You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Neither caused the other: the two PRs were open at the same time, each picked the next free number when it was written, and both were correct at the time. They collided on merge and nothing caught it — there is no gate on ADR numbering, which is why this is being filed rather than fixed in passing.
It matters more than a filename usually would, because this repository cites ADRs by number in prose and in doc comments. "ADR-0018" is now ambiguous in every one of those places.
Which one gets renumbered is a maintainer decision, and it is not obvious:
Both are already on master, so either rename breaks existing links until they are chased down.
0018-privacy-controls-and-evidence.md is the newer merge but the broader subject.
Whoever takes it should also grep -rn "ADR-0018\|adr/0018" and fix the citations in the same commit, and it is worth considering whether verify-links or a small xtask check should refuse a duplicate ADR number — that gate does not exist today, and this is the failure it would have caught.
Found while merging #168 and #169, which are stacked on #166 and are unaffected in substance.
docs/adr/currently holds two ADR-0018s:0018-cross-tenant-admin-reads.md— the/dash/v1cross-tenant admin read grant (feat(dash): a cross-tenant admin read grant, decided in ADR-0018 #166)0018-privacy-controls-and-evidence.md— the GDPR control architecture (docs: define GDPR control architecture #167)Neither caused the other: the two PRs were open at the same time, each picked the next free number when it was written, and both were correct at the time. They collided on merge and nothing caught it — there is no gate on ADR numbering, which is why this is being filed rather than fixed in passing.
It matters more than a filename usually would, because this repository cites ADRs by number in prose and in doc comments. "ADR-0018" is now ambiguous in every one of those places.
Which one gets renumbered is a maintainer decision, and it is not obvious:
master, so either rename breaks existing links until they are chased down.0018-cross-tenant-admin-reads.mdis referenced by the credential work in flight (feat(db): split credentials out of StaffMember, built for OIDC and SSO (ADR-0019) #168's ADR-0019 cites it, as doesdocs/plans/2026-09-13-dashboard-nav-notes/credentials.md).0018-privacy-controls-and-evidence.mdis the newer merge but the broader subject.Whoever takes it should also
grep -rn "ADR-0018\|adr/0018"and fix the citations in the same commit, and it is worth considering whetherverify-linksor a smallxtaskcheck should refuse a duplicate ADR number — that gate does not exist today, and this is the failure it would have caught.Found while merging #168 and #169, which are stacked on #166 and are unaffected in substance.
🤖 Generated with Claude Code