Prove #303's extension: a scenario sets a step's limit, and three scenarios show more time, a loop and the backstop #316

Merged
cmoriarty merged 1 commit from claude/prove-303 into main 2026-10-29 19:25:09 -04:00
Owner

Proves #303's extension through the engine. Until now only tests of its parts had run the wait loop's progress check.

What changed

  • A simulated-run scenario's limits: sets a step's time limit. The runner commits the pipeline Braid ships, unchanged but for those limits, as the repository's own .osf/pipelines/<name>.yaml, which osfd reads before the built-in. A test checks that nothing but the named limits differs from the shipped pipeline.
  • Three scenarios on quick-fix with a 20 s implementer limit:
    • limit-more-time: still writing files at its deadline, the step gets extended_s: 40 ("still working: 2 file write(s) in the last ten minutes"), finishes at 27 s, and the run succeeds. With the extension switched off it fails: the step is stopped at 20 s.
    • limit-loop: the same command over and over, nothing written: no more time, stopped at its deadline.
    • limit-backstop: writing every few seconds past three times its limit: more time once, then stopped at the backstop.
  • #303's change gains a simulated-runs delta (a scenario can set a step's time limit; scenarios for a step's limit) and tasks 3.1 to 3.3.

Not covered: a limit under ten minutes gets no message before its deadline, so these show the check at the deadline, not the one a little before it. Production's limits are 15 minutes to 3 hours, so #303 stays open until a real step shows it.

Checks

  • ./tools/test.sh: passed (239 backend, 63 UI, 18 browser).
  • ./tools/sim.sh: 72 of 72 passed.
  • Console, on a kept limit-more-time: the step's transcript reads ⏱ 40s more: still working: 2 file write(s) in the last ten minutes.

Only the simulated runs, a test, the sim guide and #303's OpenSpec change are touched, so the merge skips CI and doesn't deploy (run 147 is still working on production).

🤖 Generated with Claude Code

Proves #303's extension through the engine. Until now only tests of its parts had run the wait loop's progress check. **What changed** - A simulated-run scenario's `limits:` sets a step's time limit. The runner commits the pipeline Braid ships, unchanged but for those limits, as the repository's own `.osf/pipelines/<name>.yaml`, which osfd reads before the built-in. A test checks that nothing but the named limits differs from the shipped pipeline. - Three scenarios on quick-fix with a 20 s implementer limit: - `limit-more-time`: still writing files at its deadline, the step gets `extended_s: 40` ("still working: 2 file write(s) in the last ten minutes"), finishes at 27 s, and the run succeeds. With the extension switched off it fails: the step is stopped at 20 s. - `limit-loop`: the same command over and over, nothing written: no more time, stopped at its deadline. - `limit-backstop`: writing every few seconds past three times its limit: more time once, then stopped at the backstop. - #303's change gains a `simulated-runs` delta (a scenario can set a step's time limit; scenarios for a step's limit) and tasks 3.1 to 3.3. **Not covered:** a limit under ten minutes gets no message before its deadline, so these show the check at the deadline, not the one a little before it. Production's limits are 15 minutes to 3 hours, so #303 stays open until a real step shows it. **Checks** - `./tools/test.sh`: passed (239 backend, 63 UI, 18 browser). - `./tools/sim.sh`: 72 of 72 passed. - Console, on a kept `limit-more-time`: the step's transcript reads `⏱ 40s more: still working: 2 file write(s) in the last ten minutes`. Only the simulated runs, a test, the sim guide and #303's OpenSpec change are touched, so the merge skips CI and doesn't deploy (run 147 is still working on production). 🤖 Generated with [Claude Code](https://claude.com/claude-code)
A scenario's `limits:` maps a node id to its limit in seconds. The runner commits the
pipeline Braid ships, unchanged but for those limits, as the repository's own
.osf/pipelines/<name>.yaml, which osfd reads before the built-in, so a deadline comes
within seconds while the run still follows the shipped pipeline.

Three scenarios on quick-fix with a 20 s implementer limit:
- limit-more-time: still writing files at its deadline, the step gets 40 s more (as far
  as the backstop), finishes, and the run succeeds. With the extension switched off it
  fails: the step is stopped at 20 s.
- limit-loop: the same command over and over, nothing written: no more time, stopped
  at its deadline.
- limit-backstop: writing every few seconds past three times its limit: more time once,
  then stopped at the backstop.

Until now the wait loop's progress check had only run in tests of its parts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
cmoriarty deleted branch claude/prove-303 2026-10-29 19:25:09 -04:00
Sign in to join this conversation.
No reviewers
No labels
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
cmoriarty/braid!316
No description provided.