build(deps)!: uplift dig-wallet onto chia-wallet-sdk 0.34 / chia 0.36.1

[MichaelTaylor3d]

Task

Uplift dig-wallet (workspace member of dig-node) onto chia-wallet-sdk 0.34 / the chia 0.36.1 line, so it stops being the last crate behind in the ecosystem-wide chia-0.36 cascade.

Parent epic: https://github.com/DIG-Network/dig_ecosystem/issues/2610

Context

crates/dig-wallet/Cargo.toml declares (measured on origin/main @ 8c31c58):

line	declaration	target
27	chia-protocol = "0.26"	0.36.1
35	chia = "0.26"	no 0.36 line exists — replace with facade crates
38	chia-wallet-sdk = { version = "0.30", … }	0.34
41	clvmr = "0.14"	whatever SDK 0.34 resolves
88	chia-query = "0.6.2"	0.15
149	chia-traits = "0.26"	0.36.1
175	chia-sha2 = "0.26" (dev)	0.36.1
184	chia-sdk-test = { version = "0.30", … } (dev)	0.34

The chia umbrella publishes … 0.29.0, 0.32.0, 0.42.0 … — there is no 0.36. The facade set is
determined by reading chia:: import paths: chia::bls, chia::clvm_utils, chia::protocol,
chia::puzzles, chia::secp, chia::traits.

Scope

dig-wallet is on the money path (balances, coin selection, spend construction, signing). No
cross-version shim — if the API delta or the dependency graph does not line up, the deliverable is
the assessment, not a bridge. §908 binds: the node signs nothing on the user's behalf.

Stay inside crates/dig-wallet/; dig-node-core is owned by another lane.

Evidence

Every declaration cited file:line; the facade set justified from import paths; any chia type reaching
a public signature reported with how many lines it comes from; money-path byte-identity measured, not
argued.

[MichaelTaylor3d]

Verdict: BLOCKED — dig-wallet cannot move to chia 0.36.1 alone. Release-first on digstore-chain is required.

Measured on origin/main @ 8c31c58, not argued. No shim was written.

The blocker

dig-wallet depends on digstore-chain as a git dep pinned to rev 4c34f0be
(crates/dig-wallet/Cargo.toml:21, repo DIG-Network/digstore). At that rev digstore-chain declares:

digstore-chain crates/digstore-chain/Cargo.toml	version
:31 chia	0.26
:32 chia-protocol	0.26
:40 chia-wallet-sdk	0.30
:52 clvmr	0.14

digstore-chain is dig-wallet's spend-construction and signing surface, and it traffics in chia types
across the boundary — e.g. crates/digstore-chain/src/cat.rs:28 is literally
use chia_protocol::{Bytes32, CoinSpend, SpendBundle};, and:

pub fn cat_puzzle_hash(owner_puzzle_hash: Bytes32, asset_id: Bytes32) -> Bytes32
pub fn build_cat_send(keys: &WalletKeys, cats: &[Cat], asset_id: Bytes32,
                      recipient_ph: Bytes32, …) -> Result<(SpendBundle, CatSendPlan)>

These unify today only because both sides are 0.26. Move dig-wallet to 0.36.1 and they are two
different types with the same name.

The measurement

I applied the full uplift (manifest + the mechanical chia:: -> facade rewrite) and compiled. Preserved
for inspection on branch loop/2610-uplift-attempt-evidence (commit 70c5d26).

cargo check -p dig-wallet --all-targets -> exit 101, 135 errors:

error	count
E0308 mismatched types	114
E0277 trait bound	9
E0599 no method	6
E0271 type mismatch	4

132 of 135 carry a "multiple different versions of crate" note — chia_protocol (116),
chia_bls (11), chia_sdk_driver (4), chia_secp (1). The canonical failure:

error[E0308]: mismatched types
  --> crates/dig-wallet/src/lib.rs:668:72
      digstore_chain::singleton::sync_datastore_with_history(&chain, store_id)
      expected `dig_clvm::chia_protocol::BytesImpl<32>`,
         found `chia_protocol::BytesImpl<32>`

This is not an API delta. Nothing in the 0.30 -> 0.34 surface change produced these. Every failure is
one chia line meeting another at the digstore_chain::* boundary — cat_puzzle_hash, cat_balance,
build_cat_send, build_vault_create, decode_xch_address, ClawbackCoin, MakerFunds,
DIG_ASSET_ID, vc_asserter_puzzle_hash, sync_datastore_with_history. That is the byte-drift class
(#2497, #2556, #2608), on the money path. It cannot be fixed inside crates/dig-wallet/, and bridging it
is exactly what must not be done.

Required order (release-first, §4.1)

1. digstore-chain (DIG-Network/digstore) releases on chia-protocol 0.36.1 / chia-wallet-sdk 0.34.
2. dig-wallet re-pins the git rev, then this uplift applies — the patch is already written.

The chia umbrella replacement (valid, reusable)

The umbrella has no 0.36 line (… 0.29.0, 0.32.0, 0.42.0 …). The facade set, determined by reading
the chia:: module paths actually imported (72 occurrences across 19 files), not by guessing:

umbrella path	facade crate	n
chia::bls	chia-bls	23
chia::protocol	chia-protocol	19
chia::puzzles	chia-puzzle-types	13
chia::traits	chia-traits	13
chia::clvm_utils	clvm-utils	2
chia::secp	chia-secp	2

There are no grouped use chia::{…} forms, so the rewrite is exactly these six prefixes. clvmr pairs
0.16 with the 0.36.1 line (currently 0.14).

The invariant: chia types DO reach public signatures, from ONE line today

dig-wallet's crate root is chia-free (wallet_dispatch(&str,&str)->(u16,String), run()), but
pub mod sage exposes chia types:

• sage/chain.rs:234,354,371 — SpendBundle (push / decode / encode signed bundle)
• sage/coverage.rs:69, sage/custody.rs:300 — Bytes32
• sage/custody.rs:526, sage/rpc.rs:825 (watch_keys) — chia_bls::PublicKey
• sage/quorum.rs:768 — chia::protocol::Coin

All resolve from chia_protocol / the chia umbrella at 0.26 — a single line. So dig-wallet is NOT
internally split the way dig-gossip 0.29.0 and dig-nat 0.19.0 were. The uplift would keep it single-line
only if digstore-chain moves with it.

Pre-existing multi-line state (context, not caused by this work)

dig-wallet's graph already resolves two chia-protocol lines. cargo tree on unmodified main:
chia-protocol {0.26.0, 0.36.1}, chia-wallet-sdk {0.30.0, 0.34.0}, chia-bls {0.22.0, 0.26.0,
0.28.2, 0.36.1, 0.42.1, 0.45.0}, clvmr {0.14.0, 0.16.4}, chia-query {0.5.1, 0.6.2}.

Traced to root: the 0.36.1 line enters via dig-offers 0.2.0 (a direct dig-wallet dep) ->
chia-wallet-sdk 0.34 -> chia-consensus 0.36.1, and via dig-node-core -> dig-download 0.19.0 ->
dig-constants 0.10.1/0.11.0. The manifest already acknowledges the dig-offers half at :57. So the
uplift would reduce the split (unifying dig-wallet with dig-offers and dig-constants) while creating
a worse one against digstore-chain, which is the one that carries money.

Two coordination points — reported, NOT written

1. dig-node-core stays on chia-protocol 0.26 / chia-query =0.5.1 (:85, :112), pinned because
   chia-peer 0.1.3 builds against dig-chainsource-interface 0.1. Owned by the lane on PR feat(relay): module-granularity onion relay leg + absence-safe locator readers #299; not touched.
2. dig-node-service dev-deps declare chia = "0.26" (:229) explicitly "pinned to the same lines
   dig-wallet builds on so exactly one of each resolves", for the §18.6f watch-list tests that drive
   dig-wallet's watch_keys(&[PublicKey]). Uplifting dig-wallet desynchronises that pin. Test-only, but real.

Signing paths

The uplift touches signing transitively: digstore_chain::chip0002::{sign_message_with, sign_wc_coin_spends, sign_coin_spends_hex} and sign_did_spends / sign_clawback_spends, which sign via
datalayer_driver::sign_coin_spends under AugScheme with chia-bls keys. Moving chia-bls 0.26 -> 0.36.1
moves that serializer. §908 is not weakened: the node still signs nothing on the user's behalf; these
are local derivations from a user-supplied mnemonic. This is precisely why the vectors below were blessed
before any uplift.

What DID land here

Money-path golden vectors, blessed on the current line (test-only, no manifest change of its own):
crates/dig-wallet/tests/money_path_vectors.rs. See the next comment.

[MichaelTaylor3d]

Money-path golden vectors — blessed on the CURRENT line

crates/dig-wallet/tests/money_path_vectors.rs (test-only; its own commit eec9570, no manifest change).

Position: none existed. tests/vectors/ held only sage-openapi-v0.12.11.json and
sage-endpoints-v0.12.11.json — API shape fixtures, not signing/serialization/address bytes. The one
pre-existing cross-stack pin, offers.rs:741 offer_id_matches_a_locally_computed_bundle_hash, recomputes
dynamically and pins no literal. So nothing existed to "verify still passes" — they were blessed, as the
brief directs, on the line the crate builds against today.

Derived from the public BIP-39 all-abandon test vector — never a real wallet, holds nothing.

vector	value
synthetic pubkey	93c7d36e915aa1570087c9adc427c3a9bb532efe964dcc3bb04a07bc64308dbd82598a1f49f6ca86a82b32559e41380e
owner puzzle hash	d207c1e11fc3b0cd7472e8c7e53c8d2b81709516346c7baa9fbb9070ffccfe89
receive address	xch16grurcglcwcv6arjarr720yd9wqhp9gkx3k8h25lhwg8pl7vl6ysuax0gy
DIG asset id	a406d3a9de984d03c9591c10d917593b434d5263cabe2b42f6b367df16832f81
DIG CAT puzzle hash	1a0fb6b58621fb2fa657b1b0b6c75bd34a7655b463889aad17fe9425b1a9b764
indexed PH 0/1/2	d207c1e1…fccfe89 / 7ba6cfb6…d4052f / 6aaf6665…0d14a3
AugScheme signature	8acb18df…9829d7

No byte changed against the current line — these are the current line, captured and pinned.

Fixture design, not just assertions

• Indexed derivation covers a WINDOW (0..3), not index 0. A single-index fixture cannot see a
  derivation change that happens to fix index 0 while losing every other coin the wallet scans for.
• index_zero_agrees_with_default_derivation is a control against the two code paths silently
  diverging — digstore-chain documents them as byte-identical, so the fixture pins the claim.
• signature_varies_with_the_message is the anti-vacuity control: without it,
  signature_is_byte_identical would pass against a constant or empty signature.

Mutation proof — each vector reddens BY NAME, none over-fires

Committed first (git checkout is destructive on uncommitted work), then one hex character flipped per
constant, suite re-run, tree restored:

mutated	failing test	others
SYNTH_PK	synthetic_public_key_is_byte_identical	7 passed
OWNER_PH	owner_puzzle_hash_is_byte_identical	7 passed
ADDRESS	receive_address_is_byte_identical	7 passed
DIG_ASSET	dig_cat_puzzle_hash_is_byte_identical	7 passed
DIG_CAT_PH	dig_cat_puzzle_hash_is_byte_identical	7 passed
IDX_PH[1]	indexed_derivation_is_byte_identical	7 passed
SIG	signature_is_byte_identical	7 passed

Exactly one named failure each time — load-bearing, and no over-firing.

Limit, stated plainly: these mutate the expected constants, which proves the assertions are live and
correctly scoped. They cannot mutate the producer, because the producer is digstore-chain, an external
git-pinned crate. The vectors' real job is the uplift: when digstore-chain moves to 0.36.1, any of these
seven values changing is a compatibility break, and it will now fail by name instead of silently.

Suite

cargo test -p dig-wallet --no-fail-fast -> 716 passed, 0 failed, 3 ignored, exit 0
(686 lib + 22 conformance + 8 new vectors).
Pre-existing baseline on origin/main @ 8c31c58 was 708 passed / 0 failed / 3 ignored — so the delta is
exactly the 8 added, with nothing disturbed.