morluto/rea

Fork 0

mirror of https://github.com/morluto/rea.git synced 2026-11-05 06:45:22 +00:00

[PR #6306] chore(experiments): prototype MCAP summary inspection #7272

New issue

Open

opened 2026-11-01 21:46:39 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-11-01 21:46:39 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/6306
Author: @diyasharma05
Created: 10/22/2026
Status: 🔄 Open

Base: main ← Head: feat/mcap-inspection-prototype

📝 Commits (10+)

6f21e34 chore(experiments): prototype MCAP summary inspection
4d058ae fix(experiments): bound MCAP records and preserve failure evidence
3d8c034 fix(ci): include MCAP experiment in Knip cache inputs
bc23221 fix(experiments): separate Node tests from Vitest ownership
7b48000 test(experiments): cover MCAP variable-field overruns
a9c39fe chore(experiments): merge main into MCAP prototype
4d1d075 test(experiments): bind MCAP mutation to final range
14f0c2d fix(experiments): retain MCAP failure evidence
42b538f fix(mcap): preserve partial file admission evidence
cf1672f fix(ci): retain MCAP sources in lint cache inputs

📊 Changes

24 files changed (+2862 additions, -20 deletions)

View changed files

➕ experiments/mcap-inspection/README.md (+199 -0)
➕ experiments/mcap-inspection/admission-fixture.ts (+59 -0)
➕ experiments/mcap-inspection/admission.node-test.ts (+205 -0)
➕ experiments/mcap-inspection/cli.node-test.ts (+102 -0)
➕ experiments/mcap-inspection/cli.ts (+195 -0)
➕ experiments/mcap-inspection/evidence.node-test.ts (+171 -0)
➕ experiments/mcap-inspection/file-state.ts (+126 -0)
➕ experiments/mcap-inspection/fixtures.ts (+36 -0)
➕ experiments/mcap-inspection/lifecycle.node-test.ts (+159 -0)
➕ experiments/mcap-inspection/limits.node-test.ts (+86 -0)
➕ experiments/mcap-inspection/limits.ts (+126 -0)
➕ experiments/mcap-inspection/output.node-test.ts (+104 -0)
➕ experiments/mcap-inspection/package-lock.json (+444 -0)
➕ experiments/mcap-inspection/package.json (+22 -0)
➕ experiments/mcap-inspection/partial-read.node-test.ts (+118 -0)
➕ experiments/mcap-inspection/records.node-test.ts (+99 -0)
➕ experiments/mcap-inspection/records.ts (+30 -0)
➕ experiments/mcap-inspection/request.node-test.ts (+199 -0)
➕ experiments/mcap-inspection/request.ts (+98 -0)
➕ experiments/mcap-inspection/summary.node-test.ts (+121 -0)

...and 4 more files

📄 Description

Summary

This standalone MCAP experiment reports channels, schemas, declared counts/times and chunk compression from a recording's summary. --schema ID returns original schema bytes as base64. It lives under experiments/mcap-inspection, outside REA's CLI/MCP catalog and published package.

Review fixes preserve caller arguments, consumed read ranges and hashes, independent final file observations, and now partial admission evidence. A rejected directory or failed open/stat retains its resolved path and facts already observed; identity mismatch retains both metadata snapshots. Successful receipts keep their existing shape. The lint cache explicitly includes the experiment's TypeScript files after merging with upstream.

Scope and boundaries

Private package with pinned @mcap/core; no REA runtime dependency added.
Reads a caller-selected regular local file. No payload decoding, decompression, robot connection, target execution or inspection-time network access.
Reads are capped at 4 MiB before allocation; JSON at 1 MiB. Overflow preserves file/read evidence and explicitly omits the report or oversized request, retaining request JSON size/hash. Unreportable requests fail before target access.
Counts are declarations; range hashes are not whole-file hashes. Unknown facts and unsupported summaries remain explicit.
Maintainer scope/dependency approval and retained-Evidence integration remain pending. This is an exploratory prototype, not authorization for integration.

Validation performed

For cf1672f907ef5cd3f67160927b680002e4417dfd, Windows x64, Node 24.22.1 / npm 11.16.0:

npm test --prefix experiments/mcap-inspection built successfully: 47 passed, 1 existing symlink privilege skip. Six admission CLI regressions failed before the fix and passed afterward. They cover directory/missing input, open/stat errors, real pathname replacement and combined admission/close failure; guards require the intended stages, no target reads, and actual descriptor closure. A near-budget case preserves admission and cleanup facts when request provenance is omitted.
Manual directory and missing-file invocations retain partial observations. The pinned public demo succeeds with the same successful receipt shape and unchanged input hash.
npm run check:fast passed typecheck and lint. npm run verify:test-discovery found all 2,483 root test files exactly once.
The focused lint-cache inventory assertion passed. A real three-way configuration merge and Turbo 2.11.7 probe confirmed the new input survives upstream's changes and source edits invalidate the cache. The full local static-check report test file had 7 passes and 1 existing Windows jscpd path-format failure; no full-file pass is claimed.

The README records wall time, CPU, peak memory and output size; concurrent validation prevents attributing timing differences to the fix. Cold-storage and token costs are unmeasured. Root CI does not run the standalone suite. Linux was not rerun for these fixes; the earlier WSL2 run covered a prior revision. Whole-file integrity, real robot recordings, macOS and large recordings remain unverified. The historical intermittent Windows metadata change remains unexplained; failures now retain the observations.

Author readiness

Human owner: @diyasharma05

I reviewed the final diff and relevant existing code, understand the change and its edge cases, and take responsibility for the complete PR, including AI-assisted work.
I checked the evidence above, confirmed it supports the current change, and recorded meaningful validation gaps.
I declare this PR ready for initial maintainer review.

Before merge

Material scope and reader-dependency questions are resolved.
The human owner has reviewed the change and declared it ready.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/morluto/rea/pull/6306 **Author:** [@diyasharma05](https://github.com/diyasharma05) **Created:** 10/22/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `feat/mcap-inspection-prototype` --- ### 📝 Commits (10+) - [`6f21e34`](https://github.com/morluto/rea/commit/6f21e34253719ee07d740fb407811583a243002f) chore(experiments): prototype MCAP summary inspection - [`4d058ae`](https://github.com/morluto/rea/commit/4d058aee7a99a0616cbf592afa936146407f1ea1) fix(experiments): bound MCAP records and preserve failure evidence - [`3d8c034`](https://github.com/morluto/rea/commit/3d8c034f5764115647cc9127d141a5c37c5343f8) fix(ci): include MCAP experiment in Knip cache inputs - [`bc23221`](https://github.com/morluto/rea/commit/bc23221d31c2a7c061d7898167fc47260cb5de98) fix(experiments): separate Node tests from Vitest ownership - [`7b48000`](https://github.com/morluto/rea/commit/7b48000dca4c4e559d72c0bd2ff072b1dfb10ddc) test(experiments): cover MCAP variable-field overruns - [`a9c39fe`](https://github.com/morluto/rea/commit/a9c39fe24628d704f4b134b546d27610c41a1b99) chore(experiments): merge main into MCAP prototype - [`4d1d075`](https://github.com/morluto/rea/commit/4d1d075a8d95364a4d0151f1fa8b239551b7ea50) test(experiments): bind MCAP mutation to final range - [`14f0c2d`](https://github.com/morluto/rea/commit/14f0c2d3043596a1d0de684f8749422586175a6e) fix(experiments): retain MCAP failure evidence - [`42b538f`](https://github.com/morluto/rea/commit/42b538f9ce873a8d66fb34aafe232bb9287c7d8f) fix(mcap): preserve partial file admission evidence - [`cf1672f`](https://github.com/morluto/rea/commit/cf1672f907ef5cd3f67160927b680002e4417dfd) fix(ci): retain MCAP sources in lint cache inputs ### 📊 Changes **24 files changed** (+2862 additions, -20 deletions) <details> <summary>View changed files</summary> ➕ `experiments/mcap-inspection/README.md` (+199 -0) ➕ `experiments/mcap-inspection/admission-fixture.ts` (+59 -0) ➕ `experiments/mcap-inspection/admission.node-test.ts` (+205 -0) ➕ `experiments/mcap-inspection/cli.node-test.ts` (+102 -0) ➕ `experiments/mcap-inspection/cli.ts` (+195 -0) ➕ `experiments/mcap-inspection/evidence.node-test.ts` (+171 -0) ➕ `experiments/mcap-inspection/file-state.ts` (+126 -0) ➕ `experiments/mcap-inspection/fixtures.ts` (+36 -0) ➕ `experiments/mcap-inspection/lifecycle.node-test.ts` (+159 -0) ➕ `experiments/mcap-inspection/limits.node-test.ts` (+86 -0) ➕ `experiments/mcap-inspection/limits.ts` (+126 -0) ➕ `experiments/mcap-inspection/output.node-test.ts` (+104 -0) ➕ `experiments/mcap-inspection/package-lock.json` (+444 -0) ➕ `experiments/mcap-inspection/package.json` (+22 -0) ➕ `experiments/mcap-inspection/partial-read.node-test.ts` (+118 -0) ➕ `experiments/mcap-inspection/records.node-test.ts` (+99 -0) ➕ `experiments/mcap-inspection/records.ts` (+30 -0) ➕ `experiments/mcap-inspection/request.node-test.ts` (+199 -0) ➕ `experiments/mcap-inspection/request.ts` (+98 -0) ➕ `experiments/mcap-inspection/summary.node-test.ts` (+121 -0) _...and 4 more files_ </details> ### 📄 Description ## Summary This standalone MCAP experiment reports channels, schemas, declared counts/times and chunk compression from a recording's summary. `--schema ID` returns original schema bytes as base64. It lives under `experiments/mcap-inspection`, outside REA's CLI/MCP catalog and published package. Review fixes preserve caller arguments, consumed read ranges and hashes, independent final file observations, and now partial admission evidence. A rejected directory or failed open/stat retains its resolved path and facts already observed; identity mismatch retains both metadata snapshots. Successful receipts keep their existing shape. The lint cache explicitly includes the experiment's TypeScript files after merging with upstream. ## Scope and boundaries - Private package with pinned `@mcap/core`; no REA runtime dependency added. - Reads a caller-selected regular local file. No payload decoding, decompression, robot connection, target execution or inspection-time network access. - Reads are capped at 4 MiB before allocation; JSON at 1 MiB. Overflow preserves file/read evidence and explicitly omits the report or oversized request, retaining request JSON size/hash. Unreportable requests fail before target access. - Counts are declarations; range hashes are not whole-file hashes. Unknown facts and unsupported summaries remain explicit. - Maintainer scope/dependency approval and retained-Evidence integration remain pending. This is an exploratory prototype, not authorization for integration. ## Validation performed For `cf1672f907ef5cd3f67160927b680002e4417dfd`, Windows x64, Node **24.22.1** / npm **11.16.0**: - `npm test --prefix experiments/mcap-inspection` built successfully: **47 passed, 1 existing symlink privilege skip**. Six admission CLI regressions failed before the fix and passed afterward. They cover directory/missing input, open/stat errors, real pathname replacement and combined admission/close failure; guards require the intended stages, no target reads, and actual descriptor closure. A near-budget case preserves admission and cleanup facts when request provenance is omitted. - Manual directory and missing-file invocations retain partial observations. The pinned public demo succeeds with the same successful receipt shape and unchanged input hash. - `npm run check:fast` passed typecheck and lint. `npm run verify:test-discovery` found all **2,483** root test files exactly once. - The focused lint-cache inventory assertion passed. A real three-way configuration merge and Turbo 2.11.7 probe confirmed the new input survives upstream's changes and source edits invalidate the cache. The full local static-check report test file had **7 passes and 1 existing Windows jscpd path-format failure**; no full-file pass is claimed. The [README](https://github.com/diyasharma05/rea/blob/cf1672f907ef5cd3f67160927b680002e4417dfd/experiments/mcap-inspection/README.md) records wall time, CPU, peak memory and output size; concurrent validation prevents attributing timing differences to the fix. Cold-storage and token costs are unmeasured. Root CI does not run the standalone suite. Linux was not rerun for these fixes; the earlier WSL2 run covered a prior revision. Whole-file integrity, real robot recordings, macOS and large recordings remain unverified. The historical intermittent Windows metadata change remains unexplained; failures now retain the observations. ## Author readiness Human owner: @diyasharma05 - [x] I reviewed the final diff and relevant existing code, understand the change and its edge cases, and take responsibility for the complete PR, including AI-assisted work. - [x] I checked the evidence above, confirmed it supports the current change, and recorded meaningful validation gaps. - [x] I declare this PR ready for initial maintainer review. ## Before merge - [ ] Material scope and reader-dependency questions are resolved. - [ ] The human owner has reviewed the change and declared it ready. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-11-01 21:46:39 +00:00

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

morluto/rea#7272

No description provided.

Rows
Columns