[Birfy]

Owner: Dev3
Reviewer: Reviewer
Written-by: Dev3
Fixes #1077
Tip SHA: 7dcf6ca

Docs-not-needed: comment wording only, no documented behaviour changes (#1077)

Reproduce

plugins/social/src/face-rooms.js:689 said the digest was "two FNV-1a passes". Only the first pass is FNV-1a
(0x01000193); the second multiplies by 0x85ebca6b and then xor-shifts, which is Murmur3's finalizer mix.
Nothing pins the sentence, so it cannot go red — the evidence is the two constants in lineDigest (:694-706):
low = Math.imul(low ^ code, 0x01000193) and high = Math.imul(high ^ code, 0x85ebca6b) followed by
high ^ (high >>> 13).

What changed

One comment in plugins/social/src/face-rooms.js, on the #1062 bounded digest: it now names the FNV-1a prime
and the Murmur3-style finalizer constant. No code line changes, and the digest values are untouched — the second
pass only has to be a different mixing function, so changing the algorithm to match a stale comment would have
been the tail wagging the dog (and would rewrite seen digests already persisted on hosts).

Triage's scope note also named the same sentence in changelog.d/1062-a-face-event-key-carries-a-digest.changed.md.
That fragment never mentions FNV — it describes the bounded digest only — so there is nothing to rename there and
the file is untouched.

What I ran

• plugins/social node --import ./test/resolve.mjs --test test/face-rooms.test.js → 34/34, 0 skipped (the
  same count as base c653a5dc; the issue said 33/33, one case has landed since).
• diff -u against base c653a5dc: the only hunk is the comment block, 4 lines replaced by 4, no code line moved.

Red without the source change

Not applicable in the test sense: a comment has no assertion behind it, so removing it leaves every test green.
What I did instead is the check that matters for a rename — the diff contains comment lines only and the digest
tests are unchanged at 34/34, so no key value can have moved. If this wording drifts again it will stay invisible;
pinning it would mean asserting on a comment, which is not worth a test.

Not run

• oxlint / npm run verify — no binary for either on this box (same as my earlier PRs); the repo's CI
  plugins (social/tree/timeline) job covers lint and the full suite.
• arch-docs — needs git log, and there is no git here; the trailer above is in the commit message too, so the
  queue's squash commit carries it (check-arch-docs.mjs:107 reads git log base..HEAD --pretty=%B).

Risk

None to behaviour: no executable line changed, no test changed, no changelog fragment (comment-only, nothing a user
sees). The only risk is that a future reader trusts a comment that was wrong for one release; this PR removes
exactly that.

[Birfy]

Verdict: PASS at 7dcf6ca

Written-by: Reviewer

One comment, three lines, no code line touched — diff against main c653a5dc shows only plugins/social/src/face-rooms.js:689, so no digest value moves and the seen keys persisted in an owner's settings stay valid.

I checked the sentence against lineDigest (:695-706) rather than against the PR body: low starts at the FNV offset basis 0x811c9dc5 and does Math.imul(low ^ code, 0x01000193) — that is FNV-1a, prime named correctly. high starts at 0x9e3779b9 and does Math.imul(high ^ code, 0x85ebca6b) then high ^ (high >>> 13) — murmur3 fmix32's constant and shift, in a reduced form, so "Murmur3-style finalizer mix" is right and "two FNV-1a passes" was wrong. The rest of the sentence holds too: no time, no counter, only text.length and the char codes.

Ran:

• cd plugins/social && TZ=UTC node --import ./test/resolve.mjs --test test/face-rooms.test.js → 34/34.
• grep -rn FNV over the tip: the only remaining mentions are characters.js:46 (the Bot look's FNV-1a) and client.js:336 (the same, in the plugin's mirror) — neither is about this digest. No doc carries the old sentence either.
• changelog.d/1062-a-face-event-key-carries-a-digest.changed.md read: it describes the bounded digest and never names FNV, so leaving that file alone is correct, as the body says.

Not run: nothing behavioural is at stake here; no test pins a comment, so there is no knife to show. CI at this tip: tests-ok, plugins ×3, Node 24, package, merge-gate, agent-audience all success; enqueue-guard red only because no verdict named this tip until this comment.

Non-blocking: the two mixes are not two passes — both are updated in the same loop over the characters. "one FNV-1a pass and one Murmur3-style finalizer mix" reads as two loops to someone who has not opened the function. ...updated in the same loop would be exact, but this PR is about wording and I am not asking for a second round of it.