GodopuandClaude Fable 5.1
c3e6a7e96d
Issue #36: delegated prompts travel by file, a delegation that fails ends the loop with rc 1. VERSIONS.md carries the entry; the merge-side reports (round 1 @59be1c8 and round 2 @708262a with identity sidecars, base59be1c8-> head708262a, evidence review-tree) are archived under .agents/reports/<session>/. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VCgu9Fa9iT7TbJUSzdq4Ff
77 lines
5.5 KiB
Markdown
77 lines
5.5 KiB
Markdown
# Pre-Merge Independent Delta Review Report: PR #37 (`fix/issue-36` Round 2)
|
||
|
||
- **Job ID**: `3121d389`
|
||
- **Target Agent**: `agy` (Session: `herdr:reviewer-agy-01`)
|
||
- **Role**: Reviewer
|
||
- **Target Worktree Context**:
|
||
- **Loop ID**: `loop-20260924T085626Z-b8987e`
|
||
- **Base Commit**: `59be1c89209a232059aec0c5aeecc2329929cb49` (Round 1 target)
|
||
- **Head Commit**: `708262a598c4301c0ccc70efba5401093b4b1e30` (1 delta commit: +68 / −7)
|
||
- **Tree SHA**: `d601c2d01bb4c09931e5e1989e140b01d51f5b96`
|
||
- **Verification Path**: `/var/folders/q_/68sxgw3j073dgjg8wl9s08vw0000gn/T/mam-review/loop-20260924T085626Z-b8987e/1`
|
||
|
||
---
|
||
|
||
## 1. Executive Summary
|
||
|
||
This second review round examines delta commit `708262a` ("fix(issue-36): merge review - handle prompt write failure in runner; reject empty prompt file; make test_s9 stderr check portable").
|
||
|
||
The commit directly resolves all three findings identified during Round 1:
|
||
1. **R1 / H1 (Prompt Write Failure Handling in `runner.py`)**: `_run_cmd` now wraps temporary prompt file creation and writing in `try ... except OSError:`. If writing fails (e.g. `ENOSPC`, `EFBIG`), any partial prompt file is unlinked and a `CompletedProcess(..., 126, "", msg)` is returned. The wrapper is not invoked, no unhandled exceptions bubble to `LoopRunner.run()`, and the failure is logged cleanly. Verified by new test `test_s4c`.
|
||
2. **R2 / N1 (macOS Portability in `test_s9`)**: In `tests/test_issue36_prompt_transport.py`, line 637 guards the assertion `assert "File too large" in result.stderr` with `if sys.platform == "linux":`. This accommodates macOS `/bin/bash` 3.2.57 behavior while continuing to assert `result.returncode != 0`, clean file removal (`assert not leftovers`), and the portable MAM diagnostic (`assert "delegate_job_safe: cannot write the prompt file" in result.stderr`) across all platforms.
|
||
3. **H3 (Empty Prompt File Rejection)**: `multi-agent-mux-delegate-job` now asserts `[[ -s "$PROMPT_FILE" ]]` in `cmd_submit()`, rejecting empty files before registration. Verified by new test `test_s1e`.
|
||
|
||
All 23 applicable tests in `tests/test_issue36_prompt_transport.py` passed cleanly in the macOS verification environment (with 1 platform-specific test skipped on Darwin as intended).
|
||
|
||
---
|
||
|
||
## 2. Detailed Verification Against Checklist Items
|
||
|
||
### (1) Verification of the Three Delta Fixes & Side-Effect Analysis
|
||
- **Prompt File Creation & Cleanup in `runner.py` (R1/H1)**:
|
||
- `prompt_file` is initialized to `""`.
|
||
- In `_run_cmd`, `tempfile.NamedTemporaryFile` and `fh.write(p_text)` are wrapped in `try ... except OSError as e:`.
|
||
- If writing fails:
|
||
* If `prompt_file` exists, it is unlinked immediately in the `except` block.
|
||
* Returns `subprocess.CompletedProcess(["bash", str(wrapper_script), "submit"], 126, "", msg)`.
|
||
* Because it returns directly from the `except` block, the downstream `subprocess.run(cmd)` and the outer `finally:` block do not execute. The wrapper is never invoked with an incomplete file.
|
||
- On the normal path:
|
||
* The wrapper executes, and the outer `finally: os.unlink(prompt_file)` ensures the file is deleted exactly once.
|
||
- The returned exit code 126 triggers the existing retry mechanism in `delegate_job()`; if the write still fails, the error message is logged cleanly and `delegate_job()` returns `None`, terminating the loop with exit code 1 without tracebacks.
|
||
- Verified by `test_s4c_a_prompt_file_that_cannot_be_written_is_a_failed_submit`.
|
||
- **macOS Test Portability for `test_s9` (R2/N1)**:
|
||
- Scoping `assert "File too large" in result.stderr` to `sys.platform == "linux"` eliminates the macOS `AssertionError` that caused the Round 1 `NOT PASS`.
|
||
- Portable assertions across Darwin and Linux verify:
|
||
* `assert result.returncode != 0` (non-zero exit upon write failure).
|
||
* `assert not leftovers` (partial `.mam/prompt-shloop-*` files are deleted).
|
||
* `assert "delegate_job_safe: cannot write the prompt file" in result.stderr` (MAM error diagnostics printed to stderr).
|
||
- Verified by running `tests/test_issue36_prompt_transport.py::test_s9_shell_loop_removes_a_partial_prompt_file_and_says_why` on macOS (PASS).
|
||
- **Empty Prompt File Rejection (H3)**:
|
||
- `cmd_submit()` in `multi-agent-mux-delegate-job` checks `[[ -s "$PROMPT_FILE" ]] || { echo "submit: --prompt-file $PROMPT_FILE is empty" >&2; exit 1; }`.
|
||
- Rejects zero-byte prompt files fail-closed before invoking `registry.py`.
|
||
- Verified by `test_s1e_an_empty_prompt_file_is_refused`.
|
||
|
||
### (2) Impact of Empty File Rejection on Valid Workflows
|
||
- In MAM, all valid delegated prompts (task instructions, review requests, refinement briefs, compliance orders) require meaningful textual context.
|
||
- Loop brief provisioning (`prompt.md`) always writes non-empty markdown briefs containing metadata and task descriptions.
|
||
- Enforcing `-s` prevents unintended empty job dispatches without affecting legitimate system workflows.
|
||
|
||
### (3) Overall Merge Readiness
|
||
- Commit `708262a` cleanly addresses all review concerns with minimal, surgical changes (+68 / −7).
|
||
- `git diff 59be1c8...HEAD --check` reported 0 whitespace or formatting errors.
|
||
- Version bump to `v4.7.3` is scheduled for the orchestrator release commit.
|
||
- Zero blockers remaining.
|
||
|
||
---
|
||
|
||
## 3. Test Execution Summary
|
||
|
||
| Test Suite / Target | Result | Duration | Notes |
|
||
| :--- | :--- | :--- | :--- |
|
||
| `tests/test_issue36_prompt_transport.py` | **23 PASS, 1 SKIP** | 42.4s | 23 passed on macOS; `test_s1b` skipped (`linux_only` constant test); `test_s9`, `test_s4c`, `test_s1e` all pass |
|
||
| `git diff 59be1c8...HEAD --check` | **0 errors** | 0.05s | Clean formatting and no whitespace errors |
|
||
|
||
---
|
||
|
||
[VERDICT: PASS]
|