Commit Graph
3 Commits
Author SHA1 Message Date
GodopuandClaude Opus 5.5 f26952ccc3 fix(issue-36): the legacy shell loop and lib.sh's submit helpers pass prompts by file
The two remaining exec points that carried a prompt as one argv string
(capped at 131,072 bytes):

- run_loop.sh (MAM_LOOP_PY=0): delegate_job_safe, which all eleven shell
  delegations go through, turns `--prompt <text>` into `--prompt-file` under
  the control root's .mam/ and removes the file once the wrapper returns. Its
  failure lines go to stderr: callers capture stdout, and under set -e a
  failed submit ends the loop at that line with the wrapper's code, so the
  cause was being swallowed.
- lib.sh delegate_submit_job and its legacy twin (create_session's onboarding
  and --submit-job): the prompt is written under .mam/ and handed on as
  --prompt-file; lib_py.delegation.cli submit accepts --prompt-file and
  forwards it to the registry. --prompt keeps working everywhere.

S6/S6b push a 140,000-byte prompt through the CLI and both helpers; S7 runs
the real run_loop.sh with a 140,000-byte plan and a 150,000-byte diff; S7b
fixes a nonzero exit when the implementation or compliance job cannot be
registered. All six fail on 14d11bf.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-24 06:08:25 +00:00
GodopuandClaude Opus 5.5 14d11bf062 fix(issue-36): a job that cannot be registered after a rebuttal ends the loop with rc 1
Review of f5cadd9 (planner-reviewer-codex-01, R1): the compliance job's
registration failure was only logged, so a passing next round ended the
real loop process with rc 0. Compliance now returns 1 when it cannot be
registered or does not finish, as the correction step already does.
Re-adjudication and arbitration also return 1 when their job cannot be
registered; a registered job that yields no verdict keeps the designed
fail-closed fallback.

S4c (stubbed delegate_job, a fixture whose later rounds never pass) is
replaced by S3c: a real `python -m lib_py.loop` process whose wrapper runs
an executable that simulates the agents, refusing the compliance,
re-adjudication or arbitration registration while the next round passes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-24 03:38:10 +00:00
GodopuandClaude Opus 5.5 f5cadd98c0 test(issue-36): prompts over the argv cap go through the real wrapper, registry and loop; a failed delegation ends the loop with rc 1 and no traceback
S1-S1d push a 140,000-byte prompt through the real wrapper and registry by
file, with a negative control that the same prompt as an argument fails with
E2BIG. S2 runs the real runner against a recording wrapper. S3/S3b run a real
`python -m lib_py.loop` whose wrapper fails or cannot start. S4/S4b/S4c cover
the exec-error handler, the busy retry and the compliance log. S5 sends a
collected diff from the former 131,072..200,000-byte blind spot through whole.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-24 02:30:13 +00:00