morluto/rea

Fork 0

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

[PR #2007] [MERGED] fix(target): probe WASM before inferred database routing #4962

New issue

Closed

opened 2026-10-31 00:57:27 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-31 00:57:27 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/2007
Author: @SII-YDD
Created: 10/12/2026
Status: ✅ Merged
Merged: 10/13/2026
Merged by: @morluto

Base: main ← Head: fix/wasm-before-inferred-database

📝 Commits (4)

4b59977 fix(target): probe WASM before inferred database routing
79a26b5 test(wasm): execute recovery arguments from native refusal
379c9f5 test(ci): repair fixture alignment and share MCP assertions
4429398 fix(target): reconcile WASM routing with current main

📊 Changes

8 files changed (+261 additions, -151 deletions)

View changed files

📝 docs/wasm-artifacts.md (+5 -0)
📝 src/application/BinaryTargetResolver.ts (+6 -1)
📝 tests/boundary/filesystem/targetDiagnostics.test.ts (+67 -2)
📝 tests/boundary/mcp/analysisViewMcp.test.ts (+4 -9)
📝 tests/boundary/mcp/binaryDiagnosticsMcp.test.ts (+15 -23)
📝 tests/boundary/mcp/evmInterfaceMcp.test.ts (+15 -21)
📝 tests/boundary/mcp/wasmArtifactMcp.test.ts (+97 -95)
➕ tests/fixtures/advertisedMcpTool.ts (+52 -0)

📄 Description

Summary

Follow up on #1872's remaining .hop admission review; refs #1852 (already closed). Recognized WASM v1 bytes renamed to .hop now receive the existing static-inspection refusal. The recovery test executes the actual tool name and JSON arguments printed in that refusal.

Problem and expected behavior

An eight-byte empty WASM header with an inferred .hop interpretation previously returned as a native database before content detection. It should receive wasm_requires_static_inspection. Explicit database/raw-binary/DOS COM choices and non-WASM .hop routing retain their meaning.

Change and scope

Reuse the existing detector before inferred database routing, skipping it for an explicit database choice. Non-WASM artifact dispatch still occurs after the database return. Identical valid ZIP bytes provide positive .zip and .hop/.HOP compatibility controls.

Reuse the existing owned application MCP harness and schema helper. Parameterize the native-refusal case for bin/hop; parse the printed recovery tool name and arguments, check the advertised schema and exact escaped path, then invoke that continuation. Preserve missing-WABT diagnostics, same-connection ping, unchanged source bytes and no native provider startup. Remove only the redundant canonical-input schema check.

Rebased onto 9cf49c7a784ace9b79e4b4c1022c9fbf2b3f77e1. The final diff also includes the test-only CI corrections described below. An independently prepared six-file Ghidra fixture correction was found identical to upstream 18d56f5 and dropped during rebase; it is not an additional contribution in this PR.

Contract and boundary impact

Semantic owner and earliest changed stage: filesystem admission in parseBinaryTarget.
Contribution scope fit and linked maintainer decision: ordinary correction within existing read-only admission semantics; no new execution capability, so advance scope decision is not applicable.
CLI and MCP/tool-catalog contract: existing refusal and inspect_wasm_artifact continuation now reach .hop WASM; schemas unchanged.
Provider, bridge, target-format, or platform compatibility: explicit interpretations and ordinary database routing preserved; no provider implementation change.
Evidence, artifact, provenance, or reconstruction contract: source bytes and existing identity/cleanup rules preserved.
Target selection, authority, upstream behavior and cleanup: no target execution, mutation, implicit provider fallback or installation; owned harness/fixtures reused.
Generated metadata, package or installation impact: none.

Expected effects: read existing files; no added target-process launch, injection, mutation, network access or protection recovery.

Evidence and regression coverage

Unchanged earlier admission producer: the selected WASM regression fails both filesystem/MCP .hop cases (2 failed / 33 passed).
Deliberately mutating the printed tool name to an unavailable tool makes both bin/hop continuation tests fail (2 failed / 7 passed). Production source restored afterward.
Final four-file focused run: 92 passed. Latest-base Linux run before the final continuation-test strengthening: 273 passed / 4 platform skips across 19 files, covering all 11 previously failing CI files, every shared Ghidra MCP harness consumer, public JavaScript CLI/MCP, reopen, Java-launch and WASM paths.
Final compiled stdio MCP smoke read actual advertised input/output schemas, parsed the refusal's exact arguments including a quoted path, invoked the static tool, observed independent missing-WABT capability_unavailable, then pinged the same connection. Source bytes were unchanged; client/transport/fixture cleanup completed.
Representative public result: native open reports target_unavailable / unsupported_target / wasm_requires_static_inspection; unconfigured static WABT lane reports capability_unavailable.
Observed, derived and inferred claims remain distinguishable.
Artifact identity, source provenance and failed attempts remain preserved.
Unsupported, incomplete, unavailable or uncertain outcomes remain visible.

Validation performed

Pinned Node 24.18.0 / npm 11.16.0; isolated dependency/workspace environments without GitHub credentials passed to project processes.

macOS: npm run test:focused -- tests/boundary/mcp/wasmArtifactMcp.test.ts tests/boundary/filesystem/targetDiagnostics.test.ts tests/boundary/filesystem/binaryTarget.test.ts tests/boundary/filesystem/rawBinaryTarget.test.ts — 92 passed after final test strengthening.
Linux arm64 official Node container with git/procps: 19-file npm run test:focused -- PATH... — 273 passed / 4 platform skips on latest base before final strengthening. Includes src/ghidra/{GhidraDoctor,GhidraInstallation,GhidraSnapshotCache}.test.ts, applicationWorkflowCli, filesystem binaryTarget/ghidraPspExtension/rawBinaryTarget/targetDiagnostics, MCP analysisView/ghidraCompactReferences/ghidraEvidence/ghidraFailureDiagnostics/ghidraInventory/ghidraNativeApi/ghidraReopen/ghidraRouting/inlineEvidenceComposition/wasmArtifact, and ghidraJavaLaunch tests.
npm run check, npm run check:fast, npm run docs:check — passed on Linux and repeated on the final macOS tree; architecture guards across 1,297 source files, README consistency 2 passed.
npm run verify:test-discovery — 1,036 files exactly once.
Linux npm run test:focused -- tests/boundary/mcp/wasmArtifactMcp.test.ts repeated after final strengthening — 9 passed.
git diff --check and independent read-only AI review — passed; review found no actionable issue.
Final compiled stdio MCP smoke — passed as described above.

CI repair included in this revision

The previous final-head run had two root failures; coverage/timing and required-CI failed as downstream aggregates.

The IDA large universal fixture declared alignment14 for offset4096. Its layout now allows an explicit alignment, with12 only for that entry and14 retained for all other entries. Original offsets, selected-slice lengths, byte comparisons, digests and cleanup assertions remain intact; no production parser or provider change.
Repeated schema and inline Evidence assertions are now shared by WASM, binary-layout, EVM and analysis-view tests through tests/fixtures/advertisedMcpTool.ts. The helper preserves actual advertised schemas, response success, canonical output parsing, Evidence confidence/digest/normalized result/session retention. Negative-input, malformed-provider, cancellation, retained-reference and cleanup scenarios remain in their existing workflows. Duplication threshold and exclusions are unchanged.

On main ded199a, duplication fell from1,481clones/15,816lines/3.010175% to1,473clones/15,744lines/2.996620%. All42 original tests across the five affected files pass on macOS and Linux, including all7IDA and14analysis-view cases. The exact npm run check:ci gate passed, including duplication and TODO scanning; check, check:fast, docs:check, discovery and compiled MCP smoke also passed. Latest-base post-rebase verification is recorded below.

An independently prepared Ghidra fixture correction was already fixed upstream18d56f5 and remains dropped, attributed to its upstream author.

macOS process-dependent CLI tests remain SDK-blocked: an independent Foundation control also reproduces duplicate SwiftBridging definitions. Host SDK was not modified; those paths were verified in the earlier Linux19-file run. No local Windows or real WABT/Hopper/Ghidra/IDA success run; simulated native bridge/Java evidence does not establish real-engine success. Fresh full upstream CI38367170302 completed successfully on head379c9f5. All six shards, static checks, coverage/timing and required CI passed; PR checks show20success/15conditional skips/0failures. The PR remains open and unmerged.

Compatibility, safety, and release

Breaking change or migration: none; recognized WASM follows its existing contract.
Real-provider/platform coverage: admission, public CLI/MCP and fixture paths as listed; real engine and macOS SDK gaps explicit above.
Package/release metadata: none.
Security/privacy/process containment: source bytes and resource ownership preserved; no new target execution.

Implementation, testing and independent review were assisted by OpenAI Codex.

Before merge

Concrete problem, focused outcome and Conventional Commit title.
Existing scope/authority preserved; related issue/review linked.
Observable behavior, recovery, failure, compatibility and cleanup covered.
Owning docs and representative output updated where applicable.
Final diff checked for secrets, unrelated cleanup and unsupported claims.
Latest full upstream CI completed successfully (run38367170302).

Latest-base verification (9cf49c7a784ace9b79e4b4c1022c9fbf2b3f77e1, head 379c9f54b8d38f0b33c3f24b8face33574996b2f):42 focused cases passed; check:ci, check, check:fast, docs:check all passed;1,038 test files discovered exactly once; compiled stdio MCP smoke passed. Exact duplicate percentage:2.990427% with1472clones/15726duplicated lines, below the unchanged3%threshold.

_{🔄 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/2007 **Author:** [@SII-YDD](https://github.com/SII-YDD) **Created:** 10/12/2026 **Status:** ✅ Merged **Merged:** 10/13/2026 **Merged by:** [@morluto](https://github.com/morluto) **Base:** `main` ← **Head:** `fix/wasm-before-inferred-database` --- ### 📝 Commits (4) - [`4b59977`](https://github.com/morluto/rea/commit/4b5997759da9f044c1432adb2ebcac358989d7a9) fix(target): probe WASM before inferred database routing - [`79a26b5`](https://github.com/morluto/rea/commit/79a26b5cef86933f4b4fdc5d8d0028073205cacc) test(wasm): execute recovery arguments from native refusal - [`379c9f5`](https://github.com/morluto/rea/commit/379c9f54b8d38f0b33c3f24b8face33574996b2f) test(ci): repair fixture alignment and share MCP assertions - [`4429398`](https://github.com/morluto/rea/commit/4429398955d4980ddaf3cb8687c774abe19fd06c) fix(target): reconcile WASM routing with current main ### 📊 Changes **8 files changed** (+261 additions, -151 deletions) <details> <summary>View changed files</summary> 📝 `docs/wasm-artifacts.md` (+5 -0) 📝 `src/application/BinaryTargetResolver.ts` (+6 -1) 📝 `tests/boundary/filesystem/targetDiagnostics.test.ts` (+67 -2) 📝 `tests/boundary/mcp/analysisViewMcp.test.ts` (+4 -9) 📝 `tests/boundary/mcp/binaryDiagnosticsMcp.test.ts` (+15 -23) 📝 `tests/boundary/mcp/evmInterfaceMcp.test.ts` (+15 -21) 📝 `tests/boundary/mcp/wasmArtifactMcp.test.ts` (+97 -95) ➕ `tests/fixtures/advertisedMcpTool.ts` (+52 -0) </details> ### 📄 Description ## Summary Follow up on #1872's remaining `.hop` admission review; refs #1852 (already closed). Recognized WASM v1 bytes renamed to `.hop` now receive the existing static-inspection refusal. The recovery test executes the actual tool name and JSON arguments printed in that refusal. ## Problem and expected behavior An eight-byte empty WASM header with an inferred `.hop` interpretation previously returned as a native database before content detection. It should receive `wasm_requires_static_inspection`. Explicit database/raw-binary/DOS COM choices and non-WASM `.hop` routing retain their meaning. ## Change and scope Reuse the existing detector before inferred database routing, skipping it for an explicit database choice. Non-WASM artifact dispatch still occurs after the database return. Identical valid ZIP bytes provide positive `.zip` and `.hop`/`.HOP` compatibility controls. Reuse the existing owned application MCP harness and schema helper. Parameterize the native-refusal case for `bin`/`hop`; parse the printed recovery tool name and arguments, check the advertised schema and exact escaped path, then invoke that continuation. Preserve missing-WABT diagnostics, same-connection ping, unchanged source bytes and no native provider startup. Remove only the redundant canonical-input schema check. Rebased onto `9cf49c7a784ace9b79e4b4c1022c9fbf2b3f77e1`. The final diff also includes the test-only CI corrections described below. An independently prepared six-file Ghidra fixture correction was found identical to upstream [18d56f5](https://github.com/morluto/rea/commit/18d56f5321c3c5d5be40c19bcb0a0f2b25c2d459) and dropped during rebase; it is not an additional contribution in this PR. ## Contract and boundary impact - Semantic owner and earliest changed stage: filesystem admission in `parseBinaryTarget`. - Contribution scope fit and linked maintainer decision: ordinary correction within existing read-only admission semantics; no new execution capability, so advance scope decision is not applicable. - CLI and MCP/tool-catalog contract: existing refusal and `inspect_wasm_artifact` continuation now reach `.hop` WASM; schemas unchanged. - Provider, bridge, target-format, or platform compatibility: explicit interpretations and ordinary database routing preserved; no provider implementation change. - Evidence, artifact, provenance, or reconstruction contract: source bytes and existing identity/cleanup rules preserved. - Target selection, authority, upstream behavior and cleanup: no target execution, mutation, implicit provider fallback or installation; owned harness/fixtures reused. - Generated metadata, package or installation impact: none. Expected effects: read existing files; no added target-process launch, injection, mutation, network access or protection recovery. ## Evidence and regression coverage - Unchanged earlier admission producer: the selected WASM regression fails both filesystem/MCP `.hop` cases (2 failed / 33 passed). - Deliberately mutating the printed tool name to an unavailable tool makes both `bin`/`hop` continuation tests fail (2 failed / 7 passed). Production source restored afterward. - Final four-file focused run: **92 passed**. Latest-base Linux run before the final continuation-test strengthening: **273 passed / 4 platform skips across 19 files**, covering all 11 previously failing CI files, every shared Ghidra MCP harness consumer, public JavaScript CLI/MCP, reopen, Java-launch and WASM paths. - Final compiled stdio MCP smoke read actual advertised input/output schemas, parsed the refusal's exact arguments including a quoted path, invoked the static tool, observed independent missing-WABT `capability_unavailable`, then pinged the same connection. Source bytes were unchanged; client/transport/fixture cleanup completed. - Representative public result: native open reports `target_unavailable` / `unsupported_target` / `wasm_requires_static_inspection`; unconfigured static WABT lane reports `capability_unavailable`. - [x] Observed, derived and inferred claims remain distinguishable. - [x] Artifact identity, source provenance and failed attempts remain preserved. - [x] Unsupported, incomplete, unavailable or uncertain outcomes remain visible. ## Validation performed Pinned Node 24.18.0 / npm 11.16.0; isolated dependency/workspace environments without GitHub credentials passed to project processes. - macOS: `npm run test:focused -- tests/boundary/mcp/wasmArtifactMcp.test.ts tests/boundary/filesystem/targetDiagnostics.test.ts tests/boundary/filesystem/binaryTarget.test.ts tests/boundary/filesystem/rawBinaryTarget.test.ts` — **92 passed** after final test strengthening. - Linux arm64 official Node container with git/procps: 19-file `npm run test:focused -- PATH...` — **273 passed / 4 platform skips** on latest base before final strengthening. Includes `src/ghidra/{GhidraDoctor,GhidraInstallation,GhidraSnapshotCache}.test.ts`, `applicationWorkflowCli`, filesystem `binaryTarget/ghidraPspExtension/rawBinaryTarget/targetDiagnostics`, MCP `analysisView/ghidraCompactReferences/ghidraEvidence/ghidraFailureDiagnostics/ghidraInventory/ghidraNativeApi/ghidraReopen/ghidraRouting/inlineEvidenceComposition/wasmArtifact`, and `ghidraJavaLaunch` tests. - `npm run check`, `npm run check:fast`, `npm run docs:check` — passed on Linux and repeated on the final macOS tree; architecture guards across **1,297** source files, README consistency **2 passed**. - `npm run verify:test-discovery` — **1,036 files exactly once**. - Linux `npm run test:focused -- tests/boundary/mcp/wasmArtifactMcp.test.ts` repeated after final strengthening — **9 passed**. - `git diff --check` and independent read-only AI review — passed; review found no actionable issue. - Final compiled stdio MCP smoke — passed as described above. ## CI repair included in this revision The [previous final-head run](https://github.com/morluto/rea/actions/runs/38363262390) had two root failures; coverage/timing and required-CI failed as downstream aggregates. - The IDA large universal fixture declared alignment14 for offset4096. Its layout now allows an explicit alignment, with12 only for that entry and14 retained for all other entries. Original offsets, selected-slice lengths, byte comparisons, digests and cleanup assertions remain intact; no production parser or provider change. - Repeated schema and inline Evidence assertions are now shared by WASM, binary-layout, EVM and analysis-view tests through `tests/fixtures/advertisedMcpTool.ts`. The helper preserves actual advertised schemas, response success, canonical output parsing, Evidence confidence/digest/normalized result/session retention. Negative-input, malformed-provider, cancellation, retained-reference and cleanup scenarios remain in their existing workflows. Duplication threshold and exclusions are unchanged. On main ded199a, duplication fell from1,481clones/15,816lines/3.010175% to1,473clones/15,744lines/2.996620%. All42 original tests across the five affected files pass on macOS and Linux, including all7IDA and14analysis-view cases. The exact `npm run check:ci` gate passed, including duplication and TODO scanning; `check`, `check:fast`, `docs:check`, discovery and compiled MCP smoke also passed. Latest-base post-rebase verification is recorded below. An independently prepared Ghidra fixture correction was already fixed upstream18d56f5 and remains dropped, attributed to its upstream author. macOS process-dependent CLI tests remain SDK-blocked: an independent Foundation control also reproduces duplicate SwiftBridging definitions. Host SDK was not modified; those paths were verified in the earlier Linux19-file run. No local Windows or real WABT/Hopper/Ghidra/IDA success run; simulated native bridge/Java evidence does not establish real-engine success. Fresh full upstream [CI38367170302](https://github.com/morluto/rea/actions/runs/38367170302) completed successfully on head379c9f5. All six shards, static checks, coverage/timing and required CI passed; PR checks show20success/15conditional skips/0failures. The PR remains open and unmerged. ## Compatibility, safety, and release - Breaking change or migration: none; recognized WASM follows its existing contract. - Real-provider/platform coverage: admission, public CLI/MCP and fixture paths as listed; real engine and macOS SDK gaps explicit above. - Package/release metadata: none. - Security/privacy/process containment: source bytes and resource ownership preserved; no new target execution. Implementation, testing and independent review were assisted by OpenAI Codex. ## Before merge - [x] Concrete problem, focused outcome and Conventional Commit title. - [x] Existing scope/authority preserved; related issue/review linked. - [x] Observable behavior, recovery, failure, compatibility and cleanup covered. - [x] Owning docs and representative output updated where applicable. - [x] Final diff checked for secrets, unrelated cleanup and unsupported claims. - [x] Latest full upstream CI completed successfully (run38367170302). Latest-base verification (`9cf49c7a784ace9b79e4b4c1022c9fbf2b3f77e1`, head `379c9f54b8d38f0b33c3f24b8face33574996b2f`):42 focused cases passed; `check:ci`, `check`, `check:fast`, `docs:check` all passed;1,038 test files discovered exactly once; compiled stdio MCP smoke passed. Exact duplicate percentage:2.990427% with1472clones/15726duplicated lines, below the unchanged3%threshold. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-31 00:57:27 +00:00

closed this issue
added the
pull-request
label

bsakaguchi referenced this issue

2026-11-01 21:42:22 +00:00

[PR #4962] [MERGED] perf(analysis-view): reuse validated immutable parents #6695

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#4962

No description provided.

Rows
Columns