRManager.cpp.
• Fix integer-truncation in Hybrid ABR throughput calculations and add focused unit tests.
• Register HybridAbrTests with the unit test runner in CMake.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
abr/abr.cpp	Excludes iframe tracks from FragmentfailureRampdown() candidate list.
support/aampabr/HybridABRManager.cpp	Fixes throughput arithmetic and filters iframe tracks during fragment-failure rampdown.
test/utests/tests/AampAbrTests/AbrTests.cpp	Adds tests verifying ABRManager::FragmentfailureRampdown() skips iframe tracks and uses correct fallback.
test/utests/tests/AampAbrTests/HybridAbrTests.cpp	Adds tests for Hybrid ABR truncation fixes and iframe-skip behavior (plus local SpeedCache for testing).
test/utests/tests/AampAbrTests/CMakeLists.txt	Adds aamp_utest_run_add(HybridAbrTests) so the suite runs in CI.

Comments suppressed due to low confidence (1)

support/aampabr/HybridABRManager.cpp:458

• After filtering out iframe tracks, availableProfiles can become empty. The current implementation still computes len = size() - 1 and later indexes availableProfiles[len] / availableProfiles[0], which will go out-of-bounds when all profiles are iframe (or no profiles exist). Add an early return when availableProfiles.empty() (e.g., return 0 like ABRManager::FragmentfailureRampdown() does) before calculating len / sorting.

	std::vector<ProfileInfo> availableProfiles = getProfileInfo();
	availableProfiles.erase(
		std::remove_if(availableProfiles.begin(), availableProfiles.end(),
			[](const ProfileInfo &p) { return p.isIframeTrack; }),
		availableProfiles.end());
	int len = (int)availableProfiles.size() - 1;
	std::sort(availableProfiles.begin(), availableProfiles.end(), [](const ProfileInfo& a, const ProfileInfo& b) {
        return a.bandwidthBitsPerSecond < b.bandwidthBitsPerSecond;
    });
	// Iterate over profiles in descending order of bandwidth
	for (int i = (int)availableProfiles.size() -1  ;i >= 0 ; i--)
	{
		double profilePercentage = ((double)(availableProfiles[i].bandwidthBitsPerSecond) / availableProfiles[len].bandwidthBitsPerSecond) * 100.0;
		AAMPABRLOG_WARN("Index: %d, bandwidth %d , profile percentage %lf, buffer percentage %lf",i,(int)availableProfiles[i].bandwidthBitsPerSecond,profilePercentage,bufferPercentage);
		// Check if profile bandwidth percentage is less than buffer percentage ,and it should be a rampdown
		if (profilePercentage < bufferPercentage && (availableProfiles[i].bandwidthBitsPerSecond < currentbw))
		{
			desiredProfilebw  = availableProfiles[i].bandwidthBitsPerSecond;
			break;
		}
	}

	if (desiredProfilebw == 0)
	{
		// If no profile found, then return the lowest profile. Usually happens when bufferPercentage is very low or already at lowest profile
		desiredProfilebw = availableProfiles[0].bandwidthBitsPerSecond;
	}