Repository navigation
check_pr_ont_delta.sh grades an event-SNAPSHOT body against a LIVE file list, so the documented fix can never be applied without a push #3298
Description
Found while unblocking PR #3093 (PVL-1, PMAT-1099).
The two sources disagree by construction
.github/workflows/ci.yml (lines 1245-1255) hands the guard:
env:
PR_BODY: ${{ github.event.pull_request.body }} # EVENT SNAPSHOT
run: |
printf "%s" "$PR_BODY" > "$RUNNER_TEMP/ont-body.txt"
gh api "repos/${GITHUB_REPOSITORY}/pulls/${PR_NUMBER}/files" --paginate \
--jq ".[].filename" > "$RUNNER_TEMP/ont-files.txt" # LIVEThe body is frozen at the moment the event was minted; the changed-file list is fetched live at job
time. The guard then decides "is this a sweep PR" from the live list and "did it close with a delta"
from the stale body.
Why that matters, measured
The guard failed PR #3093 with FAIL no ont-delta: line. The documented remedy is to put the line
in the PR body, so:
gh pr edit 3093 --body-file …— line present, read back from the API: 1 occurrence.gh run rerun 34936163911 --failed— guard FAILED again, and the job log printed the old
body verbatim, ending in the paragraph I had just deleted.- Only
git push(an empty commit was enough) minted an event whose payload carried the new body.
A rerun replays the original payload, so a body edit can never be re-graded by a rerun. The one
action the guard tells you to take is the one action that does not work. Cost here: one full CI
cycle (~13 min) plus a commit that exists only to move an event payload.
The same pattern is at line 1214 (check_pr_closes_issue.sh), so this is a class, not one step.
Not a security trade-off here
For pull_request_target the snapshot is the safe choice. These two steps run under
pull_request, where the body is equally untrusted either way and is only ever written to a file
and grepped. Reading it with gh api repos/$GITHUB_REPOSITORY/pulls/$PR_NUMBER --jq .body, exactly
as the file list is already read, makes both inputs come from the same place at the same time.
Suggested shape
Read both from the API in the same step, and keep PR_BODY only as the fallback when
github.event_name is merge_group (no PR to ask about) — where the guard already REPORTs rather
than judges.
Not fixed in #3093: touching ci.yml again would widen that PR and lapse its 3/3 quorum.
ont-delta: none — this issue adds no entity type, shape, resolves rule or reason id; it reports a wiring defect in a guard.
Activity
noahgift commentedon Sep 23, 2026
triage 2026-09-23 (aprender-dd, standing triage owner): milestoned to 0.70.0, theme Fast Train (CI, gates, pv, release tooling, hygiene); it moved out of
backlog(the 2026-09-17 overflow bucket, not a release).noahgift commentedon Sep 23, 2026
triage 2026-09-23 (aprender-dd): kept in 0.70.0 and labelled ont-0.70. Operator, verbatim (relayed by the cop aprender-cf): "the ontology spec MUST be finished in .7". Classified as ontology/SHACL spec work: its title names ONT/ontology/SHACL/capability cells, or its body cites an ONT- row.
ci: PR-body guards read the LIVE body, not the event snapshot (#3298)
3 remaining items
merge origin/main into fold/pr-review-ci
noahgift commentedon Oct 14, 2026
P0 to P1 (2026-10-14 P0 triage): P0 now means it blocks the next release or carries the operator's own words on it; this issue does neither, so it stays on the 0.71.0 milestone at P1. The operator quote in its history is a milestone-wide triage note, not words about this issue.
noahgift commentedon Oct 14, 2026
Re-parent: former parent #3569 is closed. Moved under open epic #3269 (same 0.71 train).
noahgift commentedon Oct 14, 2026
Folded into #5308 as a checklist row (TRIAGE-100, 2026-10-14): the open-ticket cap is 100, so this ticket is now tracked on that master and closed as not planned. Nothing is lost: reopen it when its work resumes, and an open PR for it reopens it. Pushed branch:
fix/3298-pr-body-live.Agent: aprender-clerk
ci(pr-body-guard): drop the edited-trigger workflow; retry only trans…