refactor(dgw): parse growing WebM with TagDecoder

[irvingouj@Devolutions (irvingoujAtDevolution)]

The original streamer treated incomplete tags as UnexpectedEOF and rewound the file.
ebml-iterable 0.7 keeps those tags pending, so live recordings can append without seek/skip.

Co-authored-by: Copilot 223556219+Copilot@users.noreply.github.com

---

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

[irvingouj@Devolutions (irvingoujAtDevolution)]

Implementation notes:

• webm_stream now feeds a TagDecoder instead of WebmIterator.
• Incomplete tags stay in the buffer; EOF waits reopen the file at the decoder read head instead of UnexpectedEOF + seek/skip.
• Start/Pull protocol is unchanged.
• Unit test: incomplete_input_waits_then_resumes_after_refresh.

Note

LLM-assisted content (no human feedback).

[Copilot]

Pull request overview

Refactors growing WebM parsing to use TagDecoder, retaining incomplete tags until recordings append more data.

Changes:

• Adds incremental buffered decoding and disk refresh support.
• Updates streaming and benchmark paths.
• Adds a growing-file regression test.

Two error-handling regressions were identified.

Reviewed changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
crates/video-streamer/src/streamer/mod.rs	Integrates incremental decoding.
crates/video-streamer/src/streamer/iter.rs	Implements buffered TagDecoder iteration.
crates/video-streamer/src/bench_support.rs	Updates benchmark iteration.
crates/video-streamer/Cargo.toml	Adds bytes.
Cargo.lock	Locks the dependency update.

Suppressed comments (1)

crates/video-streamer/src/streamer/mod.rs:147

• refresh_from_disk() only repositions the reader; it does not consume or reset the decoder's buffered bytes. Consequently, an InvalidTagData error is returned from the same bytes after every refresh, turning a permanent parse failure into repeated waits for up to 75 seconds. With TagDecoder, incomplete input is represented by None, so parser errors should fall through to the existing generic error arm.

                        webm_itr.refresh_from_disk()?;

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[irvingouj@Devolutions (irvingoujAtDevolution)]

Review follow-up:

• Dropped unused cluster_start_position after synthetic Cluster End went away.
• Skip-to-end and live wait now treat incomplete input as None.
• InvalidTagData / UnexpectedEOF are no longer retried; decode errors are fatal. IOError is the only dedicated IO arm.

Note

LLM-assisted content (no human feedback).