Reconcile the contract feature with #755, which replaced the WalletBackend
trait with a concrete WalletBackend<C,K> struct (libwallet/src/backend.rs) and
removed impls/src/backends/lmdb.rs.
- Port the contract module and contract API entry points from the trait form
(T: WalletBackend<'a,C,K>) to the concrete struct.
- Re-home next_child_for and a parent_id+log_id get_tx_log_entry onto backend.rs.
- Re-apply the self-spend tx-log filters, cancel-loop, v5 payment-proof fields,
and Context contract fields onto staging's rewritten files.
- Adapt to fallible iter()/tx_log_iter() and self-consuming batch commit.
Full suite green (53 binaries / 191 tests).
* wallet: update lmdb from store crate, migrate single backend from trait to struct, update tests cargo deps and design doc
* backend: do not panic on iter and directory creation
* return iterator errors
* build: update to last node revision
* fix: height selection for scan, remove unused mutability for wallet
* wallet: remove unused mutability conversion
* tx: store slate state at database
* build: cargo fmt
* fix: do not return error when slate not found on tx slate state update
* lmdb: do not collect all tx log records at iterator, show read errors at log
* fix: save slate state, return error if slate not found
* tx: specify amount of bad txs on slate state update error
---------
Co-authored-by: Joerg <wiesche89@googlemail.com>
Comment-only follow-ups to the review:
- selection: a negative change output is already rejected; drop the stale TODO.
- utils: note the per-participant kernel fee is rounded up (overpay bounded to
under one unit per participant), and that add_outputs/parent_key_id are handled
by the sign step and the stored Context.
- types: drop the "custom outputs imply payjoin" TODO; make_outputs is valid
without a payjoin (e.g. a receiver splitting its output), so the check would be
incorrect.
Addresses PR #754 review (selection.rs, utils.rs, types.rs).
Comment/doc cleanups from review; the only behavior-adjacent change is the
get_net_change signature.
- get_net_change takes the slate id instead of the whole slate.
- replace the owner-API contract methods' `/// TODO` placeholders with real docs.
- narrow or remove now-resolved TODOs (slate proof fns, context, utils stored_tx,
a doc-comment that should be //, wallet_args parse notes); reword the rest as
concrete FUTURE notes.
- prune the contract actions README of items already fixed.
Addresses PR #754 review (comment hygiene).
revoke left the canceled slate's private context in place when the contract
had not been signed (signing already deletes it). Look up the revoked slate id
from its tx log entry and delete the context after cancelling, if one still
exists. Resolves the stale revoke TODOs about the cancelled context.
Addresses PR #754 review (revoke.rs).
contract view hardcoded agreed_net_change to None. When a local context exists
we've agreed on the net change, so read it from the context and report it
(None when there's no context); other errors propagate.
Addresses PR #754 review (view.rs).
The separate contract_setup step is superseded by contract_sign, which runs
setup internally. Remove the dead commented-out function (libwallet, controller,
api), its CLI subcommand (grin-wallet.yml), and its dispatch arm (wallet_args).
The contract_sign path and the parse_contract_setup_args parser are unaffected.
Addresses PR #754 review (command.rs, owner.rs).
ContractNewArgsAPI and ContractView carried a target_slate_version that was only
ever None and never read; the emitted slatepack version is chosen from the slate
content. Remove it. The legacy InitTxArgs/IssueInvoiceTxArgs fields are unrelated
and untouched.
Addresses PR #754 review (types.rs).
Adds the two regression tests requested in review:
- verify_selection_consistency: same/default args are accepted; changed
non-default args after setup are rejected.
- my_fee_contribution: a single participant pays the whole kernel cost, two
participants each pay ceil(kernel/2), and the shares cover the full cost.
Also resolves the related TODOs (selection.rs "test that this works",
utils.rs fee-encoding compatibility note).
Addresses PR #754 review (selection.rs, utils.rs).
add_keys added our participant info without checking the slate wasn't already
full, so a tampered slate could exceed num_participants. Block adding when the
participant set is already full and we aren't already in it (re-adding our own
info stays idempotent).
Addresses PR #754 review (slate.rs).
generate_invoice_signature derived the receiver's proof address from the active
account's parent key. Like the output-derivation fix, use context.parent_key_id
so the proof address belongs to the contract's account, not whatever account is
active.
Addresses PR #754 review (proofs.rs, utils.rs).
The committed-fee-exceeded check returned a GenericError. Use the existing
Error::Fee variant and fold the "cancel and retry" hint into the message.
Addresses PR #754 review (selection.rs).
A large u64 amount could wrap when cast `as i64`. Convert with i64::try_from in
the CLI net-change derivation and in contract view, returning an error instead.
get_net_change / to_api_args now return Result (the two arg panics become
errors too), and the to_api_args call moves out of the owner-session closure.
Addresses PR #754 review (command.rs, view.rs).
num_participants is divided into the kernel cost, so 0 caused a divide-by-zero
(inf -> garbage fee). Validate it in context::create, the common setup path for
every party: the initiator's value comes from args and the signer's from the
slate, so this also rejects a tampered slate. Only 1 (self-spend) and 2
(standard) are supported today; lift the upper bound when multi-party contracts
are implemented.
Addresses PR #754 review (wallet_args.rs:1039).
The foreign contract_sign API unwrapped the optional net_change field, so an
incomplete request could panic the wallet. Return an error instead, matching
the foreign contract_new path.
Addresses PR #754 review (foreign.rs).
revoke() cancelled the original tx and built the self-spend using the wallet's
active account, but the locked inputs belong to the account that created the
contract. Revoking while a different account was active would fail to find and
cancel the original tx and would recover funds into the wrong account.
- derive the account from the locked input's root_key_id; use it for cancel_tx
and as the self-spend's src_acct_name.
- test: revoke a contract while a different account is active.
Addresses PR #754 review (revoke.rs).
create_slatepack always serialized to V5, so every slatepack went out V5 and
V4-only peers could not read it. Emit V4 unless the slate carries a payment
proof: only the (opt-in) proof needs the V5-only fields (timestamp/memo) that
V4 cannot represent. The slate's declared version is set to match the emitted
structure so readers select the right variant.
- packer: V5 when payment_proof is present, else V4.
- test: a proof slate round-trips as V5 (memo preserved); a no-proof slate as V4.
Addresses PR #754 review (packer.rs).
contract `new` and the foreign `contract_new` API unwrapped the optional
net_change field, so a request omitting it panicked the wallet. Return an
error instead.
- new::compute and foreign::contract_new: net_change.ok_or_else(..).
- test: contract_new with no net change errors rather than panicking.
Addresses PR #754 review (new.rs, foreign.rs).
get_private_context() flattened "missing context" and real store/keychain
errors into one error, so the contract code created a fresh context on ANY
error (e.g. a decryption failure from a wrong keychain mask), masking the
real problem.
- add Error::NotFoundErr; the grin_store conversion maps NotFoundErr through
faithfully (other store errors stay Backend).
- contract get_or_create / get_net_change / sign create-or-skip only on
NotFoundErr and propagate every other error.
Addresses PR #754 review (context.rs, utils.rs, sign.rs).
Custom output amounts (--make-outputs) and the CLI amount parser converted
grin to nanogrin with `(g * GRIN_BASE as f64) as u64`, which can't represent
all values exactly. Use core::amount_from_hr_string — the same exact integer
parser already used by `send` and `invoice`.
- contract OutputSelectionArgs::output_amounts (covers selection.rs).
- wallet_args::parse_grin_amount.
Addresses PR #754 review (types.rs, selection.rs, wallet_args.rs).
New contract outputs were derived with next_child(), which uses the wallet's
active account. During a late-lock the active account can differ from the one the
contract context was created under, so an output could be created in the wrong
account. Derive under context.parent_key_id instead.
- WalletBackend: add next_child_for(parent_key_id, mask); next_child delegates to it.
- add_outputs_to_ctx derives via next_child_for(&context.parent_key_id, ..).
- parent_key_for returns Result and errors (UnknownAccountLabel) on an unknown
account name instead of silently falling back to the active account.
- test: a contract referencing an unknown account now errors.
Addresses PR #754 review (context.rs, utils.rs).
--encrypt-for is no longer required. `contract new` without it produces an
unencrypted slatepack; `contract sign` without it replies to the incoming slate's
sender when known, otherwise unencrypted. Passing --encrypt-for is unchanged. The
slatepack layer already produced plaintext for an empty recipient list — this only
removes the CLI-side requirement.
Bring the branch current with upstream staging (= master + #748). The earlier
"Merge branch master into contracts" used a stale local master snapshot, which
left the branch ~41 commits behind; this merges the current tip, including the
grin 5.4.0 dependency bump (#745) and the tx-delete / retrieve_txs fixes.
Only Cargo.lock conflicted (regenerated). Full cargo test --workspace green
(52 binaries / 185 tests).
OutputSelectionArgs::output_amounts() unwrap()'d the float parse of each
make-outputs value, panicking on bad input via the API path (the CLI now
validates earlier). Return Result instead; sum_output_amounts() propagates.
num_custom_outputs() now counts CSV entries without parsing, so it stays
infallible and the debug! macros that use it are unaffected.
Bad input to the contract CLI could abort the process:
- a non-numeric --make-outputs value reached output_amounts(), which
unwrap()s the float parse;
- `contract new` with neither --send nor --receive hit a panic in
get_net_change;
- `contract sign` panicked ("No address to encrypt for") when given no
--encrypt-for and an incoming slate with no embedded sender.
Validate --make-outputs and require exactly one of --send/--receive at
parse time (ParseError, before reaching the panicking paths), and return a
typed error instead of panicking when there is no address to encrypt for.
Verified at runtime: each now yields a clean "Invalid Arguments" error.
Remove the broken, commented-out SEC_KEY_FAKE key-scrubbing: a zero secret
key is rejected by secp, and the scrubbing was redundant. The signing
context holding sec_key/sec_nonce is forgotten by deleting the private
context in save_step once signed (retained only at step2 for safe cancel),
and contexts are keyed by slate.id so a nonce is never reused across
messages.
Make that guarantee enforced rather than assumed: after a signed,
non-step2 save_step, verify the private context was actually removed and
error otherwise. Document the invariant at the sign site.
Contract setup/sign and output selection used panic! on
protocol-reachable invariant violations: receiver fee exceeding
net_change, net_change / num_participants mismatch between the setup and
sign phases, a negative change output, conflicting selection args, and a
failed kernel-excess computation. A crafted or inconsistent counterparty
slate could abort the wallet. Return Error::GenericError (or propagate the
underlying error) instead; build_output_amount_list now returns Result.
The contract new/sign/revoke parsers panicked or unwrapped on user input:
conflicting --no-payjoin/--use-inputs, and unparseable receive / send /
num-participants / tx-id values would abort the process. Return
ParseError instead, and add a parse_grin_amount helper for the
Grin->nanogram conversion.
ProofWrap::read unwrapped on the sender/receiver address, promise
signature, and memo_type while parsing untrusted slatepack bytes, so a
malformed payment proof would panic the wallet. Propagate CorruptedData
instead, matching the timestamp path already in this function.
Default ProofArgs.suppress_proof to true. With it false, the receiver's
`contract sign` unconditionally built an invoice promise, but the sender
address is never plumbed through, so a bare send/receive failed with
NoSenderAddressProvided (#729). The CLI never sets this field, so every
default `contract sign --receive` hit the broken path; the basic tests
passed only because they set suppress_proof = true by hand.
Drop those workarounds from contract_srs/contract_rsr so they now exercise
the default path, and opt the proof-using tests (contract_early_proofs_srs,
contract_early_proofs_rsr, contract_accounts) into proofs explicitly.
master's mwixnet (#726) keeps the `types` submodule private and re-exports
its items, so use the public `mwixnet::MixnetReqCreationParams` path. The
merge missed this because `cargo check --workspace` does not build test
targets.
Bring the contracts feature branch up to current master, which has since
landed rust 1.80 support (#722), the canonical mwixnet onion code (#726),
and node-client proxy support (#738).
Conflict resolution:
- libwallet/src/mwixnet/*: take master's (#726), dropping the contracts
branch's older divergent copy. owner.rs::create_mwixnet_req updated to
master's API (use_test_rng) and master's flat re-export imports.
- Cargo manifests/lock/.cargo: master's deps & version; re-add serde_with
for Slate v5; keep contracts' .cargo/config.toml (Windows static-CRT).
- lib.rs: keep serde_derive AND serde_with macro_use; single mwixnet decl.
- error.rs: union of master + contracts variants; flat mwixnet re-exports.
- api/libwallet owner.rs: master's #738 + create_mwixnet_req plus the
contract_* entry points; drop duplicate older create_mwixnet_req method.
- slatepack/armor.rs, node_clients/resp_types.rs: take master's.
Workspace builds; libwallet lib tests pass (34/34). The Slate version()
bug, issue #729 fix, and de-panicking follow in subsequent commits.
* scan: save last scanned block info for initial wallet scanning
* scan: save start height at scanned block info on outputs collecting
* scan: 10k blocks batches
* fix: include last height into batch on scan
* scan: save hash of last block
* scan: add larger window (2880 blocks) to scan outputs when updating wallet state
* add test batched scan and resume, add scan doc
* rename corrupted restore test
---------
Co-authored-by: Joerg <wiesche89@googlemail.com>
jwinterm