[brettheap]

Lane: openRepoTools-5. Fixes the two indeterminate reads #60 names. Each one picked a branch of the (b) table instead of naming the read.

What was wrong

1. The register's text. register_text answers empty text and exit 0 when neither origin/<branch>:<path> nor the working file renders, because it ends in a printf. The guard's only register check was [ ! -f "$LANES_FILE" ]. A register file that exists but cannot be read, with no published copy, therefore reached the table with no rows. The table's last arm then refused it. Measured at f87fd5a with the suite's own fixtures (window repoGD-1, this uuid the row's last id, register text empty):

lanes-edit:   row      the register has no row for `repoGD-1`
lanes-edit: this window is lane repoGD-1, whose row does not name aaaa0012-1111-4000-8000-aaaa00121111 at all — the harness minted a new transcript with nobody acting (a /clear, a usage reset, a profile switch). ...
lanes-edit: run: lane-start --no-launch repoGD 1

Both the sentence and the cure are wrong for a read that did not happen.

2. The lane's binding log (drifted since the issue was filed). Since #87 the guard no longer calls guard_rename_is_newer. Its one remaining caller is the SessionStart hook (ssb_name_line). That test folded lane_binding_utc's 1 ("the log could not be read", e.g. two published logs differing only by case) into "no binding, so the rename is newer". Measured at f87fd5a: the session was renamed by a person to repoLate-1 a day ago, before the STARTED the row's own log carries, and the lane repoTwo-1 has two logs differing by case. The hook printed:

session name 'repoLate-1' is a lane name YOU set, newer than this window's binding to 'repoTwo-1' — the name guard offers the move at your next prompt (Amendment 12(h) rule 2); nothing was renamed here

That statement is about a binding nobody read, and here it is false.

What changed

• The guard now tests the register's text positively, once, before the first row read. register_text | awk 'NF { t = 1 } END { exit !t }' runs right before the THE ROW section, and on no text the guard refuses with exit 2 (only 2 blocks a prompt), naming the read and printing the safe-mode bypass:

  lanes-edit:   window   gdsess:@12 'repoGD-1'
  lanes-edit:   session  aaaa0012-1111-4000-8000-aaaa00121111 'repoGD-1' (nameSource user) pid … profile t1
  lanes-edit:   row      not read
  lanes-edit: the register's TEXT could not be read — neither `origin/main:lanes/register-60.md` nor …/register-60.md rendered a line — so the ROW cannot be read and the triple cannot be verified. A register that could not be read is not a register with no row for this lane, and an indeterminate read refuses (Amendment 12(d)). If this workstation cannot answer at all, `claude --safe-mode` runs with every hook off — …
  

  • Where it sits, and why that differs from the issue's suggestion. The issue proposed putting the test beside the [ ! -f ] refusal. I tried that first. There, neither the window nor the session had been read yet, so this refusal's triple printed window not in tmux 'none' about a session that is in tmux. At the head of the row reads, the test is still once and before the table, and the triple is accurate.
  • Cost, against the 5 s budget. The text is cached beside $SE_CACHE_FILE for the life of the process, so the test moves the first render rather than adding one. With GIT_TRACE on a register padded to 1.3 MB, the agreeing case makes 1 git show of the register and 12 git commands in all, both before and after. The added cat | awk costs about 6 ms. A $( ) capture of the same text cost about 30 ms, which is why the check streams. Agreeing-case wall time over 11 alternating runs: before median 0.507 s, after 0.523 s (load average about 18). All 22 runs exited 0 and printed nothing.
• guard_rename_is_newer now has a third state. It returns 0 newer · 1 not newer, or no nameSince · 5 the lane's log could not be read. Any lane_binding_utc status other than 0 or 8 is 5. The direction the issue asked to keep holds: an unreadable log is never "not newer", so the hook never types over a person's own /rename.
• The SessionStart hook names that read. It refuses nothing and exits 0 as before:

  session name 'repoLate-1' is a lane name YOU set, and lane repoTwo-1's object log could not be read, so whether you set it after this window's binding is NOT known — nothing was renamed here, and the name guard puts it to you at your next prompt (Amendment 12(h) rule 4)
  

• Dead code removed. gr_since in guard_run was the guard's input to guard_rename_is_newer until The name guard offers a readable, recoverable lane/session mismatch three choices, and lane-start resumes an exact transcript uuid, never a title #87, and nothing reads it. Removing it also saves one jnum per prompt.

No doc quotes either changed message. docs/README-lanes.md already promises that an unreadable register is refused by naming the read (the "Fail closed" section), and the code now does that.

Functions and regions edited

• lanes-edit.sh: guard_run: new block immediately before # ---- THE ROW, from origin/<branch>…; the gr_since line removed.
• lanes-edit.sh: guard_rename_is_newer: header comment and body.
• lanes-edit.sh: ssb_name_line: the nameSource: user rename branch, now a case on the test's status.
• tests/test_lane_helpers.sh: the Amendment 12 section, a new block after the "payload naming no cwd" case. The Amendment 15 section, a new block inside the "two published logs for one lane" fixture, before a15_unpublish.

Tests

Amendment 12 section: a register that is there and whose text could not be read. LANES_DIR keeps the logs and alias table in place, and LANES_FILE names a file on no published tree, so register_text falls through to it.

• Empty file (runs on every host and uid). Asserts exit 2, the read named, the window in the triple, row not read, not the last arm's sentence, and the safe-mode bypass.
• Mode-000 copy of the register (the issue's literal case). Asserts the same refusal. It is skipped as root, the suite's existing pattern.
• Why not a dangling symlink. The brief suggested one, but it is not a fixture for this check: [ -f ] is false for a dangling link, so the existing workspace refusal already answers it before and after this change.

Amendment 15 section: the two-logs-differing-by-case fixture. A nameSource: user rename to repoLate-1 from a day ago, in a window bound to repoTwo-1. Asserts that session-start exits 0, names "lane repoTwo-1's object log could not be read", never says "newer than this window's binding", and types nothing.

Fail-before and pass-after, proved locally in isolation. The full suite was not run locally, per AGENTS.md; CI's tests job is the record. I built a scratch runner from this branch's tests/test_lane_helpers.sh: the prologue (lines 1–971, SRC_DIR pointed at the tree under test), the definitions of E, LOGD, LANE, LANES_CMD and profiles_root, the whole Amendment 12 section, the Amendment 15 section through the two-logs fixture, and the footer. It ran in its own mktemp -d sandbox with env -u LANES_EDIT -u LANES_WORKSPACE_ROOT -u LANES_REPO -u LANES_LANE -u LANES_FILE -u LANES_DIR -u LANES_PATH -u TMUX:

• Source tree with f87fd5a's lanes-edit.sh: 343 passed, 7 failed. The 7 are exactly the new assertions (the window-in-triple one passes on both sides).
• This branch: 350 passed, 0 failed, 0 skipped.

Also passing: python3 -m pytest tests/test_guard_launch_mode.py tests/test_repo_hygiene.py (165 passed), bash -n on both changed files, and git diff --check.

Closes #60

🤖 Generated with Claude Code

Summary by Sourcery

Fail closed on unreadable register and binding-log reads so the name guard and SessionStart hook never infer state from data they could not read.

Bug Fixes:

• Make the name guard refuse with an explicit unreadable-register message instead of misreporting that the lane has no row.
• Prevent SessionStart from treating an unreadable lane binding log as evidence that a user rename is newer, and report the indeterminate read without renaming.

Enhancements:

• Remove the obsolete guard rename timestamp input and add a distinct unreadable-log state to rename comparisons.

Tests:

• Add coverage for unreadable register files and case-colliding binding logs in the guard and SessionStart hook tests.

[sourcery-ai]

Reviewer's Guide

The PR makes both guard and SessionStart behavior fail closed on indeterminate reads: the guard positively verifies register text before reading rows, while rename recency gains an explicit unreadable-log state that SessionStart reports without acting. Focused regression tests cover diagnostics, exit behavior, and prevention of accidental renames.

Sequence diagram for fail-closed register validation

sequenceDiagram
    participant Guard as guard_run
    participant Register as register_text
    participant Rows as Row reader
    participant Refusal as guard_refuse

    Guard->>Register: register_text
    Register-->>Guard: Rendered register text
    alt Text contains a line
        Guard->>Rows: Read the row
    else Text is empty or unreadable
        Guard->>Refusal: guard_refuse
        Refusal-->>Guard: Exit 2 with named read and safe-mode bypass
    end

Loading

Sequence diagram for unreadable binding log handling

sequenceDiagram
    participant Hook as ssb_name_line
    participant Recency as guard_rename_is_newer
    participant Binding as lane_binding_utc

    Hook->>Recency: guard_rename_is_newer
    Recency->>Binding: lane_binding_utc
    alt Binding log readable and rename is newer
        Binding-->>Recency: Binding timestamp
        Recency-->>Hook: Status 0
        Hook-->>Hook: Report newer rename, do not rename
    else Binding log cannot be read
        Binding-->>Recency: Read error
        Recency-->>Hook: Status 5
        Hook-->>Hook: Name unreadable log, do not rename
    end

Loading

File-Level Changes

Change	Details	Files
Fail closed when the register text cannot be read, instead of treating an empty read as a missing lane row.	Streams the cached register text through awk once before row evaluation and refuses with exit 2 when no content is rendered. Reports the unread register, preserves an accurate window/session/row diagnostic, and includes the safe-mode bypass. Removes the obsolete nameSince extraction from the guard path.	lanes-edit.sh tests/test_lane_helpers.sh
Distinguish an unreadable lane binding log from a binding that is absent or older/newer.	Extends guard_rename_is_newer with status 5 for unreadable logs while retaining statuses for newer and not-newer cases. Treats unexpected lane_binding_utc failures as unreadable rather than as evidence of rename recency.	lanes-edit.sh
Make SessionStart explicitly report unreadable binding logs without renaming or suppressing the hook.	Branches on the new status and names the affected lane log and unknown ordering. Keeps SessionStart exit status 0 and leaves the user's session name unchanged.	lanes-edit.sh tests/test_lane_helpers.sh
Add regression coverage for both indeterminate-read scenarios and their required side effects.	Tests empty and mode-unreadable register files for exit 2, accurate diagnostics, and absence of the misleading row message. Tests case-colliding published logs to verify SessionStart reports the unreadable log, does not claim recency, and performs no rename.	tests/test_lane_helpers.sh

Assessment against linked issues

Issue	Objective	Addressed	Explanation
#60	Detect when the register’s text cannot be read before consulting the name table, and refuse with a message that names the unreadable register read rather than claiming the lane has no row.	✅
#60	Preserve an unreadable lane binding log as a distinct state, so the hook neither claims a rename is newer than an unread binding nor silently treats it as not newer; name the unreadable log and leave the session name unchanged.	✅

Possibly linked issues

• The name guard's two remaining indeterminate reads pick a branch of the (b) table instead of naming the read: the register's TEXT and the lane's binding log #60: The PR directly fixes both reported defects: it refuses when register text is empty and reports unreadable lane binding logs.

---

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
0 New issues
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

[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 1 day and 9 hours by commenting @sourcery-ai review. Upgrade to get a review now.