Summary
test/ri-zero-connector-knowledge-conformance.test.ts's scanRepository() invariant ("RI production code contains zero connector/provider-specific executable knowledge") currently fails with 8 findings across 5 files. All 8 are confirmed pre-existing (predate PR #55's branch entirely — traced back through git log to Move B-era commits) and unrelated to the @pdpp/polyfill-connectors re-vendoring/resolver work in #55. Verified each is a false positive on legitimate code, not a real defect:
| File |
Line |
Rule |
Verdict |
scripts/stream-health-audit/authority.ts |
1812, 1897 |
hardcoded-provider-endpoint-url |
False positive — "https://pdpp.invalid" is the RFC 6761/IANA reserved .invalid placeholder TLD, used deliberately as a dummy base for new URL(relativeHref, base) parsing. Not a real provider endpoint. |
scripts/test-accounting/inventory.ts |
280 |
hardcoded-connector-identity-literal |
False positive — "signal" here is a NodeJS.Process-style exit-signal field name in a receipt-schema array (alongside exit_code, counts, etc.), not a reference to the signal messaging connector. |
scripts/test-accounting/inventory.ts |
719 |
unresolvable-data-resource-load |
False positive — reads a receipt/completion file by a caller-supplied, already-validated path; not connector-specific. |
scripts/test-accounting/packet.ts |
267, 580, 637 |
unresolvable-data-resource-load |
False positive — reads test-accounting manifest/lease/packet files by internally-constructed safe paths (safePath/safeLeasePath helpers). |
scripts/test-accounting/with-local-full-suite-lock.mjs |
44 |
unresolvable-data-resource-load |
False positive — reads a lock-owner file by a constructed path. |
server/polyfill-manifest-reconcile.ts |
99 |
unresolvable-data-resource-load |
False positive — reads a manifest JSON file by caller-supplied path; the whole function's purpose is manifest reconciliation, this read is its core job. |
Why filed separately from #55's connector-tree-scope work
These 8 findings are unrelated to data-connectors#75's resolver work — none of the 5 files touch @pdpp/polyfill-connectors at all, and all predate PR #55's branch (confirmed via git log on each file). They happened to share a test NAME with genuinely connector-tree-scope-related failures earlier in this PR's history, which is why they weren't separated out sooner — corrected here now that the connector-tree-scope fix (via #75) resolved the OTHER violations this same test used to report (a runtime/controller.ts finding is now gone from this test's output, confirming that part of the fix worked).
Suggested fix
The scanner (test/helpers/ri-zero-connector-knowledge-scan.ts per the test file's own references) needs either:
- A recognized allowlist/exemption for the
.invalid TLD placeholder pattern (rule 6/hardcoded-provider-endpoint-url), or
- Context-aware matching for
hardcoded-connector-identity-literal (rule 7) that doesn't flag arbitrary string literals matching a connector name without checking surrounding context (e.g., array/object key names suggesting a non-connector meaning), or
- A narrower
unresolvable-data-resource-load (rule 5) heuristic that recognizes readFileSync/readFile calls guarded by internal path-construction helpers (safePath, safeLeasePath) as resolvable/safe, not unresolvable.
Not attempting this fix as part of #55 — it's a scanner correctness question, not something this PR's re-vendor/resolver-swap work should absorb.
Current impact
2 of test reference implementation's remaining failures in PR #55 (RI production code contains zero... and its "...after the universal per-node rewrite" duplicate — both scan the same repo, so they share this same 8-finding root cause).
Summary
test/ri-zero-connector-knowledge-conformance.test.ts'sscanRepository()invariant ("RI production code contains zero connector/provider-specific executable knowledge") currently fails with 8 findings across 5 files. All 8 are confirmed pre-existing (predate PR #55's branch entirely — traced back throughgit logto Move B-era commits) and unrelated to the@pdpp/polyfill-connectorsre-vendoring/resolver work in #55. Verified each is a false positive on legitimate code, not a real defect:scripts/stream-health-audit/authority.tshardcoded-provider-endpoint-url"https://pdpp.invalid"is the RFC 6761/IANA reserved.invalidplaceholder TLD, used deliberately as a dummy base fornew URL(relativeHref, base)parsing. Not a real provider endpoint.scripts/test-accounting/inventory.tshardcoded-connector-identity-literal"signal"here is aNodeJS.Process-style exit-signal field name in a receipt-schema array (alongsideexit_code,counts, etc.), not a reference to thesignalmessaging connector.scripts/test-accounting/inventory.tsunresolvable-data-resource-loadscripts/test-accounting/packet.tsunresolvable-data-resource-loadsafePath/safeLeasePathhelpers).scripts/test-accounting/with-local-full-suite-lock.mjsunresolvable-data-resource-loadserver/polyfill-manifest-reconcile.tsunresolvable-data-resource-loadWhy filed separately from #55's connector-tree-scope work
These 8 findings are unrelated to
data-connectors#75's resolver work — none of the 5 files touch@pdpp/polyfill-connectorsat all, and all predate PR #55's branch (confirmed viagit logon each file). They happened to share a test NAME with genuinely connector-tree-scope-related failures earlier in this PR's history, which is why they weren't separated out sooner — corrected here now that the connector-tree-scope fix (via#75) resolved the OTHER violations this same test used to report (aruntime/controller.tsfinding is now gone from this test's output, confirming that part of the fix worked).Suggested fix
The scanner (
test/helpers/ri-zero-connector-knowledge-scan.tsper the test file's own references) needs either:.invalidTLD placeholder pattern (rule 6/hardcoded-provider-endpoint-url), orhardcoded-connector-identity-literal(rule 7) that doesn't flag arbitrary string literals matching a connector name without checking surrounding context (e.g., array/object key names suggesting a non-connector meaning), orunresolvable-data-resource-load(rule 5) heuristic that recognizesreadFileSync/readFilecalls guarded by internal path-construction helpers (safePath,safeLeasePath) as resolvable/safe, not unresolvable.Not attempting this fix as part of #55 — it's a scanner correctness question, not something this PR's re-vendor/resolver-swap work should absorb.
Current impact
2 of
test reference implementation's remaining failures in PR #55 (RI production code contains zero...and its "...after the universal per-node rewrite" duplicate — both scan the same repo, so they share this same 8-finding root cause).