[PR #3305] [CLOSED] fix(apple): stop JVM class lookalikes passing as FAT Mach-O #5715

New issue

Closed

opened 2026-10-31 01:08:52 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-31 01:08:52 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/3305
Author: @Sriram-PR
Created: 10/14/2026
Status: ❌ Closed

Base: main ← Head: fix/fat-jvm-class-lookalike

📝 Commits (2)

a8e2a55 fix(apple): stop JVM class lookalikes passing as FAT Mach-O
d6b1041 test(apple): pin the binary target refusal for JVM FAT lookalikes

📊 Changes

3 files changed (+53 additions, -4 deletions)

View changed files

📝 src/artifacts/apple/MachoLoadCommandReader.test.ts (+38 -2)
📝 src/domain/apple/machoContainer.ts (+7 -2)
📝 src/domain/binaryTarget.test.ts (+8 -0)

📄 Description

Summary

trace-dylib-resolution with automatic root discovery reported a valid JVM class resource as a malformed Mach-O when its constant pool happened to contain a known Mach-O CPU value. The class landed in coverage.unparsed_images and coverage dropped to partial. classifyFatContainerIdentity treated any known CPU type in the FAT-shaped table as proof of Mach-O, then reported the slice range as malformed without ever confirming a slice header.

Fixes #3176

Problem and expected behavior

Smallest trigger: a class file (0xCAFEBABE, major version 65 read as 65 FAT records) whose constant pool puts 0x00000007 (x86) in a CPU-type slot, with a declared slice offset past the end of the file. This is the javaClassFile(4000, true) fixture the Go reader already tests against. Expected: not-mach-o, so the class stays out of the image list and coverage stays complete.

Change and scope

In src/domain/apple/machoContainer.ts, a known CPU type now counts as Mach-O identity only for a slice that starts inside the file. The parsed FAT header carries fileSize so the identity check can tell.
Real FAT containers whose slice starts inside the file but is truncated, misaligned, or uses an unsupported alignment are still reported as malformed/unsupported exactly as before (the existing reader tests for those cases pass unchanged).
Other identity consumers: the artifact graph format and the Go reader only act on a confirmed mach-o identity, so they are unchanged for these inputs. binaryTarget still rejects the file, but its reason changes from FAT architecture N slice range extends beyond the file to FAT table does not identify a Mach-O image.
Trade-off, stated plainly: a FAT file truncated before any slice begins, with no other evidence, is now not-mach-o instead of malformed. Its bytes can't be told apart from the JVM lookalike, and no existing test covers that case. Unknown-CPU tables were already classified this way.

Contribution scope: ordinary fix within existing boundaries; no scope decision needed.

Contract and boundary impact

Semantic owner and earliest changed stage: classifyFatContainerIdentity in src/domain/apple/machoContainer.ts, the shared FAT identity decision.
Contribution scope fit and linked maintainer decision: not applicable (bug fix, #3176).
CLI and MCP/tool-catalog contract: no schema change.
Provider, bridge, target-format, or platform compatibility: FAT-magic files whose known-CPU slices all start past the end of the file are classified not-mach-o instead of malformed; see trade-off above.
Evidence, artifact, provenance, or reconstruction contract: dylib traces no longer list such class files as images or unparsed images, so coverage is complete when nothing else is missing.
Target selection, authority, upstream behavior, and cleanup details: not applicable.
Generated metadata, package, or installation impact: none.

Expected effects (check known effects; use documentation-only on its own).
Unchecked items do not establish absence; leave uncertainties visible.

Read existing files or captured data.
Launch or attach to a process.
Inject instrumentation or run scripts/code.
Change process memory, registers, files, or configuration.
Contact network services.
Recover protected code or work around protections.
Documentation or usability only; no target interaction.
Not sure yet; investigation or maintainer guidance needed.

Evidence and regression coverage

Tests added or updated (src/artifacts/apple/MachoLoadCommandReader.test.ts):
- the existing JVM case now also reads javaClassFile(4000, true) and expects not-mach-o;
- a new dylib trace test enumerates an app bundle containing a Mach-O executable plus a lookalike and a plain class resource, and expects only the executable in images with complete coverage.
Base reproduction: both fail on the base revision (malformed instead of not-mach-o; trace images include Contents/Resources/Look.class).
An earlier attempt that ignored known CPUs on any invalid-range slice broke the existing truncated/unaligned/alignment-exponent FAT tests, which is why the rule is limited to slices starting past the end of the file.

User-visible CLI output: an Example.app with a thin arm64 Mach-O executable plus the two class resources:

base: images ["Contents/MacOS/Example", "Contents/Resources/Look.class"]
      coverage {"status":"partial","unparsed_images":["Contents/Resources/Look.class"]}
PR:   images ["Contents/MacOS/Example"]
      coverage {"status":"complete","unparsed_images":[]}

Remaining proof gaps: ran on Linux x64 only; no Hopper or other engine involved.

For evidence-bearing changes:

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

npm run test:local -- src/artifacts/apple/MachoLoadCommandReader.test.ts — 28 passed (both new assertions fail on the base revision)
npm run test:local -- src/domain/binaryTarget.test.ts — 37 passed (new JVM lookalike refusal-reason test fails on the base revision)
npm run test:focused on the 42 test files that use FAT/universal fixtures, javaClassFile or the FAT container helpers (Mach-O, dylib, artifact inventory, binary target, Go, Hopper, Ghidra) — 41 passed, 1 skipped; 836 tests passed, 9 skipped
npm run check:fast — typecheck and lint passed
CLI comparison above, built with npm run build:cached on base and on this branch

Compatibility, safety, and release

Breaking changes or migration steps: none.
Real Hopper/Ghidra, browser, or OS coverage: not applicable; Linux x64, Node 24.18.0.
Package or release metadata impact: none.
Security, privacy, process, or containment review: none; header classification only, no target execution.

Author readiness

Human owner: @Sriram-PR

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/3305 **Author:** [@Sriram-PR](https://github.com/Sriram-PR) **Created:** 10/14/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `fix/fat-jvm-class-lookalike` --- ### 📝 Commits (2) - [`a8e2a55`](https://github.com/morluto/rea/commit/a8e2a55c29c16e87b13a94bdcc4c9af9abf45fbc) fix(apple): stop JVM class lookalikes passing as FAT Mach-O - [`d6b1041`](https://github.com/morluto/rea/commit/d6b1041d70bf1c7997d86a812ae6b712737a54e2) test(apple): pin the binary target refusal for JVM FAT lookalikes ### 📊 Changes **3 files changed** (+53 additions, -4 deletions) <details> <summary>View changed files</summary> 📝 `src/artifacts/apple/MachoLoadCommandReader.test.ts` (+38 -2) 📝 `src/domain/apple/machoContainer.ts` (+7 -2) 📝 `src/domain/binaryTarget.test.ts` (+8 -0) </details> ### 📄 Description ## Summary `trace-dylib-resolution` with automatic root discovery reported a valid JVM class resource as a malformed Mach-O when its constant pool happened to contain a known Mach-O CPU value. The class landed in `coverage.unparsed_images` and coverage dropped to `partial`. `classifyFatContainerIdentity` treated any known CPU type in the FAT-shaped table as proof of Mach-O, then reported the slice range as malformed without ever confirming a slice header. Fixes #3176 ## Problem and expected behavior Smallest trigger: a class file (`0xCAFEBABE`, major version 65 read as 65 FAT records) whose constant pool puts `0x00000007` (x86) in a CPU-type slot, with a declared slice offset past the end of the file. This is the `javaClassFile(4000, true)` fixture the Go reader already tests against. Expected: `not-mach-o`, so the class stays out of the image list and coverage stays complete. ## Change and scope - In `src/domain/apple/machoContainer.ts`, a known CPU type now counts as Mach-O identity only for a slice that starts inside the file. The parsed FAT header carries `fileSize` so the identity check can tell. - Real FAT containers whose slice starts inside the file but is truncated, misaligned, or uses an unsupported alignment are still reported as `malformed`/`unsupported` exactly as before (the existing reader tests for those cases pass unchanged). - Other identity consumers: the artifact graph format and the Go reader only act on a confirmed `mach-o` identity, so they are unchanged for these inputs. `binaryTarget` still rejects the file, but its reason changes from `FAT architecture N slice range extends beyond the file` to `FAT table does not identify a Mach-O image`. - Trade-off, stated plainly: a FAT file truncated before *any* slice begins, with no other evidence, is now `not-mach-o` instead of `malformed`. Its bytes can't be told apart from the JVM lookalike, and no existing test covers that case. Unknown-CPU tables were already classified this way. Contribution scope: ordinary fix within existing boundaries; no scope decision needed. ## Contract and boundary impact - Semantic owner and earliest changed stage: `classifyFatContainerIdentity` in `src/domain/apple/machoContainer.ts`, the shared FAT identity decision. - Contribution scope fit and linked maintainer decision: not applicable (bug fix, #3176). - CLI and MCP/tool-catalog contract: no schema change. - Provider, bridge, target-format, or platform compatibility: FAT-magic files whose known-CPU slices all start past the end of the file are classified `not-mach-o` instead of `malformed`; see trade-off above. - Evidence, artifact, provenance, or reconstruction contract: dylib traces no longer list such class files as images or unparsed images, so coverage is `complete` when nothing else is missing. - Target selection, authority, upstream behavior, and cleanup details: not applicable. - Generated metadata, package, or installation impact: none. Expected effects (check known effects; use documentation-only on its own). Unchecked items do not establish absence; leave uncertainties visible. - [x] Read existing files or captured data. - [ ] Launch or attach to a process. - [ ] Inject instrumentation or run scripts/code. - [ ] Change process memory, registers, files, or configuration. - [ ] Contact network services. - [ ] Recover protected code or work around protections. - [ ] Documentation or usability only; no target interaction. - [ ] Not sure yet; investigation or maintainer guidance needed. ## Evidence and regression coverage - Tests added or updated (`src/artifacts/apple/MachoLoadCommandReader.test.ts`): - the existing JVM case now also reads `javaClassFile(4000, true)` and expects `not-mach-o`; - a new dylib trace test enumerates an app bundle containing a Mach-O executable plus a lookalike and a plain class resource, and expects only the executable in `images` with `complete` coverage. - Base reproduction: both fail on the base revision (`malformed` instead of `not-mach-o`; trace images include `Contents/Resources/Look.class`). - An earlier attempt that ignored known CPUs on any invalid-range slice broke the existing truncated/unaligned/alignment-exponent FAT tests, which is why the rule is limited to slices starting past the end of the file. - User-visible CLI output: an `Example.app` with a thin arm64 Mach-O executable plus the two class resources: ``` base: images ["Contents/MacOS/Example", "Contents/Resources/Look.class"] coverage {"status":"partial","unparsed_images":["Contents/Resources/Look.class"]} PR: images ["Contents/MacOS/Example"] coverage {"status":"complete","unparsed_images":[]} ``` - Remaining proof gaps: ran on Linux x64 only; no Hopper or other engine involved. For evidence-bearing changes: - [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 - `npm run test:local -- src/artifacts/apple/MachoLoadCommandReader.test.ts` — 28 passed (both new assertions fail on the base revision) - `npm run test:local -- src/domain/binaryTarget.test.ts` — 37 passed (new JVM lookalike refusal-reason test fails on the base revision) - `npm run test:focused` on the 42 test files that use FAT/universal fixtures, `javaClassFile` or the FAT container helpers (Mach-O, dylib, artifact inventory, binary target, Go, Hopper, Ghidra) — 41 passed, 1 skipped; 836 tests passed, 9 skipped - `npm run check:fast` — typecheck and lint passed - CLI comparison above, built with `npm run build:cached` on base and on this branch ## Compatibility, safety, and release - Breaking changes or migration steps: none. - Real Hopper/Ghidra, browser, or OS coverage: not applicable; Linux x64, Node 24.18.0. - Package or release metadata impact: none. - Security, privacy, process, or containment review: none; header classification only, no target execution. ## Author readiness Human owner: @Sriram-PR - [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 - [x] This PR addresses a concrete problem or an agreed enhancement. - [x] Material scope and authority questions are resolved, and any required maintainer scope decision covers the final diff. - [x] The PR has one focused outcome and the title follows `type(scope): outcome`. - [x] Related issue is linked, or the reason for not linking one is stated above. - [x] Tests cover changed observable behavior and meaningful failure paths. - [x] Owning docs, contracts, and generated metadata are updated where needed. - [x] User-visible CLI/MCP changes include representative output. - [x] 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

2026-10-31 01:08:52 +00:00