Skip to content

policy: allow self sign-off on functional review (retire D30) - #119

Merged
LukasWodka merged 1 commit into
developfrom
chore/relax-d30-self-signoff
Aug 1, 2026
Merged

policy: allow self sign-off on functional review (retire D30)#119
LukasWodka merged 1 commit into
developfrom
chore/relax-d30-self-signoff

Conversation

@LukasWodka

@LukasWodkaLukasWodka commented Aug 1, 2026

Copy link
Copy Markdown
Contributor

Flips allow-self-signoff to default true in the reusable, which applies org-wide in one change since the per-repo callers pass no inputs.

What changes

Functional review still requires a human to assert it. It no longer has to be a different human from the one who opened the PR.

Why

Code review already puts a second person on the diff. Most implementation is now AI-assisted, so requiring a separate human for the functional review was ceremony rather than a control.

It was also actively counterproductive on 2026-08-01: 3 of 17 /fr-pass comments were refused as self-signoff, and the cards were then advanced by hand through the project API. The control did not hold — it was routed around. A rule that gets bypassed under normal load is worse than no rule, because it produces the appearance of a gate.

Scope

  • Unchanged: authorisation still requires repo write/maintain/admin.
  • Unchanged: release-train/* and hotfix-backmerge/* promotion PRs were already exempt.
  • Unchanged: fr-gate still blocks prod promotion until every contained item is Ready for prod.
  • A repo wanting the stricter rule sets allow-self-signoff: false on its caller.

Decision recorded in RFC-BACKEND-1405 (tracebloc/rfcs#1). Related: tracebloc/backend#1406, tracebloc/backend#1408.


Note

Medium Risk
This weakens a process gate org-wide by default; write-access and release-train promotion exemptions are unchanged, but bypassing separate FR sign-off increases reliance on code review alone.

Overview
Org-wide functional review policy changes via the reusable fr-pass-comment.yml workflow: allow-self-signoff now defaults to true (was false), so repos that call the workflow without overriding inputs let the PR/issue author use /fr-pass on their own card.

Functional review still needs a human with repo write access to assert it; only the “must be someone other than the author” (D30) rule is retired by default. Callers can set allow-self-signoff: false to keep the stricter two-pairs-of-eyes behavior for one repo. The input description is updated to document RFC-BACKEND-1405 D6 and the rationale.

Reviewed by Cursor Bugbot for commit a91e805. Bugbot is set up for automated code reviews on this repo. Configure here.

Functional review still requires a human to assert it; it no longer has to be
a different human from the author.
Rationale (RFC-BACKEND-1405 D6): code review already puts a second person on
the diff, and most implementation is AI-assisted, so requiring a *separate*
human for the functional review was ceremony rather than a control. At our
size it was also a throughput ceiling — on 2026-08-01, 3 of 17 /fr-pass
comments were refused as self-signoff and the cards had to be moved by hand,
which is strictly worse than allowing the sign-off: the control was bypassed
rather than satisfied.
Flipping the reusable's default applies this org-wide in one change, because
the per-repo callers pass no inputs. A repo that wants the stricter rule can
set allow-self-signoff: false on its caller.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@LukasWodkaLukasWodka self-assigned this Aug 1, 2026
@LukasWodka
LukasWodka merged commit dee9920 into developAug 1, 2026
5 checks passed
@LukasWodka
LukasWodka deleted the chore/relax-d30-self-signoff branch August 1, 2026 21:45
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

@LukasWodka