[yash-fadadu-simform]

Description

Fixes #527

On iOS, stopping a recording while it is paused breaks the next recording. record() succeeds and the state is recording, but onAudioChunks never fires and the live waveform stays flat. The recording after that works again, so the failures alternate: works, flat, works, flat.

Cause: RecorderBytesStreamEngine keeps a paused flag that pause() and resume() toggle. stop() (detach()) never reset it. Stopping while paused left the flag true, so the next recording's input tap dropped every buffer. Pausing that silent recording toggled the flag back to false, which is why the one after it worked.

Fix: attach() now sets paused = false before installing the tap, so every new recording starts unpaused.

macOS is not affected: its recorder sets isPaused explicitly and clears it on start and stop.

Before / After

Repro harness in the example app on the iPhone 13 simulator (iOS 18.1), 16 kHz AAC, overrideAudioSession: true. Each cycle records for 1.5 s and counts onAudioChunks events; a cycle fails when none arrive.

Recordings that got audio, all three builds run back to back:

Scenario	main	New AVAudioEngine per recording, no paused reset	This PR
record → stop → record	10/10	10/10	10/10
record → pause → stop → record	3/6 (2nd, 4th, 6th silent)	3/6 (2nd, 4th, 6th silent)	6/6

The screenshot's AFTER panel was captured on an earlier build that also created a new AVAudioEngine per recording. The middle column shows that change alone leaves the bug in place; the paused reset in this PR is what fixes it.

Checklist

• The title of my PR starts with a Conventional Commit prefix (fix:, feat:, docs: etc).
• I have followed the Contributor Guide when preparing my PR.
• I have updated/added tests for ALL new/updated/fixed functionality.
• I have updated/added relevant documentation in docs and added dartdoc comments with ///.
• I have updated/added relevant examples in examples or docs.

Breaking Change?

• Yes, this PR is a breaking change.
• No, this PR is not a breaking change.

[vasu-nageshri-simform]

Verdict: Approve. The fix is correct and minimal. No blocking issues.

Comments on lines outside the diff:

[Nit] ios/Classes/RecorderBytesStreamEngine.swift:53 (detach()): also reset paused = false in detach(), so stop leaves the engine clean and attach() doesn't depend on it. Keeping the reset in attach() as well is fine.

[Nit] ios/Classes/RecorderBytesStreamEngine.swift:15: paused is written on the main thread and read on the audio render thread without synchronization. It works in practice for a Bool, but if the setter refactor below happens, consider an atomic or a serial queue.

Heads-up: this will need a rebase onto the new ios/audio_waveforms/Sources/audio_waveforms/ paths if the Swift Package Manager PR lands first.

[yash-fadadu-simform]

Addressed the review-summary nits:

• detach() now also resets paused = false. The reset in attach() is kept too.
• paused is now guarded by an NSLock, since it is written on the main thread and read on the render thread.