VPLAY-12738 : "com.comcast.viper" crash on VIPA enabled devices cont…

[vinodkadungoth]

…ributing BERR. (#1033)

• Crash in GetCurrentMimeType() on empty period Reason for Change: Potential floating pointer with empty periods Test Procedure: Tune regression testing
  Risk: Low

• LLAMA-18029 Release stream lock before calling Stop()

Reason for Change: Fix deadlock

Test Procedure: Stress test rewind to TSB start with increased progress
frequency

Risk: Low

• VPLAY-12067 - [Linear] Crash occurred in PreFetchThread in channel change (VPLAY-12067 - [Linear] Crash occurred in PreFetchThread in channel change #746)

• VPLAY-12067 - [Linear] Crash occurred in PreFetchThread while doing sequential channel change

Reason for change: In sequential channel change scenario, stop can be initiated
when AampLicensePreFetcher::PreFetchThread is in the middle of processing the
license request(ie in the middle of executing CreateDRMSession()). This
operation internally uses StreamAbstractionAamp object(mFetchInstance)
to call UpdateFailedDRMStatus. As PrivateInstanceAAMP::Stop progresses,
it can destroy the StreamAbstractionAamp object, but this status will reflect
in AampLicensePreFetcher at a later point when mDRMLicenseManager->Stop()
is called. Meanwhile, PreFetchThread is calling UpdateFailedDRMStatus
on the destroyed object and leading to crash.

Changes:

• Setting mFetchInstance in AampLicensePreFetcher to nullptr before destroying StreamAbstractionAamp object.

Test Procedure: Refer JIRA ticket

Risks: low

• Addit Fix for Unit Test compilation issue

• Implement a less intrusive protection against stream abstraction mutex deadlock

---

[Copilot]

Pull request overview

This PR addresses crash/deadlock scenarios in AAMP’s playback control and DASH MPD track handling, particularly around edge cases like empty periods and rewind-to-TSB-start workflows.

Changes:

• Update PlayFromTsbStart() to use a non-blocking stream lock attempt to avoid a potential deadlock during rewind-to-BOS handling.
• Clear representation in MPD track context when encountering an empty period to avoid dereferencing a stale pointer (fixing GetCurrentMimeType() crash path).

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
priv_aamp.cpp	Switches PlayFromTsbStart() from blocking lock acquisition to TryStreamLock() to reduce deadlock risk.
fragmentcollector_mpd.cpp	Nulls pMediaStreamContext->representation on empty periods to prevent stale-pointer deref in MIME type lookup.

[Copilot]

TryStreamLock() is a non-blocking try_lock() (no timeout), but the comment says it uses a timeout. More importantly, rate is set to AAMP_NORMAL_PLAY_RATE and NotifySpeedChanged(rate) is called even when the lock can't be acquired, so MonitorProgress() will stop retrying PlayFromTsbStart() while the underlying seek/tune never happened. Consider only updating rate/sending the speed-changed notification after a successful lock + TuneHelper(), or keep the rewind rate unchanged when the lock isn't acquired (and/or schedule a retry) to avoid leaving playback state inconsistent.

[pstroffolino]

This is the functional fix, yes?