Skip to content

Review follow-ups for 0.10.1: a late schema defect is -32603, the missing tests, the dead clause, the pattern cost - #79

Open
HackTuah wants to merge 1 commit into
mainfrom
fix/review-follow-ups
Open

HackTuah wants to merge 1 commit into
mainfrom
fix/review-follow-ups

Conversation

@HackTuah

Copy link
Copy Markdown
Member

Sean's non-blocking points from the 0.10.1 review, one commit:

  • A tools/call against a schema the catalog made unenforceable after startup is now -32603 naming the tool and keyword, not an "invalid arguments" tool result (the test fails on 0.10.1's server).
  • Tests: additionalProperties as a schema (kills the surviving mutant), list-form type refusal, every README keyword accepted well-formed.
  • The unreachable when key in @enforced clause and its list removed.
  • Threat model: one sentence in the arguments row on pattern cost (inside the row: no threat or status moves, so the 1.0.0 clock is unaffected).
  • CHANGELOG: the change, plus 0.10.1's nil reason now reading null.

Left for later, as the review said: compiling the schema once at startup (it touches ToolSpec, a public struct). Local gate: sixteen steps pass.

…sing tests, the dead clause, the pattern cost

From Sean MacGuire's review of 0.10.1. A tools/call against a schema the catalog made
unenforceable after startup is answered -32603 naming the tool and the keyword (it was a
tool result saying "invalid arguments", blaming the client); the new test fails on 0.10.1's
server. Tests for additionalProperties as a schema (it kills the review's surviving mutant,
the branch replaced by :ok), a list-form type refusing a value of none of its types, and every
README keyword accepted well-formed; the `when key in @enforced` clause, which no keyword could
reach, and the list it read are removed. The threat model's arguments row says what `pattern`
costs (PCRE's match limit, fail closed; 80 ms measured for the worst case a lane tried), inside
the row so no threat or status moves. CHANGELOG: the change, and 0.10.1's `nil` reason
reading `null`, recorded late. The stdio test helper says why its device is latin1.
Gate: sixteen steps pass.

Signed-off-by: Ayla Croft <aylacroft@proton.me>
@HackTuah
HackTuah requested a review from minitru September 26, 2026 19:28
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