InviteToFamily previously rejected any second invitation for a (family, node)
pair with PendingInvitationAlreadyExists, even once the existing invitation had
expired and was left inert in the pending map. Now a still-valid invitation still
blocks a duplicate, but an expired one is archived under the new terminal status
FamilyInvitationStatus::Expired and superseded by the fresh invitation.
Regenerated the contract JSON schema and updated the openspec capability.
* experiment: add openspec details for node families contract
* add openspec for the ecash contract
* fix(ecash): correct latest_deposit off-by-one
DepositStorage::latest_deposit() returned the counter value, but the
counter holds the *next* free id (after next_id() saves counter+1). The
GetLatestDeposit handler then tried try_load_by_id(counter), which
always returned None — meaning the query yielded { deposit: None }
both on a fresh contract and after every successful deposit.
Fix: return counter.checked_sub(1) so latest_deposit() yields the most
recently assigned id (or None on a fresh contract). The
getting_latest_deposit unit test is updated to assert Some(0) and
Some(1) after one and two next_id() calls respectively.
No downstream consumer was relying on the buggy semantics
(validator-client exposes the query as a passthrough trait method that
nothing currently calls).
* experiment: add openspec details for ecash contract
Reverse-engineered openspec change `ecash-contract-spec` documenting
the existing CosmWasm contract at `contracts/ecash/`. Mirrors the
node-families workflow: docs-only deliverable, no migration, no
dependency changes. Archived as
openspec/changes/archive/2026-05-21-ecash-contract-spec/ and promoted
to openspec/specs/ecash-contract/spec.md as the canonical reference.
The spec captures 25 normative requirements with 64 scenarios covering
instantiation, migration, deposit submission (default + reduced tier),
RequestRedemption + redemption-proposal reply, legacy RedeemTickets
(dead code retained), stubbed blacklist surface, the ticketbook-size
invariant tripwire, the full query surface, and the public storage /
event / error surface.
Key documented points the source-of-truth phrasing pins down:
- The contract stores claimed ed25519 pubkeys opaquely; ownership is
enforced off-chain by nym-api signers via `validate_deposit`.
- Per-signer-local de-duplication via `state.already_issued`; no
on-chain "issued" state.
- Raw 32-byte deposit storage under the `"deposit"` namespace; deposit
ids are sequential `u32` starting at 0.
- Statistics invariant: default_count + sum(custom_count) = total.
- `cw_controllers::Admin` is used as a generic address-equality helper
for the `multisig` slot (the wrapper's full admin semantics are not
exercised on that slot).
- `RedeemTickets` is dead code retained on the public surface; flagged
as a candidate for removal.
Stubbed-blacklist final disposition is the only Open Question left for
the redesign change owner.
* docs(ecash): add rustdoc derived from archived ecash-contract spec
Drop short doc-comments on the ecash contract surface — handlers,
storage slots, message variants, error variants, event constants,
shared types — derived from the canonical spec at
openspec/specs/ecash-contract/spec.md (archived 2026-05-21).
Coverage:
- contracts/ecash/src/*.rs: crate-root summary, both DepositStorage
and DepositStatsStorage with their invariants called out, every
#[sv::msg(...)] handler in contract/mod.rs, reply id constants,
Config + invariants snapshot, migration entry point.
- common/cosmwasm-smart-contracts/ecash-contract/src/*.rs: every
ExecuteMsg / QueryMsg variant, every reachable EcashContractError
variant (with unreachable-but-preserved variants flagged), every
event constant, every response type, Deposit + DepositId.
Explicitly out of scope (separate concerns):
- Removing event_attributes::BANDWIDTH_PROPOSAL_ID (dead constant,
documented as such for now).
- Removing ExecuteMsg::RedeemTickets (dead handler, documented as such;
removal is a breaking-schema change).
- contracts/ecash/Cargo.toml version bump (docs-only).
No behaviour change; all 38 contract tests pass and cargo doc emits
no warnings on the touched crates.
Jędrzej Stuczyński