Read the clause (c) TAIL case on a checkout at a short fixed path, and drive lane-start's row stamp at three checkout lengths, so the case no longer answers by the length of the run root (#180)

[brettheap]

Closes #180

What was wrong

tests/test_lane_helpers.sh's clause (c) TAIL case asserted that repoA11b-1's row stamp ends dir $HOME/projects/repoA11b; window a11start:3 @42. lane-start's stamp is ONE line under ratified decision O1's 240-character cap (Amendment 13(a)). cut_to_line cuts it at 230 and drops the last partial word. The tail is the checkout path, so the case only passed while $HOME was short. Under plain pytest's /tmp it was green. Under tests/run.sh's per-run root (TMPDIR about 71 characters, so $HOME about 91) the line passed 230, and the cut landed on the ; right after the path, so the path went whole: the row read … window renamed, no launch; dir .... On macOS's runner the case was at the edge.

Reproduced outside the shell suite with a real lane-start, started twice --no-launch, under a 92-character $HOME:

• checkout under $HOME (107 characters): … window renamed, no launch; dir <107 chars>; .... The window is gone and the head is intact.
• the same $HOME with the checkout passed as --dir /tmp/ort-tail.XXXXXXXX/repoZ (28 characters): … no launch; dir /tmp/ort-tail.…/repoZ; window testsess:0 @1, 197 characters in all.

Why (A), the test-side fix, and not (B)

The cut is behaving as designed. Nothing in the stamp is broken under long paths.

• The design puts dir and window last on purpose. lane-start's Amendment 13(a) comment above the stamp (:3668-3677) orders the tail launch first, then dir and window, so "a line long enough to be cut loses what a reader can look up rather than what they cannot". Both facts are sub-fields of the same act's STARTED/RESUMED line in the lane's log, which lane-dir reads and the new test asserts in full. Both are also columns of Amendment 11's listing. No reader parses dir out of the state cell (checked with grep across lane-start, lane-end, lanes-edit.sh, lane, lanes, lane-rename, lane-worktrees, link-estates).
• The cut is already routine in production. Any launching resume (launching claude --name <lane> --resume <uuid>) is past 230 before dir even starts. So "the stamp drops dir/window under a long line" is today's normal behaviour, not an edge case of long paths.
• (B) would write a non-path after dir. Today a cut that lands inside the path drops it whole (dir ...), so the row never carries a truncated path that reads as a real one. (B)'s "truncate the dir, keep the facts" would put an elided path there. That changes the stamp format of a governed artefact, which is for the protocol to decide, not for a test fix.
• (B) alone cannot fix this case. The case asserts the FULL dir $HOME/…, which cannot fit under the cap at run.sh's length whatever the stamp does.
• cut_to_line is pinned. test_the_three_row_writers_carry_the_same_one_line_cut_byte_for_byte pins it byte for byte across lane-start, lane-end and lane-handoff.

So the case reads the tail on a checkout whose path has the same length wherever the suite runs. No product code changes. tests/run.sh is untouched: lane 1's #195 edits it.

What changed

tests/test_lane_helpers.sh (edited in place, never at the end; ranges are at this head):

• :66-67: A11T_ROOT="" beside SANDBOX=, before the traps, so cleanup never acts on an inherited value.
• cleanup() :101-103: removes $A11T_ROOT. It runs on EXIT, INT and TERM. Open PR A silent tmux is no longer read as a dead window, the companion verdict is read off the lane's live window from any window, and the guard's cure for a lane named before Rule 4 says what it cannot fill in (#73) #222 (at 2946a41) also adds a line to cleanup() here, so the two conflict in this one hunk of tests/test_lane_helpers.sh (~:101). The resolution is to keep both lines, and it falls to whichever PR lands second.
• clause (c) block :5356-5389: replaces the old :5351-5356.
  • repoA11b-1's second start keeps its exit, , no launch and head assertions. These do not depend on the root. One description now reads "…and its stamp says what actually happened" because its antecedent moved.
  • The TAIL assertion moves to a new lane repoA11t-1, whose checkout is mktemp -d <base>/ort-a11t.XXXXXX + /repoA11t. <base> is the sandbox's own $TMPDIR when that is at most 40 characters (plain pytest on Linux, where the checkout then lives inside $SANDBOX), and /tmp when it is not (run.sh, macOS). So the checkout is at most 65 characters, an assertion holds it there, and the line stays under about 210 wherever the root is. It runs in its own fake window a11tail:5 @43, which nothing else in the suite uses.
  • Only a SIGKILLed run under a long root leaves an ort-a11t.* behind, and it is in /tmp, outside run.sh's per-run root. Any other exit is cleaned up by cleanup.
  • The lane is started with --dir, then started again, and the second start reads the checkout back from the lane's log, the same shape as before.
  • repoA11b-1 keeps $HOME/projects/repoA11b as its recorded directory, because the later DIRECTORY PRECEDENCE dry-run (:5668) reads it back.

tests/test_lane_start_row_stamp.py (new): three parametrized cases, each a real lane-start started twice --no-launch in a sandbox whose $HOME is at least run.sh's length. The checkout is at an exact length under a /tmp mkdtemp, so the answer does not depend on pytest's own basetemp. The workspace seed is a data table written in a loop, which also clears SonarCloud's duplication gate against tests/test_lane_start_claude_current.py:

• short-fixed-checkout: the tail keeps dir and window.
• window-cut-dir-kept (100 characters): dir survives and the window ref is gone.
• path-dropped-whole-as-in-180 (160 characters): the path is dropped whole (no launch; dir ...) and never truncated. This is what tests/run.sh: a long run root truncates the row stamp so test_lane_helpers.sh's clause (c) TAIL case fails locally while CI's plain pytest passes #180's case read under run.sh's root.

All three also assert the head, the 240 cap, and the lane log's full dir …; window ….

Tests

• Local, under the shared lock (flock -w 1800 ${TMPDIR:-/tmp}/openrepotools-pytest.lock, lock taken 20:40:57Z, at the fix round's tree):
  • python3 -m pytest tests/test_lane_start_row_stamp.py -q: 3 passed in 17.69s (short-fixed-checkout, window-cut-dir-kept, path-dropped-whole-as-in-180).
  • The clause (c) block run on its own, with the suite's own setup (lines 1 to the first case), under a 72-character TMPDIR, the same shape as run.sh's root:
    • base f87fd5a: 8 passed, 1 failed, and the one failure is the TAIL case. This is tests/run.sh: a long run root truncates the row stamp so test_lane_helpers.sh's clause (c) TAIL case fails locally while CI's plain pytest passes #180 reproduced.
    • this branch under that 72-character TMPDIR: 12 passed, 0 failed.
    • this branch under TMPDIR=/tmp: 12 passed, 0 failed.
    • no /tmp/ort-a11t.* was left after any of those runs.
• CI at 085a234 (before this fix round): the Linux tests job passed, 1422 passed and 9 skipped. Two sibling PRs on the same base that add no pytest cases show 1419 passed and 9 skipped, so the difference is the three new cases. The shell suite's wrapper passed too. Every other job passed, except tests-macos, which waits for the ready label.
• SonarCloud at 085a234 failed its quality gate on 4.7% new-code duplication: the register seed, :164-177, against tests/test_lane_start_claude_current.py:198-211. The seed is restructured at this head.

🤖 Generated with Claude Code

Summary by Sourcery

Stabilize lane row-stamp coverage by testing checkout paths at controlled lengths instead of relying on the surrounding run root.

Bug Fixes:

• Make the clause (c) TAIL test independent of the run-root and home-directory path length by using a short fixed checkout path.
• Add coverage for row-stamp behavior across short, boundary-length, and long checkout paths, verifying paths are dropped whole rather than truncated.

Enhancements:

• Preserve validation of the complete directory and window details in the lane log while testing the capped row stamp.

Tests:

• Add parametrized integration tests that start lanes twice with real checkout paths and verify row-stamp caps, preserved heads, and path-dependent tail cuts.

[sourcery-ai]

Sorry @brettheap, you've used your own review budget of 250,000 diff characters for the last 7 days.

You can request another review in 22 hours and 56 minutes by commenting @sourcery-ai review. Upgrade to get a review now.

[sourcery-ai]

Reviewer's Guide

The PR fixes flaky clause (c) TAIL coverage by decoupling the full-tail assertion from variable $HOME/run-root lengths, without changing product code, and adds isolated parametrized tests that document and verify how lane-start stamps are cut at short, boundary, and long checkout paths.

File-Level Changes

Change	Details	Files
Make the shell-suite TAIL assertion independent of the run-root length by using a short, fixed checkout path.	Move the full dir/window TAIL assertion to a dedicated checkout under /tmp. Start the replacement lane twice so the second invocation reads the checkout from its log, preserving the original scenario. Add cleanup for the fixture created outside the suite sandbox while retaining the existing assertions on the original lane’s head and launch status.	tests/test_lane_helpers.sh
Add deterministic coverage for row-stamp truncation behavior across checkout path lengths.	Build an isolated workstation, register, fake tmux, and Git environment for real lane-start executions. Exercise short, 100-character, and 160-character checkout paths with two --no-launch starts. Verify the 240-character cap, intact stamp head, expected tail retention or whole-path omission, and full dir/window facts in the lane log.	tests/test_lane_start_row_stamp.py

Assessment against linked issues

Issue	Objective	Addressed	Explanation
#180	Make the clause (c) TAIL case pass independently of the length of the test run root by reading the row stamp on a checkout at a short, fixed path.	✅
#180	Exercise lane-start’s row-stamp behavior at checkout lengths that fit, partially exceed, and exceed the stamp’s available space, verifying the head remains intact and the directory is not misleadingly truncated.	✅

Possibly linked issues

• tests/run.sh: a long run root truncates the row stamp so test_lane_helpers.sh's clause (c) TAIL case fails locally while CI's plain pytest passes #180: The PR gives the TAIL assertion a short fixed checkout path and adds coverage for stamp behavior at longer checkout paths.
• tests/run.sh: a long run root truncates the row stamp so test_lane_helpers.sh's clause (c) TAIL case fails locally while CI's plain pytest passes #180: The PR fixes issue tests/run.sh: a long run root truncates the row stamp so test_lane_helpers.sh's clause (c) TAIL case fails locally while CI's plain pytest passes #180 by moving the TAIL assertion to a fixed short checkout and testing stamp behavior across path lengths.

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud