morluto/rea

Fork 0

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

[PR #7222] [MERGED] fix(javascript): honor effective symbolic descriptor values #7726

New issue

Closed

opened 2026-11-03 13:39:55 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-11-03 13:39:55 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/7222
Author: @SII-YDD
Created: 10/26/2026
Status: ✅ Merged
Merged: 10/26/2026
Merged by: @gensokyo-dev

Base: main ← Head: fix/symbolic-descriptor-values

📝 Commits (10+)

0080007 fix(javascript): honor effective symbolic descriptor values
7aba953 fix(javascript): preserve empty descriptor spreads and CI path identity
32c3715 test(pwntools): retain completed evidence in cleanup regressions
6cc5469 chore: synchronize upstream and retain its Wakaru path contract
e857317 test: preserve native paths and workspace admission distinctions
9a786e3 chore: reconcile descriptor fix with current main
61fe689 fix(javascript): invalidate replaced symbolic descriptor exports
23f468a chore: refresh descriptor validation on current main
bb88f0c fix(javascript): classify shared descriptor facts before exporting
4a40b54 test(javascript): verify string spread RPC fields against runtime

📊 Changes

12 files changed (+1109 additions, -94 deletions)

View changed files

📝 src/domain/javascript/electronStaticAnalysisBrowser.ts (+7 -7)
📝 src/domain/javascript/javascriptAstValues.test.ts (+145 -2)
📝 src/domain/javascript/javascriptAstValues.ts (+211 -30)
📝 src/domain/javascript/javascriptRpcEndpoints.test.ts (+24 -5)
📝 src/domain/javascript/javascriptSemanticAmbientWrites.test.ts (+98 -1)
📝 src/domain/javascript/javascriptSemanticAmbientWrites.ts (+132 -11)
📝 src/domain/javascript/javascriptSemanticChildProcesses.ts (+2 -2)
📝 src/domain/javascript/javascriptSymbolicBindings.test.ts (+178 -0)
📝 src/domain/javascript/javascriptSymbolicBindings.ts (+51 -30)
📝 src/domain/webBundleAnalyzerAst.ts (+3 -2)
📝 src/domain/webBundleAnalyzerInspection.ts (+3 -3)
📝 tests/boundary/cli/electronWrappedCallPaths.test.ts (+255 -1)

📄 Description

Summary

Fixes #7219. Ordered Object.defineProperty descriptor values could leave stale bundled exports, producing false IPC senders and missing values written last. Shared CLI/MCP analysis now uses effective ordered object facts and records unknown writes instead of reusing stale exports.

The object-member owner parses keys once and supplies related value/get/set queries from one snapshot. Literal spreads, array holes/unknown iterable suffixes, symbolic primitive indices and null/ignored/unknown prototypes retain their distinctions without per-index expansion. One normalized data-value field preserves original ordinary method nodes across symbolic, child-process and browser consumers; accessors stay unresolved. Duplicate prototype interpretation and stale fallback code are removed.

Native descriptor semantics require a retained, read-specific global identity proof. Exact replacements accept canonical direct/global-object spellings only while namespace identity is preserved. Conditional, aliased, parent, mutating or possibly repeated effects remain unknown; only a proven inert empty or exact primitive-literal-return override preserves the previous export. Competing unconditional replacements use the existing definite-order fact for both selection and confirmation; effectful parameter binding stays unknown. Matching ambient facts are cached within the reader. No invocation-count heuristic or expanded call graph was added.

Validation base 5508b2aff is integrated; final head e7e8d7bb645b67cb117c6257601a67198503af1e. Twelve related files change; unrelated conflict resolutions retain main unchanged.

Validation and limits

384 tests pass: 358 symbolic/AST/RPC/shape/ambient/child-process/browser tests and 26 public CLI/MCP acceptance/boundary tests. check:fast, docs:check and diff checks pass.
Original and review regressions were reproduced before their producing-boundary fixes. Actual Node confirms that the second invocation can read a replacement installed at the end of the first; unknown is required without a single-invocation proof. The directly affected RPC golden uses an independent runtime oracle.
Node 24.22.1 executes 35 test-owned CommonJS/TypeScript bundles (31060 bytes, 5308 AST nodes), with native type erasure and a recording Electron substitute: 26 channels execute, two mixed descriptors throw TypeError. Main has five false channels and misses six runtime channels. Fixed CLI/MCP produce 25 correct channels and explicitly leave the possibly repeated patch-after-sender factory unknown. This is not native Electron or third-party application verification. That deliberate static coverage gap remains; it is not counted as complete recovery.
Summary plus 3 javascript-facts pages projects four fields from 150 observations, with exhausted paging. Every row matches the full parent’s node/observation identity, source metadata, evidence, values and field omission. Both completed delivery workflows export the exact same full parent record (ev_79710da7151e46c6115bd2238cb7f0f578934b2490197004391796cda2d0f76f), root digest 239c4dd8b130e1d7be4682626f6b0464cc7eb90a5da03ec58afedad46f0ff660. No tool catalog is fetched.
An expanded complete reply correctly exceeds the existing 10 MiB transport budget. Recovery uses retained Evidence and paging without reanalysis; no limit is raised and complete Evidence remains accessible.

Shipped application control

Official Obsidian 1.14.4 ASAR matches the pinned corpus digest 146ef8470d8cbb4a6db209bdc19bf2bad13350444186c0fd93c31087b142b92f (27477966 bytes). Select exactly main.js and enhance.js: 216931 source bytes, 32508 AST nodes, 47 IPC channels and 47 main handlers. Base and fixed complete Evidence records are byte-for-byte structurally equal; MCP exports that same complete parent. The application was not executed. Coverage stays partial: 36 unselected files and bounded alias-selection unknowns remain.

Single completed-task controls with a 512 MiB worker and the same two-CPU guard: base CLI 18.552s wall / 24.736s CPU / 846.98 MiB family peak; fixed CLI 16.877 / 23.416 / 816.02; fixed MCP 18.234 / 24.418 / 1107.77. These samples do not establish a speed gain. MCP uses 3 calls (summary, 3 sample rows, full export), 56054 reply bytes and no catalog. Its sample honestly covers 3/94 rows with a retained next offset; it is not exhaustive. Complete 28.4 MB Evidence stays on disk. Source selection, identities, provenance and omitted fields are verified.

Resource evidence

Two-CPU affinity, bounded worker heap, fresh analysis processes, warm runtime builds, uncontrolled OS caches; single completed-task samples do not establish a speed difference.

Workflow	Tool calls	Reply bytes	Wall / CPU seconds	Sampled family peak MiB
Main CLI + capture	—	—	9.296 / 11.327	568.23
Fixed CLI + capture	—	—	9.152 / 11.073	559.04
Summary + 3 pages + full export	5	571742	9.764 / 11.603	619.10
Complete request + retained recovery + full export	5	549034	10.620 / 11.714	628.12

CPU/memory/wall include initialization, extraction and export; reply bytes exclude requests/initialization. RSS sampling can double-count shared pages and miss short peaks. Agent/schema tokens were not measured. Source bytes remain unchanged; complete controls and resource reports are retained locally.

Final head CI 40381159489 passes source/static/docs/build, all eight end-to-end shards, real JavaScript recovery, Windows controls and the Linux native package baseline. Codex reports no major issues; Greptile passes; all review threads are resolved. Earlier canceled runs are not counted as passes. The unchecked human readiness declaration is left unchanged and will use the explicitly authorized bypass; it does not assert human review.

Author readiness

Human owner: @SII-YDD

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

This PR addresses a concrete problem or an agreed enhancement.
Material scope and authority questions are resolved, and any required maintainer scope decision covers the final diff.
The PR has one focused outcome and the title follows type(scope): outcome.
Related issue is linked, or the reason for not linking one is stated above.
Tests cover changed observable behavior and meaningful failure paths.
Owning docs, contracts, and generated metadata are updated where needed.
User-visible CLI/MCP changes include representative output.
I checked the final diff for secrets, unrelated cleanup, and unsupported claims.

_{🔄 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/7222 **Author:** [@SII-YDD](https://github.com/SII-YDD) **Created:** 10/26/2026 **Status:** ✅ Merged **Merged:** 10/26/2026 **Merged by:** [@gensokyo-dev](https://github.com/gensokyo-dev) **Base:** `main` ← **Head:** `fix/symbolic-descriptor-values` --- ### 📝 Commits (10+) - [`0080007`](https://github.com/morluto/rea/commit/008000776a64dbc7a476cfb890abaa786f68069a) fix(javascript): honor effective symbolic descriptor values - [`7aba953`](https://github.com/morluto/rea/commit/7aba953300ed3a4a55ad8b0a8f69b01930f0ec5e) fix(javascript): preserve empty descriptor spreads and CI path identity - [`32c3715`](https://github.com/morluto/rea/commit/32c37153a71d8d9adb679dd486745d0b724a687f) test(pwntools): retain completed evidence in cleanup regressions - [`6cc5469`](https://github.com/morluto/rea/commit/6cc54690a3ad246f4a1f55ad53fc9016969627c4) chore: synchronize upstream and retain its Wakaru path contract - [`e857317`](https://github.com/morluto/rea/commit/e8573176676ecd59c2a6c3386e7dd48dd706bb09) test: preserve native paths and workspace admission distinctions - [`9a786e3`](https://github.com/morluto/rea/commit/9a786e393c24f7f9877cb192614e1ab89ad0137b) chore: reconcile descriptor fix with current main - [`61fe689`](https://github.com/morluto/rea/commit/61fe6892c74bf2d6527e5589153172df93ae970e) fix(javascript): invalidate replaced symbolic descriptor exports - [`23f468a`](https://github.com/morluto/rea/commit/23f468a1878deb6d1f0b0614a3f9445a7af89a16) chore: refresh descriptor validation on current main - [`bb88f0c`](https://github.com/morluto/rea/commit/bb88f0cd59d168c69248a51332d8e69c331ac1c6) fix(javascript): classify shared descriptor facts before exporting - [`4a40b54`](https://github.com/morluto/rea/commit/4a40b541a01c8cc3707cb3e2e3da497e879fc784) test(javascript): verify string spread RPC fields against runtime ### 📊 Changes **12 files changed** (+1109 additions, -94 deletions) <details> <summary>View changed files</summary> 📝 `src/domain/javascript/electronStaticAnalysisBrowser.ts` (+7 -7) 📝 `src/domain/javascript/javascriptAstValues.test.ts` (+145 -2) 📝 `src/domain/javascript/javascriptAstValues.ts` (+211 -30) 📝 `src/domain/javascript/javascriptRpcEndpoints.test.ts` (+24 -5) 📝 `src/domain/javascript/javascriptSemanticAmbientWrites.test.ts` (+98 -1) 📝 `src/domain/javascript/javascriptSemanticAmbientWrites.ts` (+132 -11) 📝 `src/domain/javascript/javascriptSemanticChildProcesses.ts` (+2 -2) 📝 `src/domain/javascript/javascriptSymbolicBindings.test.ts` (+178 -0) 📝 `src/domain/javascript/javascriptSymbolicBindings.ts` (+51 -30) 📝 `src/domain/webBundleAnalyzerAst.ts` (+3 -2) 📝 `src/domain/webBundleAnalyzerInspection.ts` (+3 -3) 📝 `tests/boundary/cli/electronWrappedCallPaths.test.ts` (+255 -1) </details> ### 📄 Description ## Summary Fixes #7219. Ordered `Object.defineProperty` descriptor values could leave stale bundled exports, producing false IPC senders and missing values written last. Shared CLI/MCP analysis now uses effective ordered object facts and records unknown writes instead of reusing stale exports. The object-member owner parses keys once and supplies related value/get/set queries from one snapshot. Literal spreads, array holes/unknown iterable suffixes, symbolic primitive indices and null/ignored/unknown prototypes retain their distinctions without per-index expansion. One normalized data-value field preserves original ordinary method nodes across symbolic, child-process and browser consumers; accessors stay unresolved. Duplicate prototype interpretation and stale fallback code are removed. Native descriptor semantics require a retained, read-specific global identity proof. Exact replacements accept canonical direct/global-object spellings only while namespace identity is preserved. Conditional, aliased, parent, mutating or possibly repeated effects remain unknown; only a proven inert empty or exact primitive-literal-return override preserves the previous export. Competing unconditional replacements use the existing definite-order fact for both selection and confirmation; effectful parameter binding stays unknown. Matching ambient facts are cached within the reader. No invocation-count heuristic or expanded call graph was added. Validation base `5508b2aff` is integrated; final head `e7e8d7bb645b67cb117c6257601a67198503af1e`. Twelve related files change; unrelated conflict resolutions retain main unchanged. ## Validation and limits - **384 tests pass:** 358 symbolic/AST/RPC/shape/ambient/child-process/browser tests and 26 public CLI/MCP acceptance/boundary tests. `check:fast`, `docs:check` and diff checks pass. - Original and review regressions were reproduced before their producing-boundary fixes. Actual Node confirms that the second invocation can read a replacement installed at the end of the first; unknown is required without a single-invocation proof. The directly affected RPC golden uses an independent runtime oracle. - Node 24.22.1 executes 35 test-owned CommonJS/TypeScript bundles (31060 bytes, 5308 AST nodes), with native type erasure and a recording Electron substitute: 26 channels execute, two mixed descriptors throw `TypeError`. Main has five false channels and misses six runtime channels. Fixed CLI/MCP produce 25 correct channels and explicitly leave the possibly repeated `patch-after-sender` factory unknown. **This is not native Electron or third-party application verification.** That deliberate static coverage gap remains; it is not counted as complete recovery. - Summary plus 3 `javascript-facts` pages projects four fields from 150 observations, with exhausted paging. Every row matches the full parent’s node/observation identity, source metadata, evidence, values and field omission. Both completed delivery workflows export the exact same full parent record (`ev_79710da7151e46c6115bd2238cb7f0f578934b2490197004391796cda2d0f76f`), root digest `239c4dd8b130e1d7be4682626f6b0464cc7eb90a5da03ec58afedad46f0ff660`. No tool catalog is fetched. - An expanded complete reply correctly exceeds the existing 10 MiB transport budget. Recovery uses retained Evidence and paging without reanalysis; no limit is raised and complete Evidence remains accessible. ## Shipped application control Official Obsidian 1.14.4 ASAR matches the pinned corpus digest `146ef8470d8cbb4a6db209bdc19bf2bad13350444186c0fd93c31087b142b92f` (27477966 bytes). Select exactly `main.js` and `enhance.js`: 216931 source bytes, 32508 AST nodes, 47 IPC channels and 47 main handlers. Base and fixed **complete Evidence records are byte-for-byte structurally equal**; MCP exports that same complete parent. The application was not executed. Coverage stays partial: 36 unselected files and bounded alias-selection unknowns remain. Single completed-task controls with a 512 MiB worker and the same two-CPU guard: base CLI 18.552s wall / 24.736s CPU / 846.98 MiB family peak; fixed CLI 16.877 / 23.416 / 816.02; fixed MCP 18.234 / 24.418 / 1107.77. These samples do not establish a speed gain. MCP uses 3 calls (summary, 3 sample rows, full export), 56054 reply bytes and no catalog. Its sample honestly covers 3/94 rows with a retained next offset; it is not exhaustive. Complete 28.4 MB Evidence stays on disk. Source selection, identities, provenance and omitted fields are verified. ## Resource evidence Two-CPU affinity, bounded worker heap, fresh analysis processes, warm runtime builds, uncontrolled OS caches; single completed-task samples do not establish a speed difference. | Workflow | Tool calls | Reply bytes | Wall / CPU seconds | Sampled family peak MiB | |---|---:|---:|---:|---:| | Main CLI + capture | — | — | 9.296 / 11.327 | 568.23 | | Fixed CLI + capture | — | — | 9.152 / 11.073 | 559.04 | | Summary + 3 pages + full export | 5 | 571742 | 9.764 / 11.603 | 619.10 | | Complete request + retained recovery + full export | 5 | 549034 | 10.620 / 11.714 | 628.12 | CPU/memory/wall include initialization, extraction and export; reply bytes exclude requests/initialization. RSS sampling can double-count shared pages and miss short peaks. Agent/schema tokens were not measured. Source bytes remain unchanged; complete controls and resource reports are retained locally. Final head CI `40381159489` passes source/static/docs/build, all eight end-to-end shards, real JavaScript recovery, Windows controls and the Linux native package baseline. Codex reports no major issues; Greptile passes; all review threads are resolved. Earlier canceled runs are not counted as passes. The unchecked human readiness declaration is left unchanged and will use the explicitly authorized bypass; it does not assert human review. ## Author readiness Human owner: @SII-YDD - [ ] 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 - [ ] This PR addresses a concrete problem or an agreed enhancement. - [ ] Material scope and authority questions are resolved, and any required maintainer scope decision covers the final diff. - [ ] The PR has one focused outcome and the title follows `type(scope): outcome`. - [ ] Related issue is linked, or the reason for not linking one is stated above. - [ ] Tests cover changed observable behavior and meaningful failure paths. - [ ] Owning docs, contracts, and generated metadata are updated where needed. - [ ] User-visible CLI/MCP changes include representative output. - [ ] I checked the final diff for secrets, unrelated cleanup, and unsupported claims. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-11-03 13:39:55 +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#7726

No description provided.

Rows
Columns