Skip to content

docs(adr): correct ADR 0024's stale build status and its no-401-handling claim - #381

Merged
wshallwshall merged 1 commit into
mainfrom
claude/b2-adr-0024
Aug 14, 2026
Merged

docs(adr): correct ADR 0024's stale build status and its no-401-handling claim#381
wshallwshall merged 1 commit into
mainfrom
claude/b2-adr-0024

Conversation

@wshallwshall

Copy link
Copy Markdown
Collaborator

Sixth ADR in the same accuracy sweep as #380, which merged before this one arrived -- so it lands as its own PR, rebased onto current main rather than stacked on the merged branch.

ADR 0024 (SMART Backend Services token provider) carried a stale build status and asserted the provider does not handle 401 responses. Both are false against the shipped code.

Scope

files      1, docs/adr/0024-smart-backend-services-token-provider.md
commits    1, rebased onto origin/main
.py, BACKLOG, other docs/   ZERO

No banner flip, deliberately

Dispatched as an accuracy fix rather than a numbered item, and authoring ledger content is not the builder's. Flagged explicitly because the same visible shape -- a PR touching no ledger file -- was a real defect on #372 earlier today and is correct compliance here. The PR body is the only thing that distinguishes them.

Verification

Docs-only, no code path touched; no tests run for that reason. Commit hooks passed. Same accuracy-fix class as the five in #380, where every audit row understated its own site count -- an audit row tells you where the reader looked, not where the claim lives.

…ing claim

Two accuracy defects, both understating shipped behaviour. Found while predicate-checking
this ADR's clean verdict and deliberately not acted on then, because they were outside
that pass; dispatched afterwards so they have an owner.

DEFECT 1 -- the 401 backstop. Section 4 stated in the present tense that _post has "no"
401-specific handling and dead-letters a 401 as a permanent NegativeAckError, then
described adding a branch as future work. Both branches shipped.

  THE INTERESTING PART IS WHY THIS SURVIVED A CHECK. The two branches do not share a
  spelling: the REST arm tests a `status` variable, the FHIR arm tests the exception's
  `code`. A grep for `exc.code == 401` returns 2 in fhir.py and ZERO in rest.py -- so a
  single-spelling search reads as "not shipped on REST" and would have confirmed the
  stale sentence. Caught only by printing every 401 site rather than counting one
  pattern. That warning is now written into the ADR so the next reader does not repeat it.

DEFECT 2 -- the status contradiction. The Status line says accepted, built and shipped,
while the section heading still read "Decision (proposed)" and a "To resolve on
acceptance" list stood open. A reader skimming headings concludes the decision is
unmade.

  Heading corrected. The to-resolve list is RETAINED rather than deleted, marked resolved,
  because the questions record what acceptance turned on -- that is history worth keeping,
  and deleting it would lose the reasoning while fixing the tense.

SITE COUNT CHECKED BEFORE EDITING, per the pattern in the previous five: exactly two
proposed/unbuilt markers in this file, both fixed. Every one of those five understated its
own count, so counting first is now the default rather than a precaution.

Verified: no ledger file touched, no engine code touched. cp1252 encodability run against
the ADDED lines only -- 15 added lines, ZERO non-ASCII.

  A ZERO SEEN-COUNT IS NOT SELF-EVIDENCING, so the scanner was separately proved to
  discriminate against a planted corpus: an em dash (safe punctuation) and an arrow
  (U+2192, unsafe) in the same three lines returned seen=2, unsafe=1 at 0x2192. The
  control also demonstrated the rule's own rationale by crashing this console with
  UnicodeEncodeError when it first tried to print the offending character.

No banner flip and no BACKLOG citation: dispatched as an accuracy fix, not a numbered
item, and authoring ledger content is not the builder's.
@wshallwshall
wshallwshall enabled auto-merge (squash) August 14, 2026 00:27
@wshallwshall
wshallwshall merged commit f7fe42f into main Aug 14, 2026
33 of 34 checks passed
@wshallwshall
wshallwshall deleted the claude/b2-adr-0024 branch August 14, 2026 00:30
Sign up for free to 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