morluto
2cf9c28f27
Move plist projection into the graph owner and identify generated scalar wrappers by producer identity through every dictionary/UID consumer. Remove the intermediate JSON clone that erased those facts. Literal wrapper-shaped dictionaries remain dictionaries; scalar $top values fail as malformed. Keep numeric projection after graph interpretation, preserve unknown counts, and retain binary prototype-key partial facts with an explicit limitation. Adapt @Jax367's relevant scalar proposal and public regression from PR #4317. Add scalar-root, literal-root and binary prototype controls. Correct only the captured Foundation data node expectation and its manifest digest; original producer bytes and source are unchanged. Unrelated PR changes excluded. Validation: 175 focused checks across 23 files pass, one skipped; build, check:fast and docs:check pass. Public CLI and stdio MCP cover both encodings. Baseline scalar controls fail; 24 tiny completed CLI cost comparisons retain ordinary full replies and Evidence IDs byte for byte. No statistical speed claim, heavy profiling, target execution or real deep-provider verification. Fixes #4295 Fixes #4785
Captured format and protocol fixtures
These are actual tool outputs, not records assembled to match the parser.
provenance.json records the capture host, producer versions, source digests
and immutable capture/expectation digests. Compiled binaries and catalogs are
not committed.
Foundation archive
keyed-archive/foundation.xml is the XML representation of a binary archive
written by Foundation NSKeyedArchiver, using the source-owned Swift fixture.
plutil represents binary UID entries as CF$UID dictionaries. The captured
archive has six objects: nil, the root NSArray, one shared custom record,
its four-byte data payload and two class descriptors.
keyed-archive/graph.json is the reviewed complete expected graph. The source
establishes two references from the array to object 2, object 2's self-cycle,
its nil conditional reference and the payload 00 01 02 ff. The reader reports
serialized classes as unknown and does not instantiate them. The synthetic
malformed-reference tests remain separate.
Object 3 is a plist <data> value in the captured XML, produced by the Swift
fixture's Data([0, 1, 2, 255]). Its reviewed graph kind is scalar; the JSON
$plist_type wrapper does not make it a serialized dictionary. This expectation
correction updates only the graph digest in provenance.json, without
recapturing or changing the original XML or its producer metadata.
To recapture the input on an approved macOS host:
capture_root=$(mktemp -d)
xcrun swift -module-cache-path "$capture_root/modules" \
tests/conformance/native/keyed-archive.swift "$capture_root/model.plist"
plutil -convert xml1 -o tests/fixtures/golden/keyed-archive/foundation.xml \
"$capture_root/model.plist"
rm -rf "$capture_root"
Review expected graph changes against the source and XML before updating the expectation or provenance. Tests never rewrite the golden.
Asset catalog
assetutil.json is the unmodified output of /usr/bin/assetutil --info for an
actual actool catalog. The source colorset defines FixtureColor with sRGB
components [0.125, 0.5, 0.875, 1]. Preserve private fields and producer metadata
in this fixed input, including its original timestamp; the test checks that the
projection does not discard them.
capture_root=$(mktemp -d)
xcrun actool tests/conformance/native/Assets.xcassets \
--compile "$capture_root" --platform macosx --minimum-deployment-target 12.0
/usr/bin/assetutil --info "$capture_root/Assets.car" \
> tests/fixtures/golden/assetutil.json
rm -rf "$capture_root"
The real E2E lane builds fresh catalogs and compares all their metadata against the live utility output. It does not compare volatile timestamps or producer versions against this historical capture.
MCP JSON-RPC request
mcp-initialize.json captures the initialize request emitted by the pinned
@modelcontextprotocol/client 2.0.0 SDK on 2026-10-04. The client connected to
the actual scripts/rea.mjs mcp subprocess and completed listTools (the current catalog).
The recorder intercepted transport.send, serialized the message, and delegated
the unchanged message to the real SDK transport. The committed JSON is formatted
for review; whitespace is not a wire identity claim. No provider was launched.
The decoder expectation separately specifies every field path, JSON value and observed-versus-inferred status. It is a captured decoder regression, not proof of a live protocol-capture product workflow.