AudioProcessor: don't abort() on a non-frame-aligned Consume()

[OzGav]

Fixes #90.

AudioProcessor::Consume() asserts that the input length is a whole multiple
of the channel count:

assert(length % m_num_channels == 0);

When a caller passes a buffer that splits the audio mid-frame, this assertion aborts the entire host process instead of failing gracefully. It's easy to hit when feeding PCM in fixed-size byte chunks (the chunk boundary rarely lands
on a frame boundary), and especially likely with multi-channel sources where a frame spans more bytes. A library should never abort() the process that embeds it on account of input framing.

Reported downstream where an AcoustID scan of a 5.1 file took the whole application down with:

Assertion length % m_num_channels == 0 failed.

Consume() now tolerates input that isn't aligned to a whole number of frames. Any trailing partial frame is buffered in a new m_leftover member and prepended to the next call, so the audio is processed losslessly and channel
interleaving stays aligned across calls. The existing whole-frame processing loop is unchanged, it's simply moved into a private ConsumeAligned() helper that Consume() calls once the input has been split into a whole-frame portion
and a carried remainder.

Consume() splits input into complete frames + a carried remainder ConsumeAligned() the original Consume() body (still asserts alignment, which is now always satisfied) m_leftover holds a sub-frame remainder between calls; cleared in Reset() Behaviour for callers that already pass whole frames is identical as the leftover buffer stays empty, remainder is always 0, and the exact same processing path runs.

[coderabbitai]

Walkthrough

AudioProcessor gains a private ConsumeAligned() helper and an m_leftover buffer (std::vector<int16_t>). Consume() now completes any pending partial frame from previous calls before forwarding aligned input to ConsumeAligned(), storing any trailing remainder back into m_leftover. Reset() clears m_leftover.

Changes

Partial Frame Buffering in AudioProcessor

Layer / File(s)	Summary
Header contract: ConsumeAligned and m_leftover src/audio_processor.h	Declares private ConsumeAligned(const int16_t*, int) and adds m_leftover (std::vector<int16_t>) to buffer partial frames between calls.
Implementation: leftover logic in Consume() and Reset() src/audio_processor.cpp	Reset() clears m_leftover. Consume() drains any existing partial frame into ConsumeAligned(), processes the aligned bulk, then stashes the trailing remainder. Debug message corrected to say ConsumeAligned().

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	No description was provided by the author, making it impossible to assess whether it relates to the changeset.	Add a description explaining what the crash was, how the buffering mechanism fixes it, and why handling partial frames across calls is necessary.
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title accurately describes the core fix: handling non-frame-aligned Consume() calls without aborting, which matches the implementation changes of buffering incomplete samples in m_leftover.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[lalinsky]

Thank you!

[trostli]

Great! I've run into this error many times. @lalinsky would it be possible to issue a new release with this fix included?