# End-to-end examples and provenance

Run commands from the module root. Every bundle's XML paths resolve relative to that bundle.

## Evidence integrity workflow

The README's detach/restore-pack/verify-pack commands use the same unmodified MIT Jest fixture. Run detach with a new output directory; verify-pack exits 1 and restores the identical archive. E2E coverage creates new temporary packs, refuses existing destinations, extra members, byte changes, missing accepted members and directories masquerading as XML, then restores successfully after correction. A pack of the missing-bundle example still verifies as incomplete (3). All pack code/test/index metadata is newly authored Apache-2.0; member XML retains its upstream license. The output directory is not committed.

`moon run scripts/render-demo.mbtx wasm` also writes `output/jest-integrity.json` as UTF-8 from the unmodified Jest archive. Run `moon run --target wasm cmd/moontestledger -- check-integrity output/jest-evidence.json output/jest-integrity.json`: expected exit **1** with the unchanged six-case/four-failure summary. Integrity success preserves the failed test gate. Supply `native` to both commands for native execution with a working compiler.

Keep the expected manifest in a separately trusted location before transferring evidence. The E2E evaluator verifies the receipt SHA-256 against the independently recorded upstream file hash below, accepts archive JSON reformatting, changes only original XML whitespace to obtain `IntegrityMismatch` (2), checks incomplete evidence still exits 3, preserves denied recovery (1) and successful recovery (0), and recovers after a malformed manifest. No altered upstream fixture is committed. These tests and manifest workflow are project-authored Apache-2.0; the original XML retains its MIT license. Checksums authenticate neither the creator nor the execution environment.

## Unmodified upstream reports

`upstream/collection.json` is a project-authored Apache-2.0 full expected-plan/path mapping to the unchanged MIT Jest XML in this directory. Run `collect examples/upstream/collection.json examples/upstream`, save the output separately, then verify: expected gate 1. Using an empty root keeps the same plan and yields gate 3. The E2E evaluator copies the exact XML into a nested temporary job directory, collects it deterministically, preserves the existing summary, refuses malformed bytes and undeclared test identities, rejects traversal before directory access, and recovers after correction. It deliberately includes an unlisted malformed text file to verify that only declared XML is input. No altered upstream report is committed.

`retry/collection.json` reuses the existing synthetic payment plan/first/second XML and explicitly sets deny-recovered. It collects both exact declared paths and verify remains 1. Both new collection specifications are newly authored binding metadata, not captured production configurations.

Repository: [dorny/test-reporter](https://github.com/dorny/test-reporter), commit [cc52f364a405169afe07bb00dab3d375b451229d](https://github.com/dorny/test-reporter/commit/cc52f364a405169afe07bb00dab3d375b451229d), MIT license. Files were downloaded on 2026-10-13 without transformation. The complete upstream license is retained at upstream/LICENSE.dorny-test-reporter.

| Local file | Upstream path | SHA-256 |
| --- | --- | --- |
| upstream/jest-junit.xml | [__tests__/fixtures/jest-junit.xml](https://github.com/dorny/test-reporter/blob/cc52f364a405169afe07bb00dab3d375b451229d/__tests__/fixtures/jest-junit.xml) | 8e18da3a6e4252c34fe7899fad28f8eceeb43849329989b36f2ff68dc7698e58 |
| upstream/pytest-counter-mismatch.xml | [__tests__/fixtures/python-xunit-pytest.xml](https://github.com/dorny/test-reporter/blob/cc52f364a405169afe07bb00dab3d375b451229d/__tests__/fixtures/python-xunit-pytest.xml) | 4b6b57757b6ce395c8491739aaf50199f1d5f3e0235de87cd51e9194d811a560 |

The Jest producer is inspectable upstream: its [package configuration](https://github.com/dorny/test-reporter/blob/cc52f364a405169afe07bb00dab3d375b451229d/reports/jest/package.json) runs Jest with jest-junit and writes this fixture. Its [main tests](https://github.com/dorny/test-reporter/blob/cc52f364a405169afe07bb00dab3d375b451229d/reports/jest/__tests__/main.test.js) and [second tests](https://github.com/dorny/test-reporter/blob/cc52f364a405169afe07bb00dab3d375b451229d/reports/jest/__tests__/second.test.js) match the six observed cases. This is captured framework output for a public sample suite, not a claim of production traffic.

The pytest fixture declares tests="15" but contains ten testcase elements. It deliberately remains unmodified here. Its capture workflow must reject the mismatch with MTL_JUNIT_COUNT; a fixture being in an upstream repository is not enough to establish internal consistency.

~~~sh
moon run --target wasm cmd/moontestledger -- audit examples/upstream/plan.json examples/upstream/bundle.json
moon run --target wasm cmd/moontestledger -- audit examples/upstream/plan.json examples/upstream/missing-bundle.json
moon run --target wasm cmd/moontestledger -- capture examples/upstream/invalid-bundle.json
~~~

Expected outcomes: complete six-case report with exit 1; one missing report/six missing cases with exit 3; counter diagnostic with exit 2.

The JSON plan/bundles are newly authored example bindings. The upstream-macos-jest environment describes the captured report's paths, not a locally repeated macOS test run.

## Synthetic selective retry

Retry/first.xml and retry/second.xml are newly authored boundary fixtures under the project Apache-2.0 license. They model a payment-test shard with pass, failure and skip, followed by a retry of only the failed case.

~~~sh
moon run --target wasm cmd/moontestledger -- audit examples/retry/plan.json examples/retry/bundle.json
moon run --target wasm cmd/moontestledger -- audit examples/retry/plan.json examples/retry/missing-bundle.json
moon run --target wasm cmd/moontestledger -- audit examples/retry/plan.json examples/retry/bundle.json --deny-recovered
~~~

Expected outcomes: complete with one recovered testcase and both attempts retained (0); missing declared retry (3); complete recovery rejected by policy (1).

Bundle.json presents attempt 2 before attempt 1, verifying declared retry order rather than collection order.

The newly authored Apache-2.0 `retry-only-bundle.json` contains just attempt 2. Archive it and `missing-bundle.json` separately against the same full plan, then merge and verify as shown in the root README. This checks collection from independent producers without shrinking the expected inventory.

## Automated evaluator

`continuation/` contains project-authored Apache-2.0 initial inventory, bundle and retry identities. It reuses the existing synthetic first/second payment XML to demonstrate a complete failed run -> newly declared selective retry -> pending archive -> appended retry evidence. The original failure is retained and only the final adverse testcase is scheduled. The README commands run without a test runner or private service; in production the caller's runner executes the generated inventory.

`repetitions/` contains newly authored Apache-2.0 synthetic manifests and a passing-repeat XML. Run A references the existing synthetic failed payment report; run B has a distinct run ID at the same revision/environment and moved shard. The three identities are unchanged, with opposite payment outcomes, a passed case and a skipped case. The README archive/repetitions workflow exits 1 for exactly one observed mixture. `missing-bundle.json` demonstrates absent evidence yielding 3, not a reliability estimate. These examples are not production measurements or upstream test code.

The newly authored `history/` fixtures are Apache-2.0 synthetic runs at two revisions, with a payment regression and a refund improvement. Run the three archive/compare commands in the main README. Neither fixture is altered upstream output. The evaluator also compares an unmodified upstream Jest archive to itself and checks that an incomplete archive cannot pass a comparison.

~~~sh
moon run scripts/e2e.mbtx wasm
moon run scripts/e2e.mbtx native
~~~

The MoonBit script invokes the actual file CLI and verifies exit codes, parsed summary JSON, the six-case upstream report, missing-artifact behavior, capture, retained retry attempts, recovery policy and malformed-counter diagnostics. No server, CI account, token or private infrastructure is required. Native execution requires a working native toolchain.

`moon run scripts/render-demo.mbtx wasm` creates `output/jest-report.html` and `output/jest-evidence.json` directly as UTF-8. The report uses the same unmodified MIT fixture. The renderer and demonstration workflow are project-authored Apache-2.0 code; no external UI assets are used. Open the HTML locally and expand attempt/source sections; the original four failures remain visible. The automated evaluator checks HTML for real failing evidence, missing reports, recovered retries, deterministic rendering and invalid-input diagnostics on Wasm/native.