morluto/rea

Fork 0

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

[PR #4098] [MERGED] fix(kuna): report output beyond REA's buffer as deterministic and final #6057

New issue

Closed

opened 2026-10-31 01:11:31 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-31 01:11:31 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/morluto/rea/pull/4098
Author: @kaoru0822-kitauji
Created: 10/15/2026
Status: ✅ Merged
Merged: 10/15/2026
Merged by: @kaoru0822-kitauji

Base: main ← Head: fix/kuna-output-limit-not-retryable

📝 Commits (1)

cb702ea fix(kuna): report output beyond REA's buffer as deterministic and final

📊 Changes

5 files changed (+122 additions, -7 deletions)

View changed files

📝 docs/kuna-provider.md (+1 -1)
📝 src/kuna/KunaCommand.ts (+14 -4)
📝 src/kuna/KunaReferenceQueries.ts (+16 -1)
📝 tests/boundary/providers/kuna/fakeKuna.ts (+9 -1)
➕ tests/boundary/providers/kuna/kunaCommandFailures.test.ts (+82 -0)

📄 Description

Summary

When Kuna's output overflowed REA's fixed buffer for a command (graph 32 MiB, xrefs and single-function text 16 MiB, inventory 128 MiB), REA reported a capacity resource constraint. MCP/CLI project capacity as retryable: true with "Wait for an active analysis to complete, then retry within the reported admission limits." That advice is wrong: the same target and Kuna executable produce the same output, so every retry overflows again. For call queries, each retry also repeats the whole-program kuna decompile-graph walk, and the error carried about 4 KB of the cut-off JSON document. Refs #1957.

Problem and expected behavior

Smallest real trigger: Kuna 1.833 on macOS arm64, lipo -thin arm64 of /Applications/Firefox.app/Contents/MacOS/XUL (169 MB Mach-O, 403,795 inventory entries), then MCP procedure_callers for any function (0x46e5d4):

procedure_callers  113489 ms  6306 bytes  code=resource_constraint retryable=true
  "kuna decompile-graph wrote more than 33554432 bytes of output."
  remediation: "Wait for an active analysis to complete, then retry within the reported admission limits."
  details.captured_output.stdout: 4096 chars from the middle of the graph JSON
procedure_callers (again)  113115 ms  same error

For scale, a 14 MB Sublime Text arm64 binary produces a 31.8 MB graph (41,773 functions, 139,662 edges), just under the budget.

Expected: a non-retryable error that says why repeating cannot help, and no second whole-program walk in the same session.

Change and scope

KunaCommand: an output-limit failure becomes AnalysisResourceConstraintError with resource memory, as the WABT and pcap adapters already classify their fixed output budgets. Its remediation reads "REA buffers at most N bytes of kuna output, so repeating this request for this target fails the same way. Use another provider for it." memory projects as retryable: false.
The same branch no longer returns the overflowed stdout tail (captured_output.stdout: "", truncated: true); stderr, exit status and signal remain.
KunaReferenceQueries remembers a graph that overflowed (#graphFailure, cleared on close), so later call queries in the session return that failure at once.
docs/kuna-provider.md: the graph budget paragraph names a real image that exceeds it and states the error behavior.

Not included: raising the 32 MiB graph budget or parsing the graph as a stream. The budget bounds REA's memory (raw JSON plus materialized indexes) and was chosen with the call-relationship feature; changing it is a design decision with its own memory measurements. XUL's call relationships therefore remain unavailable from Kuna. Per-address xref overflows are not memoized.

Scope: an ordinary fix within the existing Kuna provider boundary; no maintainer scope decision applies.

Contract and boundary impact

Semantic owner and earliest changed stage: KunaCommand.commandFailure (the classifier for every Kuna run); KunaReferenceQueries (graph reuse).
Contribution scope fit and linked maintainer decision: not applicable.
CLI and MCP/tool-catalog contract: no schema change. An output overflow's details.resource changes from capacity to memory, and retryable from true to false.
Provider, bridge, target-format, or platform compatibility: Kuna only.
Evidence, artifact, provenance, or reconstruction contract: reported_limits.output_limit_bytes is unchanged. The overflowed stdout fragment is dropped; nothing else from the failure is.
Target selection, authority, upstream behavior, and cleanup details: fewer repeated Kuna runs; cleanup is unchanged.
Generated metadata, package, or installation impact: none.

Expected effects:

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.

(Existing kuna decompile-graph runs; this change runs fewer of them.)

Evidence and regression coverage

Tests added or updated: new kunaCommandFailures.test.ts:
- KunaTargetCommands.run with a 100-byte buffer over the fake inventory gives a memory constraint, retryable: false, the exact remediation, and an empty, truncated stdout capture;
- with a new oversized fake graph mode (33 MiB of output), two procedure_callers calls both fail as memory/not retryable after a single decompile-graph run.
Base reproduction: both tests fail on the base revision (git stash of src/kuna).

Real Kuna 1.833, XUL arm64 slice as above, after the fix:

procedure_callers  115322 ms  813 bytes  retryable=false
  remediation: "REA buffers at most 33554432 bytes of kuna decompile-graph output, so repeating this request for this target fails the same way. Use another provider for it."
  captured_output: {stdout: "", stderr: "", truncated: true, exit_code: 0, signal: null}
procedure_callers (again)  2 ms  same error
procedure_callees          3 ms  same error

Each error response drops from 6306–6383 to 813 bytes. Model tokens were not measured.

Remaining proof gaps: no Linux run; the overflow is exercised for the graph and inventory paths, while single-function and xref overflows go through the same classifier without their own tests.
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:focused -- tests/boundary/providers/kuna/*.test.ts: 136 passed, two consecutive runs on the final change.
New tests on the base revision: 2 failed.
npm run check:fast: typecheck and lint passed.
npm run docs:check: passed.
REA_KUNA_COMMAND=… npm run verify:kuna (real Kuna 1.833, darwin/arm64): passed, cleanup: "complete" (run before the stdout change, which affects only output-limit failures).
Real XUL runs before and after, above.

Compatibility, safety, and release

Breaking changes or migration steps: none. Only how an output overflow is classified and repeated changes.
Real Hopper/Ghidra, browser, or OS coverage: real Kuna on macOS arm64 only.
Package or release metadata impact: none.
Security, privacy, process, or containment review: no new side effects; fewer repeated Kuna processes.

Author readiness

Human owner:

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.

🤖 Generated with Claude Code

_{🔄 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/4098 **Author:** [@kaoru0822-kitauji](https://github.com/kaoru0822-kitauji) **Created:** 10/15/2026 **Status:** ✅ Merged **Merged:** 10/15/2026 **Merged by:** [@kaoru0822-kitauji](https://github.com/kaoru0822-kitauji) **Base:** `main` ← **Head:** `fix/kuna-output-limit-not-retryable` --- ### 📝 Commits (1) - [`cb702ea`](https://github.com/morluto/rea/commit/cb702ea52814491506d590dd96a1543917b97927) fix(kuna): report output beyond REA's buffer as deterministic and final ### 📊 Changes **5 files changed** (+122 additions, -7 deletions) <details> <summary>View changed files</summary> 📝 `docs/kuna-provider.md` (+1 -1) 📝 `src/kuna/KunaCommand.ts` (+14 -4) 📝 `src/kuna/KunaReferenceQueries.ts` (+16 -1) 📝 `tests/boundary/providers/kuna/fakeKuna.ts` (+9 -1) ➕ `tests/boundary/providers/kuna/kunaCommandFailures.test.ts` (+82 -0) </details> ### 📄 Description ## Summary When Kuna's output overflowed REA's fixed buffer for a command (graph 32 MiB, xrefs and single-function text 16 MiB, inventory 128 MiB), REA reported a `capacity` resource constraint. MCP/CLI project `capacity` as `retryable: true` with "Wait for an active analysis to complete, then retry within the reported admission limits." That advice is wrong: the same target and Kuna executable produce the same output, so every retry overflows again. For call queries, each retry also repeats the whole-program `kuna decompile-graph` walk, and the error carried about 4 KB of the cut-off JSON document. Refs #1957. ## Problem and expected behavior Smallest real trigger: Kuna 1.833 on macOS arm64, `lipo -thin arm64` of `/Applications/Firefox.app/Contents/MacOS/XUL` (169 MB Mach-O, 403,795 inventory entries), then MCP `procedure_callers` for any function (`0x46e5d4`): ``` procedure_callers 113489 ms 6306 bytes code=resource_constraint retryable=true "kuna decompile-graph wrote more than 33554432 bytes of output." remediation: "Wait for an active analysis to complete, then retry within the reported admission limits." details.captured_output.stdout: 4096 chars from the middle of the graph JSON procedure_callers (again) 113115 ms same error ``` For scale, a 14 MB Sublime Text arm64 binary produces a 31.8 MB graph (41,773 functions, 139,662 edges), just under the budget. Expected: a non-retryable error that says why repeating cannot help, and no second whole-program walk in the same session. ## Change and scope - `KunaCommand`: an `output-limit` failure becomes `AnalysisResourceConstraintError` with resource `memory`, as the WABT and pcap adapters already classify their fixed output budgets. Its remediation reads "REA buffers at most N bytes of kuna <subcommand> output, so repeating this request for this target fails the same way. Use another provider for it." `memory` projects as `retryable: false`. - The same branch no longer returns the overflowed stdout tail (`captured_output.stdout: ""`, `truncated: true`); stderr, exit status and signal remain. - `KunaReferenceQueries` remembers a graph that overflowed (`#graphFailure`, cleared on close), so later call queries in the session return that failure at once. - `docs/kuna-provider.md`: the graph budget paragraph names a real image that exceeds it and states the error behavior. Not included: raising the 32 MiB graph budget or parsing the graph as a stream. The budget bounds REA's memory (raw JSON plus materialized indexes) and was chosen with the call-relationship feature; changing it is a design decision with its own memory measurements. XUL's call relationships therefore remain unavailable from Kuna. Per-address xref overflows are not memoized. Scope: an ordinary fix within the existing Kuna provider boundary; no maintainer scope decision applies. ## Contract and boundary impact - Semantic owner and earliest changed stage: `KunaCommand.commandFailure` (the classifier for every Kuna run); `KunaReferenceQueries` (graph reuse). - Contribution scope fit and linked maintainer decision: not applicable. - CLI and MCP/tool-catalog contract: no schema change. An output overflow's `details.resource` changes from `capacity` to `memory`, and `retryable` from true to false. - Provider, bridge, target-format, or platform compatibility: Kuna only. - Evidence, artifact, provenance, or reconstruction contract: `reported_limits.output_limit_bytes` is unchanged. The overflowed stdout fragment is dropped; nothing else from the failure is. - Target selection, authority, upstream behavior, and cleanup details: fewer repeated Kuna runs; cleanup is unchanged. - Generated metadata, package, or installation impact: none. Expected effects: - [x] Read existing files or captured data. - [x] 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. (Existing `kuna decompile-graph` runs; this change runs fewer of them.) ## Evidence and regression coverage - Tests added or updated: new `kunaCommandFailures.test.ts`: - `KunaTargetCommands.run` with a 100-byte buffer over the fake inventory gives a `memory` constraint, `retryable: false`, the exact remediation, and an empty, truncated stdout capture; - with a new `oversized` fake graph mode (33 MiB of output), two `procedure_callers` calls both fail as `memory`/not retryable after a single `decompile-graph` run. - Base reproduction: both tests fail on the base revision (`git stash` of `src/kuna`). - Real Kuna 1.833, XUL arm64 slice as above, after the fix: ``` procedure_callers 115322 ms 813 bytes retryable=false remediation: "REA buffers at most 33554432 bytes of kuna decompile-graph output, so repeating this request for this target fails the same way. Use another provider for it." captured_output: {stdout: "", stderr: "", truncated: true, exit_code: 0, signal: null} procedure_callers (again) 2 ms same error procedure_callees 3 ms same error ``` Each error response drops from 6306–6383 to 813 bytes. Model tokens were not measured. - Remaining proof gaps: no Linux run; the overflow is exercised for the graph and inventory paths, while single-function and xref overflows go through the same classifier without their own tests. - [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:focused -- tests/boundary/providers/kuna/*.test.ts`: 136 passed, two consecutive runs on the final change. - New tests on the base revision: 2 failed. - `npm run check:fast`: typecheck and lint passed. - `npm run docs:check`: passed. - `REA_KUNA_COMMAND=… npm run verify:kuna` (real Kuna 1.833, darwin/arm64): passed, `cleanup: "complete"` (run before the stdout change, which affects only output-limit failures). - Real XUL runs before and after, above. ## Compatibility, safety, and release - Breaking changes or migration steps: none. Only how an output overflow is classified and repeated changes. - Real Hopper/Ghidra, browser, or OS coverage: real Kuna on macOS arm64 only. - Package or release metadata impact: none. - Security, privacy, process, or containment review: no new side effects; fewer repeated Kuna processes. ## Author readiness Human owner:  - [ ] 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. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-31 01:11:31 +00:00

closed this issue
added the
pull-request
label

bsakaguchi referenced this issue

2026-11-01 21:45:48 +00:00

[PR #6143] [MERGED] fix(cancellation): preserve caller abort reasons through late finalization #7158

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#6057

No description provided.

Rows
Columns