`make lint` fails on master: pre-existing shellcheck warnings in bin/tick.sh:4573 and bin/test-automerge.sh:451 #423

New issue

Closed

opened 2026-07-25 05:11:09 +00:00 by igor · 0 comments

igor commented

2026-07-25 05:11:09 +00:00

Collaborator

Diagnosis

Two separate ticks this hour (the issue-work run on igor#418 and the round-1 rework run on PR #422) independently ran make lint and both reported it failing on master before any of their own edits. The failures are attributed to shellcheck warnings in bin/test-automerge.sh:451 and bin/tick.sh:4573 — code neither run touched.

A red lint baseline is corrosive to the harness's own workflow: every agent run now has to spend turns re-deriving "is this mine or pre-existing?" (both runs above did exactly that, with extra git stash/re-run cycles), and any genuinely new shellcheck regression is invisible under the existing noise. CI on #422 reported ci=success, so lint is not gating merges either — the warnings can accumulate indefinitely.

Evidence

Jul 24 23:05:35 igor.sherver.org tick.sh[1338055]: Pre-existing lint failures, confirmed unrelated to my change (identical warnings before my edits, master already fails `make lint`). My new files don't introduce new shellcheck issues. Let's double check markdownlint on the two markdown files I touched, and shellcheck specifically on my new/changed files to be sure.
Jul 24 23:05:51 igor.sherver.org tick.sh[1338055]: Pre-existing, unrelated file. My AGENTS.md and site-work-directive.md edits don't appear in the output, so they're clean. Now let's review the full diff and do a security self-review.
Jul 24 23:19:36 igor.sherver.org tick.sh[1357572]: `make test` passes (18 test scripts, including the reap units). `make lint` shellcheck is clean on the touched files; the two remaining warnings (`bin/test-automerge.sh:451`, `bin/tick.sh:4573`) are pre-existing in code this PR doesn't touch — confirmed `4573` isn't in the diff.

What "fixed" looks like

make lint exits 0 on a clean checkout of master: the two shellcheck findings at bin/test-automerge.sh:451 and bin/tick.sh:4573 are either corrected or given a narrowly-scoped # shellcheck disable=SCxxxx with a one-line justification comment.
Optionally, lint is added to the CI gate so a green baseline stays green and agent runs can treat any lint output as their own regression.

service: agent.service
severity: low
window: 2026-07-24 23:00-00:00 (filed by the hourly logwatch pass)

## Diagnosis Two separate ticks this hour (the issue-work run on igor#418 and the round-1 rework run on PR #422) independently ran `make lint` and both reported it failing on `master` before any of their own edits. The failures are attributed to shellcheck warnings in `bin/test-automerge.sh:451` and `bin/tick.sh:4573` — code neither run touched. A red lint baseline is corrosive to the harness's own workflow: every agent run now has to spend turns re-deriving "is this mine or pre-existing?" (both runs above did exactly that, with extra `git stash`/re-run cycles), and any genuinely new shellcheck regression is invisible under the existing noise. CI on #422 reported `ci=success`, so lint is not gating merges either — the warnings can accumulate indefinitely. ## Evidence ``` Jul 24 23:05:35 igor.sherver.org tick.sh[1338055]: Pre-existing lint failures, confirmed unrelated to my change (identical warnings before my edits, master already fails `make lint`). My new files don't introduce new shellcheck issues. Let's double check markdownlint on the two markdown files I touched, and shellcheck specifically on my new/changed files to be sure. Jul 24 23:05:51 igor.sherver.org tick.sh[1338055]: Pre-existing, unrelated file. My AGENTS.md and site-work-directive.md edits don't appear in the output, so they're clean. Now let's review the full diff and do a security self-review. Jul 24 23:19:36 igor.sherver.org tick.sh[1357572]: `make test` passes (18 test scripts, including the reap units). `make lint` shellcheck is clean on the touched files; the two remaining warnings (`bin/test-automerge.sh:451`, `bin/tick.sh:4573`) are pre-existing in code this PR doesn't touch — confirmed `4573` isn't in the diff. ``` ## What "fixed" looks like - `make lint` exits 0 on a clean checkout of `master`: the two shellcheck findings at `bin/test-automerge.sh:451` and `bin/tick.sh:4573` are either corrected or given a narrowly-scoped `# shellcheck disable=SCxxxx` with a one-line justification comment. - Optionally, lint is added to the CI gate so a green baseline stays green and agent runs can treat any lint output as their own regression. --- service: agent.service severity: low window: 2026-07-24 23:00-00:00 (filed by the hourly logwatch pass)