Adopt distinct ported:/parity: provenance markers in both repos

[jeremydmiller]

Part of #5118 (compliance wave growth), epic #5110. Process issue — no test code.

Trimmed 2026-08-04. Most of the original scope shipped in #620 and the epic edits. What remains is item 1 below. The completed parts are kept at the bottom so nobody re-does them.

What is left

Adopt two distinct provenance markers across both repos — ported: for a file that duplicates a test belonging in the compliance library and is awaiting absorption, and parity: (or a bare marten#NNNN / polecat#NNNN) for a file that merely cites the issue its feature tracks.

The two meanings are conflated today, and that is not cosmetic — it is what made the original exit criterion unmeasurable. polecat#399 re-measured and found the marker grep counted 16 files where only 3 were real ports; the other 13 were provenance citations. Worse, the count grows as parity work succeeds: Events/binary_event_serialization_tests.cs joined the list the day polecat#402 merged, purely by citing marten#4515.

Apply as files are touched, not as a bulk sweep — a mechanical rewrite would churn both repos and lose the human judgement about which marker each file actually deserves.

Already done — do not redo

• ✅ Exit criterion restated on Epic: cross-store event sourcing compliance test library (test harness standardization Phase 3) #5110 and Master: test harness standardization → cross-store event sourcing compliance library #5119: no longer "no file carries a ported-from-Marten marker", now the child-suite checklist on Epic: cross-store event sourcing compliance test library (test harness standardization Phase 3) #5110.
• ✅ Coverage inventory in JasperFx.Events.ComplianceTests/README.md — the table of behavior areas the library covers, suite by suite.
• ✅ Out-of-scope boundary documented in the same README: storage layout and DDL, partitioning, node distribution / HotCold, high-water detection internals, and the whole document-db side (LINQ, patching, session semantics). Previously a reader had to infer the boundary from suite names, and every wave re-litigated "should this be shared?".
• ✅ "New cross-store behavior lands as a compliance suite first" written down in the README rather than living only in issue bodies.
• ✅ Local dev loop documented (-p:ComplianceSourceDir=...), and the switch added to Marten to match the one Polecat already had.

All of the above shipped in #620 (JasperFx 2.39.1).

Still-blocked port, for the record

Polecat.Tests/Linq/additional_linq_operator_tests.cs (19 tests, explicitly "ported from Marten") stays put until a document-db compliance library exists. LINQ is not event sourcing, and Polecat re-declares IDocumentStore/IDocumentSession/IQuerySession in its own namespace with no shared contract. That convergence question settles in Weasel.Storage first and is tracked as a later phase on #5119, not here.

🤖 Generated with Claude Code

[jeremydmiller]

Approved (Jeremy, this session)

Adopting ported: / parity: as the two provenance markers across both repos.

Per the issue's own guidance this is applied as files are touched, not as a bulk sweep — a mechanical rewrite would churn both repos and throw away the human judgement about which marker each file actually deserves. So this stays open as a standing convention rather than a unit of work, and closes when the remaining ported: files have been absorbed.

Recording the distinction once more so it is unambiguous at the point of use:

• ported: — this file duplicates a test that belongs in JasperFx.Events.ComplianceTests and is awaiting absorption. Counts toward the epic's remaining work.
• parity: (or a bare marten#NNNN / polecat#NNNN) — this file merely cites the issue whose feature it tracks. Provenance, not duplication. Does not count, and grows as parity work succeeds — which is precisely what made the old single-marker count move away from zero as the program went well.

[jeremydmiller]

Status: effectively blocked, not an open convention question

The ported: / parity: distinction this issue established is settled and in use. What is left is the closing condition — "closes when the remaining ported: files have been absorbed" — and that is now down to a single file that this program cannot absorb.

Current inventory in Polecat:

	Count
ported: / "ported from Marten"	1
marten#NNNN parity provenance	17

The one remaining port is Polecat.Tests/Linq/additional_linq_operator_tests.cs (19 tests). It is document-db, not event sourcing, so no amount of further work on JasperFx.Events.ComplianceTests will retire it. It is gated on the Weasel.Storage IDocumentStore convergence — Polecat re-declares IDocumentStore/IQuerySession rather than sharing them, which is the same blocker the deferred document-db compliance phase sits behind (#5119, and the W2 extraction epic #4821).

So this issue is not waiting on judgement or on a convention decision; it is waiting on that architectural work, and will close with it.

The distinction itself, restated once more so it stays unambiguous at the point of use:

• ported: — this file duplicates a test that belongs in a compliance library and is awaiting absorption. Counts toward remaining work.
• parity: (or a bare marten#NNNN / polecat#NNNN) — provenance for the issue whose feature the file tracks. Not duplication, does not count, and grows as parity work succeeds — which is precisely what made the original single-marker count move away from zero as the program went well.

Leaving open, retitled in effect as "blocked on the document-db convergence" rather than standing.

[jeremydmiller]

Closing — two decisions, taken together.

1. LINQ comes out of compliance scope permanently

Not "deferred until the document-db contract converges" — out, because the store query languages diverge structurally enough that a shared suite would be pinning coincidence rather than contract.

This restores the original day-one non-goal rather than inventing a new position. The Phase 3 design doc listed it explicitly:

No LINQ/query-provider compliance (store query languages diverge structurally).

Somewhere between then and now that hardened into "blocked pending the Weasel.Storage IDocumentStore convergence," which quietly reclassified a permanent exclusion as pending work — and that reclassification is what kept Polecat.Tests/Linq/additional_linq_operator_tests.cs (19 tests) on the books as an unabsorbed port. It is not an unabsorbed port. It is a Polecat-owned test file, and it stays one.

That also removes the last item this issue described as a "still-blocked port," which is what made the remaining scope hollow.

2. The marker convention is not an issue-shaped thing

What was left here after the 2026-08-04 trim was one item: adopt ported: vs parity: markers across both repos, applied as files are touched, not as a bulk sweep. That instruction is correct and is also why this cannot close on its own terms — there is no finish line, no owner, and no trigger. It would sit open indefinitely looking like unfinished work while the thing it was actually meant to enable moved elsewhere.

The measurement it existed to support already moved: the exit criterion is the child-suite checklist on #5110, and that criterion has been met — the ES compliance backlog is empty as of wave 8 (JasperFx 2.45.0, #5205), 28 suites / 230 tests on both stores with zero capability gates.

The underlying insight is worth keeping, so restating it here rather than in a tracker: a marker grep is not an exit criterion, because provenance citations grow as parity work succeeds. polecat#399 measured 16 marked files where only 3 were real ports, and Events/binary_event_serialization_tests.cs joined the list the day polecat#402 merged purely by citing marten#4515. A count that moves away from zero exactly when you are winning is measuring the wrong thing.

Everything else originally scoped here shipped in #620 (JasperFx 2.39.1): the restated exit criterion, the coverage inventory, the out-of-scope boundary, the "new cross-store behavior lands as a compliance suite first" rule, and the -p:ComplianceSourceDir= dev loop.

Follow-up

The library README's out-of-scope section still hedges — it lists the document-db side as out "because no shared document store contract exists yet," which is the deferral framing decision 1 rejects. That wants a one-line change in jasperfx to say LINQ is out on its merits. Noted rather than done here, since it is a different repo.