Skip to content

Replace bespoke JSON tx-spec with CCL TxPlan (YAML): offline build, Plutus, full intent coverage#2

Merged
matiwinnetou merged 30 commits into
developfrom
feature/txplan-yaml
Jun 12, 2026
Merged

Replace bespoke JSON tx-spec with CCL TxPlan (YAML): offline build, Plutus, full intent coverage#2
matiwinnetou merged 30 commits into
developfrom
feature/txplan-yaml

Conversation

@matiwinnetou

@matiwinnetou matiwinnetou commented Jun 11, 2026 •
edited
Loading

Copy link
Copy Markdown

Summary

Replaces the bridge's bespoke intermediate JSON transaction format with CCL's native TxPlan (YAML) format, builds transactions fully offline, and verifies — both offline and against a real Cardano node — that the bridge produces node-acceptable transactions across the full intent surface (payments, staking, DRep, voting, governance, pools, metadata, and Plutus).

Net: ≈−13,200 lines (the bespoke spec, its mappers, the provider path, and the four per-language fluent builders are gone).

Upgrades CCL 0.7.2 → 0.8.0-pre4 (backward-compatible for the bridge; only the deprecated ScriptTx path changes).

What changed

Core — TxPlan in, YAML out

  • Deleted the bespoke spec + mappers + provider path (~2,400 LOC). QuickTxService now: TxPlan.from(yaml) → QuickTxBuilder(staticUtxoSupplier, () -> protocolParams, null) → compose(plan) → build() → CBOR. Fully offline; never submits.
  • Entry point: ccl_quicktx_build(thread, yaml, utxos_json, protocol_params_json, exec_units_json). YAML in, YAML out; chain data is caller-supplied (no provider).

Plutus script transactions (offline)

  • Execution units are a fourth caller-supplied input (like UTXOs/params); the bridge wires CCL's StaticTransactionEvaluator to stamp them on — it never runs the script. Callers compute them with any evaluator (Ogmios/Blockfrost/Aiken/Scalus); see TODO.md §2b for the planned helper layer.
  • native-image: registered the Plutus + cert + serializer reflection Jackson needs (these pass on the JVM but fail in a native image without registration — caught by the Go E2E suite).

Stake / DRep-key signing

  • New ccl_account_sign_tx_multi(…, keys) signs with any subset of payment / stake / drep / committee_cold / committee_hot (CCL Account.signWith*Key). Fixes the MissingVKeyWitnessesUTXOW rejection for stake/vote/DRep certs; the original payment-only ccl_account_sign_tx is unchanged. Wired through all wrappers (sign_tx_with_keys / SignTxWithKeys / signTxWithKeys).

Wrappers — thin YAML pass-through

  • Deleted the fluent builders (~10k LOC across Python/Go/Rust/JS); each exposes build(yaml, utxos, protocolParams, execUnits?) and parses the YAML result.

Intent coverage — every intent, every wrapper, end-to-end

  • test-fixtures/quicktx-intents/: a shared, version-controlled fixture set (generated by the JVM QuickTxIntentsTest via TxPlan.from(tx).toYaml() — CCL's exact intent shapes, single source of truth), incl. compose (multi-sender) and the Plutus mint/spend.
  • Table-driven E2E tests in Go, Python, Rust, JS drive every fixture through the native library. Covers: payment, metadata, native minting, donation, stake registration/deregistration/delegation/withdrawal, DRep registration/deregistration/update, voting, voting_delegation, governance_proposal, pool registration/update/retirement, native_script, collect_from, reference_input, compose, and both Plutus paths.

On-chain validation (Yaci DevKit)

  • A Go submit-suite (CI "Integration Tests (DevKit)" job) does build → sign (with the right key roles) → submit → assert on-chain for the intents, turning "builds offline" into "a real node validates and accepts it", with the required on-chain setup sequenced (e.g. register a stake address before a proposal; register a pool before delegating; vote-delegate before withdrawing). Depth checks: mints assert the asset landed at the receiver, the Plutus spend asserts the script UTXO was consumed, and a negative test proves a DRep registration signed with the payment key only is rejected (so the extra signing witness is genuinely required).

Decisions (agreed during development)

  • YAML in and out; client-side UTXOs + protocol params only (no provider); Plutus exec units are caller-supplied; wrappers are thin pass-throughs.

Verification

  • :core:test + :core:nativeCompile green; all four wrapper suites green (per-intent E2E + Plutus).
  • CI green on Linux x86_64, macOS ARM64, Windows. The DevKit job submits the intent set to a live devnet.

Known limitations / follow-ups (not blockers)

  • Pluggable exec-unit evaluators (TODO.md §2b): client-side Blockfrost/Ogmios/Aiken/Scalus helpers + examples for obtaining the units to pass in. (Deferred — the one remaining real feature.)
  • Cross-wrapper submit: only Go submits to DevKit; since the CBOR is produced by the shared native lib, Go-only proves the txs are node-acceptable, and the other three wrappers have build E2E parity.
  • Depth: most intent E2E tests assert "builds"; the DevKit suite adds on-chain acceptance for the covered set.

Mateusz Czeladka and others added 19 commits June 11, 2026 15:04
@claude
Core: replace bespoke JSON tx-spec with CCL TxPlan (YAML), build offline
d2ad9e2 
Upgrade CCL 0.7.2 -> 0.8.0-pre4 (backward compatible; only the deprecated
ScriptTx path changes) and adopt CCL's native TxPlan YAML format.

- Delete the bespoke spec + mappers + provider path (~2400 LOC):
  TxSpec, TxOperation, TxItemSpec, TxSpecMapper, ScriptTxSpecMapper,
  ProviderConfig, Yaci{Utxo,ProtocolParams}Supplier, YaciTransactionEvaluator.
- Rewrite QuickTxService to: TxPlan.from(yaml) -> QuickTxBuilder(static
  utxoSupplier, () -> protocolParams, null) -> compose(plan) -> build()
  -> CBOR. Fully offline, never submits. Reuses StaticUtxoSupplier.
- New entrypoint signature: ccl_quicktx_build(thread, yaml, utxos_json,
  protocol_params_json); result stays {tx_cbor, tx_hash, fee} JSON.
- Add --initialize-at-build-time=org.yaml.snakeyaml for native-image.
- Rewrite QuickTxApiTest to build TxPlan YAML txs offline (payments,
  multi-intent, variable substitution, insufficient-funds).

Plutus script txs are deferred (no offline exec-unit evaluator in pre4).
Wrappers still target the old signature and are updated next.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Go wrapper: thin TxPlan YAML build + native reflection config
c633e2c 
End-to-end proven: TxPlan YAML -> offline build -> CBOR, through the Go
wrapper and native lib.

- Go: delete the fluent builder (~1640 LOC); QuickTxApi.Build(yaml, utxos,
  protocolParams) marshals the chain data to JSON and calls the new
  3-arg ccl_quicktx_build. Example rewritten to TxPlan YAML.
- native-image: add reflect-config for CCL's TxPlan deserialization
  classes (TransactionDocument + nested, all 27 intent classes, Amount).
  Jackson cannot construct these reflectively in a native image otherwise
  ("Cannot construct instance of TransactionDocument").
- QuickTxApi now surfaces the wrapped root cause in error messages
  (the generic "Failed to deserialize YAML" hid the real problem).

Verified: `go run ./examples/transaction` builds + signs a payment from
YAML offline. Go tests + the other wrappers are updated next.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Go tests: migrate QuickTx tests to the TxPlan YAML API
b90494b 
- ccl_test.go: replace the builder-based QuickTx unit tests with TxPlan
  YAML builds (simple/multi payment, variable substitution, insufficient
  funds); keep the reusable testProtocolParams/makeUtxos/assertTxResult
  helpers and all non-QuickTx tests.
- quicktx_integration_test.go: rewrite the DevKit build->sign->submit
  tests to YAML; drop the provider-config tests (provider is deferred).
- README: document bridge.QuickTx.Build(yaml, utxos, protocolParams);
  remove the deleted Amount helpers.

go vet clean; go test green (unit builds offline, integration skips
without DevKit).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Return the build result as YAML instead of JSON
041870a 
Per decision, ccl_quicktx_build now returns {tx_cbor, tx_hash, fee} as a
YAML document (via CCL's YamlSerializer), matching the YAML input.

- core: QuickTxService serializes the result with YamlSerializer;
  QuickTxApiTest parses it with the YAML mapper.
- Go: TxResult uses yaml tags; Build parses the result with gopkg.in/yaml.v3
  (aliased goyaml to avoid clashing with the `yaml` parameter). Adds the
  yaml.v3 dependency.

Verified: core tests green; `go run ./examples/transaction` round-trips
YAML in -> YAML out (build + sign); go test green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Python wrapper: migrate to thin TxPlan YAML build
843707e 
Same pattern as Go. Delete the ~1700-line fluent builder (quicktx.py)
and the provider module; QuickTx.build(yaml, utxos, protocol_params)
calls the 3-arg ccl_quicktx_build and parses the YAML result via pyyaml.

- _ffi.py: ccl_quicktx_build argtypes -> 3 char* (yaml, utxos, pp).
- __init__.py: export only QuickTx (drop builder/provider classes).
- pyproject.toml: add pyyaml dependency.
- tests: test_quicktx.py + test_quicktx_integration.py rewritten to YAML;
  delete provider/compose/new-features integration tests (builder/provider
  based — YAML equivalents are a follow-up; the core governance fixes stay
  in the Java).
- example + README updated to the YAML API.

Verified: 47 passed / 8 skipped (integration skips without DevKit);
example builds + signs a payment from TxPlan YAML offline.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Rust wrapper: migrate to thin TxPlan YAML build
5ea5344 
Same pattern as Go/Python. Delete the ~2140-line fluent builder from
lib.rs (TxBuilder/ScriptTxBuilder/ComposeTxBuilder/Amount/etc.);
QuickTxApi::build(yaml, utxos, protocol_params) calls the 3-arg
ccl_quicktx_build and parses the YAML result via serde_yaml.

- ffi.rs: ccl_quicktx_build -> 3 char* args.
- Cargo.toml: add serde_yaml; drop now-unused imports.
- tests: integration_test.rs + quicktx_integration_test.rs QuickTx tests
  rewritten to YAML (offline unit + DevKit), provider/metadata dropped.
- example + README updated to the YAML API.

Verified: cargo test green (29 offline tests incl. YAML builds; DevKit
tests skip); `cargo run --example transaction` round-trips YAML in -> out.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
JS wrapper: migrate to thin TxPlan YAML build
e4fb1c0 
Last wrapper. Delete the ~1300-line fluent builder from index.js
(TxBuilder/ScriptTxBuilder/ComposeTxBuilder/Amount/etc.) and the
provider module; QuickTxApi.build(yaml, utxos, protocolParams) calls
the 3-arg ccl_quicktx_build and parses the YAML result via the `yaml`
package.

- index.js: ccl_quicktx_build FFI -> 3 cstrings; drop provider export.
- package.json: add `yaml` dependency.
- index.d.ts: replace builder/provider types with the thin QuickTxApi.
- tests: ccl.test.js QuickTx tests + quicktx.integration.test.js rewritten
  to YAML; delete provider/compose/new-features integration tests.
- example + README updated to the YAML API.

Verified: bun test green (47 pass / 0 fail; integration skips without
DevKit); `bun examples/transaction.js` round-trips YAML in -> out.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
CI: run on develop too (push + pull_request)
1133d53 
PRs into develop (e.g. the TxPlan refactor) got no CI because the
workflows only triggered on main. Add develop to both the unit CI and
the DevKit integration-tests triggers so feature->develop PRs run the
full matrix.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
docs: rewrite quicktx.md for TxPlan (YAML)
1c4c0c0 
The doc described the deleted bespoke JSON spec. Rewrite it for the
TxPlan YAML flow: the new ccl_quicktx_build(yaml, utxos, protocol_params)
signature, the YAML result, the TxPlan document structure, the real
intent `type` discriminators, verified payment/variable examples, the
caller-supplied UTXO/protocol-params JSON, and per-wrapper build+sign
snippets. Notes provider removal and the deferred Plutus path.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
CI: install wrapper YAML deps (pyyaml, bun yaml)
270b946 
The first full PR run went red: Python tests failed with
ModuleNotFoundError: yaml, and the JS tests with Cannot find package
'yaml' — CI never installed the YAML parsers the wrappers gained in
this refactor.

- ci.yml + integration-tests.yml: pip install pyyaml alongside pytest.
- wrappers/js/build.gradle: `bun install` before `bun test` (unit +
  integration) so the `yaml` package is present.

Go (yaml.v3) and Rust (serde_yaml) are fetched automatically by
go test / cargo test.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Test metadata intent (verified shape) + document it
3412bdd 
The metadata intent's value is a scalar string the deserializer
auto-detects; passing it as a JSON string ('{"674": {...}}') is the
working shape (the earlier nested-map attempt threw a YAML parse error).

- QuickTxApiTest: add paymentWithMetadata — builds a payment+metadata
  TxPlan and asserts the tx body carries an auxiliary data hash, i.e.
  the metadata is actually attached (not just parsed). Confirmed in the
  native lib too (reflect-config already covers the metadata classes).
- docs/quicktx.md: add the verified metadata YAML example.

Closes the deferred metadata follow-up.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Plutus script transactions: offline build via caller-supplied exec units
60c5876 
Plutus spends/mints need each redeemer's execution units (mem + CPU
steps), which requires running the script in a UPLC evaluator. Rather
than embed an evaluator, the bridge takes the units as a fourth
caller-supplied input — exactly like UTXOs and protocol parameters — and
wires CCL's StaticTransactionEvaluator to stamp them onto the redeemers,
fully offline. The caller computes them with whatever they like (Ogmios,
Blockfrost, Aiken, Scalus). See TODO §2b for the planned pick-and-choose
evaluator helpers/examples.

- core: QuickTxService.buildTransaction gains exec_units_json -> List<ExUnits>
  -> withTxEvaluator(StaticTransactionEvaluator); ccl_quicktx_build entrypoint
  is now 4-arg (yaml, utxos, protocol_params, exec_units).
- native-image: register the Plutus reflection that Jackson needs — RedeemerTag
  / PlutusVersion enums (their @JsonProperty string forms), the plutus.spec
  types, and the plutus.spec.serializers (custom PlutusData ser/deser). Without
  these a script build fails in the native image though it passes on the JVM.
- wrappers: optional 4th arg through all four — Python/JS default param, Go
  variadic, Rust Option<&Value>. Existing 3-arg calls unchanged (Go/JS/Python);
  Rust call sites pass None.
- tests: QuickTxApiTest builds a real Plutus mint (always-succeeds V2 policy)
  offline and asserts the redeemer carries the supplied units, and that it
  fails without them; Python wrapper has the same pair. Verified end-to-end in
  the native lib (build OK with units, -10 without).
- docs/quicktx.md + TODO.md updated.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Test governance/staking/DRep/voting intents end-to-end (Go + fixtures)
2698794 
Restores the coverage dropped in the builder->TxPlan migration, where it
matters most: the native + wrapper path (the JVM/TxPlan layer is CCL's own,
covered upstream).

- QuickTxIntentsTest (JVM): builds each op with CCL, serializes via
  TxPlan.from(tx).toYaml(), builds it through the bridge, and emits the
  exact YAML to build/intent-yamls/<name>.yaml as a fixture. Covers
  stake_registration/deregistration/delegation/withdrawal, donation,
  drep_registration/deregistration/update, voting, voting_delegation,
  governance_proposal (11).
- wrappers/go/ccl/intents_test.go: table-driven test that drives every
  testdata/intents/*.yaml fixture through the native library via the Go
  wrapper, asserting tx_cbor/tx_hash/fee. This is the real bridge check —
  it confirms the native-image reflection config covers the governance/
  staking/DRep classes (it does; these intents serialize to string fields).

11/11 pass end-to-end in the native lib. Pools + Plutus spend next.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Cover stake pools end-to-end (Go) + register cert reflection
7ec4eff 
pool_registration / pool_update / pool_retirement added to the intent
fixtures and Go E2E. Pool registration surfaced native-image reflection
gaps the JVM test cannot — exactly what the Go E2E is for:
- util.serializers (HexToByteArrayDeserializer / ByteArrayToHexSerializer
  / InetAddress*) for the byte-array + relay-IP fields, and
- transaction.spec.cert.* + spec.UnitInterval for the certificate /
  margin types.

All 14 intent fixtures now build end-to-end through the native lib via Go.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Cover Plutus script spend (script_collect_from) end-to-end (Go)
4d454ba 
A Plutus spend collects from a script-address UTXO with a redeemer +
datum and attaches the spending validator. QuickTxIntentsTest builds it
and emits the fixture; script_spend_test.go drives it through the native
lib with the script UTXO (+ datum hash), a fee/collateral UTXO, and the
caller-supplied execution units — asserting it builds with units and
fails without. Reuses the Plutus reflection config from the mint (no
new native config needed).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Cover native mint / native-script / explicit + reference inputs (Go E2E)
f2cc845 
Adds the last intent types to the fixtures + Go end-to-end table:
native `minting` (NativeScript policy), `native_script` attachment,
`collect_from` (explicit input selection), and `reference_input`
(read-only inputs). The Go table now supplies a second small UTXO for
the reference-input fixture to read.

Every TxPlan intent type is now exercised end-to-end through the native
library via Go (18 in the table + Plutus spend + the payment/metadata/
mint cases elsewhere).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Cover Plutus script mint end-to-end (Go) — full intent parity
8ec0ecc 
Adds script_minting to the fixtures + a Go E2E test (build with exec
units, fail without), mirroring the spend. Go now exercises every TxPlan
intent type end-to-end through the native library.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Cover metadata intent end-to-end (Go) — 100% intent parity
64fff8a 
Adds a metadata fixture (payment + attached CIP-20 metadata) to the Go
E2E table. Every TxPlan intent type is now exercised end-to-end through
the native library via Go.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Intent E2E parity across all wrappers (Python, Rust, JS) + shared fix…
c35624e 
…tures

Brings the per-intent native end-to-end coverage that Go had to every
wrapper, so each language proves it builds all TxPlan intents through the
native library.

- test-fixtures/quicktx-intents/: shared, version-controlled fixtures
  (19 generic intents + plutus/{script_minting,script_collect_from}),
  generated by the JVM QuickTxIntentsTest. Single source of truth.
- Go: repointed intents_test.go / script_spend_test.go at the shared dir;
  dropped the duplicated wrappers/go/ccl/testdata copies.
- Python (test_quicktx_intents.py), Rust (intents_test.rs), JS
  (intents.e2e.test.js): table-driven tests over the shared fixtures +
  dedicated Plutus mint/spend tests (build with exec units, fail without).
- JS gradle test task now also runs intents.e2e.test.js.

All four wrappers: every intent type builds end-to-end. No native changes
needed — the reflection config proven by the Go suite already covers it.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@matiwinnetou matiwinnetou changed the title WIP: Replace bespoke JSON tx-spec with CCL TxPlan (YAML), build offline Replace bespoke JSON tx-spec with CCL TxPlan (YAML): offline build, Plutus, full intent coverage Jun 12, 2026
Mateusz Czeladka and others added 10 commits June 12, 2026 14:39
@claude
docs: update CCL version references to 0.8.0-pre4
ba4552b 
The bridge has been on 0.8.0-pre4 since the TxPlan refactor, but README,
CLAUDE.md, and TODO.md still said 0.7.2. Mark the 0.7.2->0.8.0 upgrade
item done and re-frame TODO §6 (the upstream modules are all available on
the current dependency now).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Cover compose mode (multiple senders -> one tx) end-to-end
5e143cf 
Compose was a feature of the old bespoke format that the TxPlan migration
left untested. TxPlan's `transaction` list supports it natively (multiple
`tx` entries, each with its own `from`, one fee_payer).

- QuickTxIntentsTest.compose builds a 2-sender compose via
  TxPlan.from(List<AbstractTx>).toYaml() and emits the compose.yaml fixture.
- All four wrapper intent tables pick it up; each supplies a second
  sender's UTXO so the compose builds end-to-end through the native lib.
- docs/quicktx.md gets a compose example.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
Stake/DRep-key signing: ccl_account_sign_tx_multi + wrappers
f356ef6 
Governance/staking intents build offline but couldn't be submitted: their
certificates must be witnessed by the stake (or DRep) key, while
ccl_account_sign_tx adds only the payment key — the node rejects them with
MissingVKeyWitnessesUTXOW.

- core: new ccl_account_sign_tx_multi(..., keys) signs with any subset of
  payment/stake/drep/committee_cold/committee_hot (CCL Account.signWith*Key),
  applied in order. The original ccl_account_sign_tx is unchanged.
- wrappers: sign_tx_with_keys (Python/Rust), SignTxWithKeys (Go, variadic
  roles), signTxWithKeys (JS); keys as a list/CSV.
- tests: Go asserts payment+stake adds a witness vs payment-only and that an
  unknown role errors; Python has the parity test. Rust/JS bindings verified
  to compile + load.
- TODO.md marks the item done; docs/quicktx.md documents it.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-tests for 6 intents (build -> sign -> submit -> on-chain)
7eeb3a6 
Converts "builds offline" into "a real node accepts it" for the
straightforward intents: stake_registration, drep_registration, donation,
governance info proposal, metadata, and Plutus mint.

Each resets the devnet, funds the fixed test account, builds the intent's
fixture with its real UTXOs, signs with the required key roles (e.g.
payment+stake for staking, payment+drep for DRep, payment for the rest),
submits, and asserts the tx is retrievable on-chain. Skips when DevKit is
not running, so it runs only in the CI "Integration Tests (DevKit)" job.

Native mint (no-key policy) and Plutus spend (lock-then-spend) follow once
this batch confirms the pattern on CI.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-tests: verify via node acceptance; fix donation + deposits
f7f8cdf 
First CI run surfaced real issues (all fixable):
- stake_registration / metadata / Plutus mint already submitted fine; the
  on-chain check used a garbled hash (the devnet returns a chunked body, so
  devkitSubmitTx's "hash" was the chunk-size prefix). Verify via submit
  success (HTTP 200/202 = the node validated + accepted the tx) instead.
- donation: the Conway cert asserts the stated treasury equals the chain's;
  regenerate the fixture with currentTreasuryValue=0 (fresh devnet).
- drep_registration / governance_proposal: DevKit's /epochs/parameters omits
  drep_deposit / gov_action_deposit; inject them so the build can compute the
  certificate deposits (the node validates them on submit).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-tests: set Conway deposits unconditionally
a575dbb 
drep_registration / governance_proposal still failed: DevKit's
/epochs/parameters returns drep_deposit and gov_action_deposit as null
(present, not absent), so the if-absent guard skipped the injection. Set
them unconditionally so the build can compute the certificate deposits;
the node validates the values on submit.

(stake_registration, metadata, Plutus mint, donation now pass on-chain.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-tests: register stake address before the info proposal
b8b7177 
5/6 now pass on-chain. The proposal failed with ProposalReturnAccountDoesNotExist:
a Conway proposal's deposit-return account must be a registered stake address.
Register it first, then submit the proposal in the next block.

Refactor: extract devnetPP() (params + Conway deposits) and signSubmit()
(build->sign->submit) so the proposal can run two sequential txs on one devnet.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-test for native mint (no-signature policy)
2c294b1 
The minting fixture used a random policy key, so the account couldn't sign
it for submission. Regenerate it under an empty ScriptAll policy (script_hex
820180) that requires no signature, and add TestIntegrationNativeMint which
the fee payer alone submits.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-test for Plutus spend (lock-then-spend)
4154be7 
Adds a plutus_lock fixture (payToContract: pays a UTXO to the
always-succeeds script address carrying the datum hash) and
TestIntegrationPlutusSpend, which:
  1. locks 10 ADA at the script address,
  2. finds the locked UTXO on-chain,
  3. repoints the script_collect_from fixture's utxo_ref at it, and
  4. spends it with the caller-supplied execution units.

Completes the 8 straightforward submit-tests (the other 6 + native mint
are green on-chain).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-tests for the prerequisite-heavy intents
a34020c 
Adds a submit-test per remaining concern, doing the on-chain setup each
certificate requires (sequenced txs on one devnet):

- voting_delegation  : register stake -> delegate voting power (to abstain)
- drep_update         : register DRep  -> update DRep
- drep_deregistration : register DRep  -> deregister DRep
- stake_withdrawal    : register stake -> withdraw (zero) rewards
- stake_delegation    : repoint the fixture at a real devnet pool, register+delegate
- voting              : register DRep + stake -> submit info proposal (its build
                        tx hash is the gov action id) -> vote on it
- pool_registration   : key the pool to the account's stake key (operator/owner/
                        reward account) so the stake-key signature witnesses it;
                        register the reward stake address first

Also inject pool_deposit (DevKit returns it null). Blind, CI-verified.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@claude
DevKit submit-tests: fix stake withdrawal/delegation + add depth checks
c058657 
Fixes (from the last run's ledger errors):
- stake_withdrawal: Conway requires the stake address to be vote-delegated
  before it can withdraw, so register stake -> delegate voting -> withdraw.
- stake_delegation: DevKit exposes no pool-list endpoint (/pools 404), so
  register a pool keyed to the account and delegate to it (the fixture is
  now delegate-only; the pool id is captured from StakePoolId).

Depth (turn "node accepted" into "verifiably did the thing"):
- native mint / Plutus mint now assert the minted asset is present at the
  receiver (assertMintedAssetAt).
- Plutus spend asserts the locked script UTXO was actually consumed.
- TestIntegrationDRepKeyRequired: a DRep registration signed with the
  payment key only must be rejected by the node, proving the extra
  sign_tx_with_keys witness is genuinely required.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@matiwinnetou matiwinnetou marked this pull request as ready for review June 12, 2026 15:40
@matiwinnetou matiwinnetou merged commit 1310a3a into develop Jun 12, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@matiwinnetou