[PR #7021] [MERGED] test(web): align manifest path errors with artifact ownership #7578

New issue

Closed

opened 2026-11-03 13:38:51 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-11-03 13:38:51 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/7021
Author: @gensokyo-dev
Created: 10/24/2026
Status: ✅ Merged
Merged: 10/24/2026
Merged by: @gensokyo-dev

Base: main ← Head: test/web-manifest-contract-20261024

📝 Commits (1)

cfc1eb8 test(web): align manifest path errors with artifact ownership

📊 Changes

1 file changed (+16 additions, -13 deletions)

View changed files

📝 tests/boundary/filesystem/webScriptManifestPaths.test.ts (+16 -13)

📄 Description

Problem and outcome

Main CI and subsequent Windows qualification fail two Web manifest path assertions. Main's 4f212c6a3 correctly distinguishes caller-selected input from source metadata derived from a selected manifest; these tests still expect the derived path refusal to blame script_index as invalid_request.

Assert the existing public artifact_operation_failed contract, exact operation and reason: path, refusal detail and unchanged source-path spelling. Reuse the fixture's path value in its POSIX identity assertion. Runtime and CLI/MCP behavior are unchanged; no adapter alias or special-case error mapping is added.

Executed validation

Fresh base 41cf869949, Linux x64, Node 24.22.1 / npm 11.16.0, two CPUs:

Baseline: two failed, seven passed in the affected file; only the stale error assertions fail (19.85 s wall / 26.74 s family CPU / 1125.1 MiB sampled family RSS, including build preparation).
Fixed related suite: 59 passed in six files, including real CLI and stdio MCP error provenance, caller-selected map failures, symlink/relocation identity, source integrity and read ownership (59.72 s / 74.44 s CPU / 1435.2 MiB).
Negative control: changing the real producer's post-selection ownership back to callerSelected=true makes both assertions fail with the incorrect invalid_request. Mutation restored; only this test file differs.
Final npm run check and git diff --check: passed (47.04 s / 85.93 s CPU / 2601.0 MiB).

The saved-capture CLI/MCP tests execute the real applications with authored fixture artifacts; this is not a new live-browser claim. Local platform substitution exercises Windows path admission, not native Windows I/O; the native Windows lane has now passed, including both formerly failing assertions: 777 curated tests passed, 48 platform skips, followed by the native and installed-package verifiers. This assertion-only change makes no analysis speedup or token-use claim; runtime payloads and analysis costs are unchanged. Timings include different test scopes and are not a before/after throughput comparison. RSS sampling sums descendants at 100 ms and may double-count shared pages or miss brief/reparented processes; it is not a hard aggregate cap. Tests had a 3 GiB sampled stop, static checks an 8 GiB stop, plus timeouts; no OOM occurred.

Codex and Greptile reviewed the exact cfc1eb8f head without actionable findings. Complete selected CI finished with 25,312 tests passed, five failed, 164 skipped. Only unchanged commandFinalizationCancellation cases fail after main's 731109b7d projection-owner refactor: their legacy export spy misses the prepared projection's lexical calls. Running those unchanged files locally also gives five failed, 15 passed; the producer and test are byte-identical to the base. This is the next isolated issue. This workflow is recorded as failure, not green qualification.

Author readiness

Human owner: @gensokyo-dev

The owner explicitly delegated readiness declarations and agent-managed review. Inspection and execution were performed by Codex; no separate human inspection is claimed.

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.

_{🔄 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/7021 **Author:** [@gensokyo-dev](https://github.com/gensokyo-dev) **Created:** 10/24/2026 **Status:** ✅ Merged **Merged:** 10/24/2026 **Merged by:** [@gensokyo-dev](https://github.com/gensokyo-dev) **Base:** `main` ← **Head:** `test/web-manifest-contract-20261024` --- ### 📝 Commits (1) - [`cfc1eb8`](https://github.com/morluto/rea/commit/cfc1eb8f3e4aeb7a3b39e62ddbfc2ce6a26b5281) test(web): align manifest path errors with artifact ownership ### 📊 Changes **1 file changed** (+16 additions, -13 deletions) <details> <summary>View changed files</summary> 📝 `tests/boundary/filesystem/webScriptManifestPaths.test.ts` (+16 -13) </details> ### 📄 Description ## Problem and outcome [Main CI](https://github.com/morluto/rea/actions/runs/40105159162/job/120906282001) and subsequent Windows qualification fail two Web manifest path assertions. Main's `4f212c6a3` correctly distinguishes caller-selected input from source metadata derived from a selected manifest; these tests still expect the derived path refusal to blame `script_index` as `invalid_request`. Assert the existing public `artifact_operation_failed` contract, exact operation and `reason: path`, refusal detail and unchanged source-path spelling. Reuse the fixture's path value in its POSIX identity assertion. Runtime and CLI/MCP behavior are unchanged; no adapter alias or special-case error mapping is added. ## Executed validation Fresh base `41cf869949`, Linux x64, Node 24.22.1 / npm 11.16.0, two CPUs: - Baseline: **two failed, seven passed** in the affected file; only the stale error assertions fail (19.85 s wall / 26.74 s family CPU / 1125.1 MiB sampled family RSS, including build preparation). - Fixed related suite: **59 passed in six files**, including real CLI and stdio MCP error provenance, caller-selected map failures, symlink/relocation identity, source integrity and read ownership (59.72 s / 74.44 s CPU / 1435.2 MiB). - Negative control: changing the real producer's post-selection ownership back to `callerSelected=true` makes both assertions fail with the incorrect `invalid_request`. Mutation restored; only this test file differs. - Final `npm run check` and `git diff --check`: passed (47.04 s / 85.93 s CPU / 2601.0 MiB). The saved-capture CLI/MCP tests execute the real applications with authored fixture artifacts; this is not a new live-browser claim. Local platform substitution exercises Windows path admission, not native Windows I/O; the [native Windows lane](https://github.com/morluto/rea/actions/runs/40109654433/job/120919026803) has now passed, including both formerly failing assertions: 777 curated tests passed, 48 platform skips, followed by the native and installed-package verifiers. This assertion-only change makes no analysis speedup or token-use claim; runtime payloads and analysis costs are unchanged. Timings include different test scopes and are not a before/after throughput comparison. RSS sampling sums descendants at 100 ms and may double-count shared pages or miss brief/reparented processes; it is not a hard aggregate cap. Tests had a 3 GiB sampled stop, static checks an 8 GiB stop, plus timeouts; no OOM occurred. Codex and Greptile reviewed the exact `cfc1eb8f` head without actionable findings. [Complete selected CI](https://github.com/morluto/rea/actions/runs/40109654433) finished with **25,312 tests passed, five failed, 164 skipped**. Only unchanged `commandFinalizationCancellation` cases fail after main's `731109b7d` projection-owner refactor: their legacy export spy misses the prepared projection's lexical calls. Running those unchanged files locally also gives **five failed, 15 passed**; the producer and test are byte-identical to the base. This is the next isolated issue. This workflow is recorded as **failure**, not green qualification. ## Author readiness Human owner: @gensokyo-dev The owner explicitly delegated readiness declarations and agent-managed review. Inspection and execution were performed by Codex; no separate human inspection is claimed. - [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. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-11-03 13:38:51 +00:00

closed this issue
added the
pull-request
label

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#7578

No description provided.

Rows
Columns