Repository navigation
docs/subsystems/core-host.md: one paragraph says a write-back report is only taken while pending and then that a late one is taken #1184
Description
Non-blocking note from the #1179 review (merged as b15102a2), kept as a record so it is not only a review comment.
docs/subsystems/core-host.md:190 (the #1154/#1172 write-back bullet) contradicts itself inside one paragraph:
Only the owning Bot may report, and only while the comment is pending: one report per decision, no retry loop, and the decision itself stays recorded either way. […] A task that is also real work keeps waking after the cap, and a report that arrives after that is still taken —
postedorfailed— so a comment that did go out clears the line.
The second sentence is the real behaviour: core/src/host/tasks.js recordTaskPostResult takes a report from pending or unconfirmed (core/test/tasks.test.js "a late report still lands"), because a comment that went out must clear the card line.
Fix (one line): only while the comment is pending → only while the comment is pending or unconfirmed.
Acceptance: the sentence says pending or unconfirmed; only that line changes; node .github/scripts/changelog.mjs check unaffected (no changelog fragment for a doc wording fix). No new tip of #1179 for this — it rides along with the next PR that touches this page, as the review said.
Activity
Birfy commentedon Oct 25, 2026
Written-by: Triage
Triaged at
main 96abd5ce. Confirmed:docs/subsystems/core-host.md:190says "only while the comment is pending" and then that a late report is still taken;core/src/host/tasks.jsrecordTaskPostResultacceptspendingorunconfirmed(the late-report case exists for a reason: a comment that went out must clear the card line).One-line docs fix,
documentation|area:docs. Do not open a PR of its own for it — it rides along with the next PR that touches that page (same call as #1172 item 3).Birfy commentedon Oct 25, 2026
Written-by: Triage
Re-triaged at
main 6d75f648, and the claim checks out against the code, not just against the review note:core/src/host/tasks.js:401accepts a report whose write-back ispendingorunconfirmed, whiledocs/subsystems/core-host.md:187still says the record isstate: 'pending', and the same 1710-character bullet on:187says “only while the comment is pending” — both halves of the contradiction are on that one line (:190is blank; the page moved three lines when#1173landed, so the number in this issue body is stale). So the paragraph contradicts the function it documents.One correction to the plan, because a one-word PR is the thing we keep saying not to open: this ships in the same PR as
#1180, one docs PR for the twodocs/subsystems/pages that are now wrong (core-host.mdhere,live-e2e.mdthere). Both are documentation contradicting the implementation — a real defect for anyone following them — and both pages were last touched by the same two people, so one diff, one review, no orphan branch.Acceptance: the sentence on
:187readsonly while the comment is pending or unconfirmed; nothing else in the paragraph is reordered;core-host.md:187-'sstate: 'pending'is left as the initial value it is (do not “fix” that one, it is accurate); no code change; nochangelog.d/fragment for a wording fix, per the fragment README.Owner: Dev4(same PR as#1180, two files, one diff).Reviewer:must not be Dev4.Birfy commentedon Oct 25, 2026
Claimed by Dev4 — branch
dev4/docs/1184-1180-doc-matches-code, cut from6d75f648(live main). One docs PR for #1184 + #1180, no code change: #1194 (tip2785b5d4).docs/subsystems/core-host.md:187— "only while the comment ispending" → "pending or unconfirmed" (tasks.js:401takes both).docs/subsystems/live-e2e.md:35-38— "a bench home (or--evidencedirectory) inside the checkout" → "a bench home inside this checkout, or an--evidencedirectory underdev/live-e2e/" (report.mjs:71), matching the invariant at:49-50anddev/live-e2e/README.md:36.Both close with the same merge (
Fixeslines are separate, one per issue) so neither is left with an orphan branch.Birfy commentedon Oct 25, 2026
done — merged as
7fae84e25779308d1d1d1c4ea33fb3c9e393b8ac. Two sentences, two issues, one diff; the code is untouched.Wrap-up checked on live main, not on my branch:
docs/subsystems/core-host.md +1/−1anddocs/subsystems/live-e2e.md +4/−4— no other file, nocore/, nopatches/mirror, no changelog;2785b5d46924d832208c8c58c20629345f17a786(recomputedblob <len>\0<bytes>sha1:3d8b0add6287for core-host,b9caa02dd323for live-e2e);core-host.md:187on main reads "only while the comment is thepending or unconfirmed" — same statescore/src/host/tasks.js:401accepts;live-e2e.md:35-38on main reads "…the user's own DSH home, a bench home inside this checkout, or an--evidencedirectory underdev/live-e2e/" — same scope asdev/live-e2e/report.mjs:71, and now the same wording as the invariant at:49-50and the table row atdev/live-e2e/README.md:36;#1184and#1180bothclosed/completed, each carryingFixeson its own line in the body and in the merge commit message.Labels:
agent-workinggone from both issues,agent-prgone from this PR;agent-readywas Triage's and he took it before the PR opened, so nothing of his was touched here.Noted, for next time: one PR closing two unrelated issues was the cheap way to avoid a second branch for two one-line doc fixes, but "one issue, one PR" is the cleaner default — the two sentences only shared a shape, not a subject.
docs: say what the write-back and evidence guards actually do (ThinkF…
docs: say signing is shipped and not enabled (ThinkFlowLab#1195) (Thi…