feat(header): BlockInfo timeslot + epoch-mark length prefix + preimage-hash extrinsic commitment (GP v0.8.0) (#1014)

[HanaYukii]

Part of #1012 · Closes #1014

Stacked on #1027 (← #1026). Base is feat/reporting-1016-workcontext-fields; review only the #1014 commit here. Retarget to 1012-update-to-v080 as the stack merges.

What — three consensus-visible v0.8.0 changes

#	change	GP ref
1	BlockInfo (recent history) gains a 4-byte timeslot between state_root and reported; the appended item carries H_t	eq:recenthistoryspec, eq:recenthistorydef, C(3)
2	EpochMark's validator-key sequence is now length-prefixed (var{k}); was fixed-length	encodeepochmark (serialization.tex)
3	Extrinsic hash: preimages component becomes p = E(var[(E4(s), blake(d))]) — each blob committed by its Blake2b hash, replacing the full C.15 encoding	header.tex

Everything else in the issue is already aligned (verified against the v0.7.2→v0.8.0 tex diff): epoch/winning-tickets marker presence+content rules, the ancestors requirement, and the author-index bound are semantically unchanged.

Changes

• types.BlockInfo + Encode / Decode / UnmarshalJSON (+ jam_types struct kept in sync; reflection registries stay on v0.7.x per feat(work-package): align GP v0.8.0 work packages and bundles #1015 precedent)
• EpochMark.Encode/Decode — emit/read the length prefix (decode still validates it equals ValidatorsCount)
• block_serialization.go — new p() helper (mirrors g()) used by CreateExtrinsicHash; full-block serialization (C.15) unchanged — only the hash commitment changes
• recent_history.NewItem takes the block slot; both STF call sites pass block.Header.Slot

Tested

• ✅ TestEncodeBlockInfo_V080Timeslot, TestEncodeEpochMark_V080LengthPrefix (new) — round-trip + byte-offset asserts
• ✅ TestExtrinsicHashPreimagesComponent_V080 (new) — count prefix + E4 requester + blake(blob), empty case
• ✅ go build / go vet / gofmt clean; internal/types, recent_history, safrole, statistics, store, extrinsic, input/jam_types green
• ⏭️ v0.7.x vector fallout skipped until official v0.8.0 vectors land (Update to v0.8.0 #1012): history + safrole STF modes (v080IncompatibleModes), all traces (embed full serialized state), and the matching unit vector tests (skipV080VectorIncompat + per-test skips)
• ℹ️ Pre-existing, unrelated: internal/header (TestCreateParentHash / TestCreateExtrinsicHash) and internal/blockchain (TestGetGenesisBlock / TestGetGenesisState) fail identically on the pristine 1012-update-to-v080 base (verified at 6d93ac4) — stale v0.7.x fixtures, untouched here

[yu2C]

Summary

Thanks for the v0.8.0 wire-format updates (BlockInfo timeslot, EpochMark length prefix, preimage-hash extrinsic commitment). The codec round-trip unit tests look correct.

One architectural suggestion around the new p() helper in block_serialization.go — please address before merge if possible.

---

Issue Verification

Test	Status	Evidence
go test ./internal/recent_history/... -count=1	PASS (skipped vectors)	TestRecentHistoryTestVectors skipped pending v0.8.0 vectors
go test ./internal/stf/... -count=1 -run Header	PASS (no tests)	internal/stf has no test files
TestEncodeBlockInfo_V080Timeslot	PASS	codec round-trip
TestEncodeEpochMark_V080LengthPrefix	PASS	codec round-trip
TestExtrinsicHashPreimagesComponent_V080	PASS	p() byte layout

Linked issues: #1014 (Part of #1012)

---

internal/utilities/block_serialization.go

[SUGGESTION] Move p() into internal/types/encode.go

p() is only called from CreateExtrinsicHash (used by validate_header.go and header_controller.go). It does not participate in block wire serialization (BlockSerialization still uses C.15 via ExtrinsicPreimageSerialization).

The v0.8.0 extrinsic-hash preimages component (header.tex p = E(var[(E4(s), blake(d))])) is intentionally not the same as PreimagesExtrinsic.Encode (C.15: E4(s) + length-prefixed blob). That distinction is correct — please do not change the existing Encode/Decode.

However, placing p() in block_serialization.go (alongside deprecated hand-rolled *Serialization helpers) makes it look like legacy code. The same file already has this split:

• C.15 block wire → types.PreimagesExtrinsic.Encode / Decode
• Extrinsic-hash component → p() (hash-only, no decode)

Please move the extrinsic-hash encoding into internal/types/encode.go as a separately named method, e.g.:

// EncodeForExtrinsicHash implements header.tex p =
// E(var[(E4(s), blake(d)) for (s, d) in E_P]).
// NOT the C.15 block-wire encoding; see PreimagesExtrinsic.Encode.
func (p *PreimagesExtrinsic) EncodeForExtrinsicHash(e *Encoder) error { ... }

Then have CreateExtrinsicHash call that method via the shared encoder, and remove the private p() helper. (Same pattern could apply to g() in a follow-up, but p() is in scope for this PR.)

Why encode.go?

• types.go — struct definitions only
• encoder.go — low-level Encoder machinery
• encode.go — per-type encode methods ✅
• decode.go — no decode needed for hash-only commitment

---

internal/utilities/extrinsic_hash_v080_test.go

[SUGGESTION] Merge test into block_serialization_test.go

Please drop the standalone extrinsic_hash_v080_test.go and move TestExtrinsicHashPreimagesComponent_V080 into block_serialization_test.go alongside the other serialization tests. If p() moves to types, rename the test accordingly (e.g. TestPreimagesExtrinsic_EncodeForExtrinsicHash_V080 in internal/types/encode_test.go).

---

Files reviewed (no issues)

• internal/types/types.go, decode.go, encode.go, unmarshal_json.go — BlockInfo timeslot, EpochMark length prefix ✅
• internal/recent_history/recent_history_controller.go — NewItem passes block.Header.Slot ✅
• internal/input/jam_types/jam_types.go — struct sync ✅
• testdata/reader.go — v0.8.0 skip lists updated ✅

---

Severity: 0 CRITICAL · 2 SUGGESTION · 0 NIT

[HanaYukii]

Thanks @yu2C — both suggestions applied in HEAD:

• p() → types.PreimagesExtrinsic.EncodeForExtrinsicHash (encode.go), doc comment spelling out it is the header.tex extrinsic-hash component and NOT the C.15 wire encoding (which is untouched). One wrinkle vs the sketch: types cannot import utilities/hash (that package imports types), so the method uses x/crypto/blake2b directly — same primitive Blake2bHash wraps. Added a small Encoder.EncodeFunc so CreateExtrinsicHash can run the alternate encoding through the pooled encoder; the private p() is gone. (g() can migrate the same way in a follow-up, as you noted.)
• Test moved to internal/types/encode_test.go as TestPreimagesExtrinsic_EncodeForExtrinsicHash_V080; extrinsic_hash_v080_test.go deleted.