Release Zaino 0.6.0 (Ironwood)#1401
Merged
Merged
Conversation
`podman volume inspect` reads podman's database, not the filesystem, so a named volume can exist in the DB while its backing directory is gone (after a partial ~/.local/share/containers cleanup or an interrupted `podman system reset`). The old inspect-only guard then skipped `podman volume create`, and the later container-test mount failed with "failed to validate if host path is dangerous: lstat .../volumes/zaino-container-target: no such file or directory", producing no nextest summary for the live partitions. Guard on the actual Mountpoint directory instead: when the DB record is missing or its backing dir is gone, `podman volume rm --force` drops the dangling record before recreating it, keeping the DB and on-disk storage in sync. Recovers automatically on the next `makers test` rather than requiring a full `podman system reset`. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two production sites hard-coded the finalised/non-finalised seam depth as the pre-zebra-10 value. zebra-10 raised MAX_BLOCK_REORG_HEIGHT 99->1000, so the real NON_FINALIZED_DEPTH is 1001 and both were wrong: - non_finalised_state.rs: the reorg-detection recursion bound used `height + 100`, aborting on any reorg deeper than 100 blocks; now uses MAX_NFS_DEPTH, matching its sibling bounds at 553/863. - chain_index.rs get_compact_block_range: the cache/finalised handoff used `tip - 99`, routing non-finalised blocks (tip-NON_FINALIZED_DEPTH .. tip-99) to the finalised stream that does not hold them; now derives from finalized_height_floor(tip). Unchanged for chains shorter than the depth (tip-99 == floor+1 at depth 100). Also update production doc comments describing the seam as a literal "100 blocks" to reference NON_FINALIZED_DEPTH / MAX_BLOCK_REORG_HEIGHT. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…e maturity Add a consensus module holding the workspace's authoritative, upstream-derived constants, so nothing hard-codes them: - MAX_NONFINALISED_DEPTH = zebra_chain MAX_BLOCK_REORG_HEIGHT + 1 - COINBASE_MATURITY = zebra_chain MIN_TRANSPARENT_COINBASE_MATURITY - FAST_TEST_MAX_NONFINALISED_DEPTH = MAX_NONFINALISED_DEPTH / 10 (tractable test seam) Both consts live in zaino-common, which already depends on zebra-chain, so this adds no new workspace dependency. Re-source zaino-state's NON_FINALIZED_DEPTH from the mod (production = real MAX_NONFINALISED_DEPTH; in-crate tests select the derived FAST_TEST value so short mock fixtures still exercise a moving seam), and the tx_out_set_accumulator maturity split from COINBASE_MATURITY. Behaviour-preserving: in-crate test depth stays 100, production stays 1001; the values are now derived from a single upstream source rather than literals. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… seam The FAST_TEST_MAX_NONFINALISED_DEPTH shrink was only reachable by zaino-state's own cfg(test) build. Cross-crate live tests (e2e, clientless) drive a real zaino-state instance whose seam depth is a compile-time constant, so the only way for them to use the tractable seam is a build feature. Add fast-test-seam to zaino-state: NON_FINALIZED_DEPTH now selects FAST_TEST under cfg(test) OR the feature, else the real MAX_NONFINALISED_DEPTH. e2e and clientless enable it via [dev-dependencies] only; resolver = "2" keeps it out of production graphs (verified: zainod's zaino-state does not carry the feature) and default-members excludes the live-test crates from production builds. This un-vacuums the clientless chain_cache eviction/finalisation tests: at depth 100 their ~150-block chain has a real finalised region instead of saturating to genesis at the production depth of 1001. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The zaino-state-local const is a cfg/feature selector — it resolves to the authoritative MAX_NONFINALISED_DEPTH in production and to FAST_TEST_MAX_NONFINALISED_DEPTH under cfg(test)/fast-test-seam — not a source of truth. The old name read like a source and diverged in spelling from the finalised_state module (FINALIZED vs FINALISED). OPERATIONAL_NFS_DEPTH names it as the effective/selected depth the code operates on. Pure rename, compiler-verified (cargo check -p zaino-state --tests); behaviour unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
e2e and clientless live tests run zaino-state under fast-test-seam, so their seam-relative boundaries and prose must reference the mod's FAST_TEST_MAX_NONFINALISED_DEPTH (the depth those instances actually use), not hard-coded literals or the retired NON_FINALIZED_DEPTH name. - e2e get_outpoint_spenders: FINALITY_DEPTH = FAST_TEST_MAX_NONFINALISED_DEPTH + 5. - e2e prose (send_to_transparent_finalization, get_outpoint_spenders ignore reason): retarget the stale NON_FINALIZED_DEPTH mentions to the seam const. - clientless chain_cache: express the finalised-floor / retention boundaries as FAST_TEST_MAX_NONFINALISED_DEPTH +/- test margins (values preserved). MAX_NFS_DEPTH stays a zaino-state internal (retention window = seam + margin). Value-preserving and cargo-check clean; no NON_FINALIZED_DEPTH remains in live-tests. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The load-bearing advance mined 99 blocks to push a send into the finalised DB, but a send at height H finalises only once the tip reaches H + seam. At the fast-test-seam depth (100) a 99-block advance falls one block short, so the test would still query the non-finalised chain if un-ignored. Advance FAST_TEST_MAX_NONFINALISED_DEPTH + 5 blocks (single-sourced, crosses the seam with margin), and make the surrounding prose seam-relative so it can't drift. Still #[ignore]d (waits on cheap transparent filler, round-3 P2). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The zcashd matrix (devtool_zcashd.rs, cfg feature zcashd_support) carries twins of the zebrad finalization tests, and cargo check without that feature never compiled them — so the send_to_transparent_finalization 99-block advance had the same off-by-one, and send_to_all_finalization used a magic 100-block advance. Both now advance FAST_TEST_MAX_NONFINALISED_DEPTH + 5 (single-sourced, crosses the seam), with the prose made seam-relative. Verified with cargo check -p e2e --tests --features zcashd_support. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The round-2 P1 / round-3 P2 / round-3 P3 / round-2 P0 tags reference an internal roadmap document and are meaningless in the source. Replace each with the plain capability it named — devtool transparent-coinbase shielding, cheap transparent filler-block mining, unconfirmed (mempool) wallet balances, zcashd activation-height alignment — or drop the tag where the surrounding text already states the fact. Compiles default and with --features zcashd_support; fmt clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The front-door test set is named for the `packages/` directory it runs, so the plural is the faithful name. Hard-cut: `makers test package` now errors with `unknown set` rather than aliasing. Updates every reference site: Makefile.toml (default, case matcher, dispatch, descriptions, comment block), help.sh, docs/testing.md, README.md, and the live-tests/CONTEXT.md glossary term (now records the retired singular under _Avoid_). ADR-0004 gets a one-line supersession footer; its historical text is preserved. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Harden init-podman-volumes against stale volume records
Fix nfs depth bug
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Phase 3 of the zingo-common disintegration. zaino-common loses its direct dependency on the shared vocabulary crate: the two From impls in config/network.rs move to zaino-testutils as named boundary functions targeting zcash_local_net's re-exported types, following the existing local_network_from_activation_heights precedent. The zcash_local_net pin moves to the add_zingo_consensus superset rev, which contains all of add_client_support plus the zingo-consensus migration. Production zaino now has no dependency on the activation heights vocabulary at all. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Move the infrastructure pin from the add_zingo_consensus branch (0a003b4f) to b09d9c11, the head of remove_unnecessary_deps. The new commit replaces zcash_protocol::PoolType with zingo_consensus::MinerPool as the harness's mine-to-pool selector, so the pin bump carries the matching rename through zaino-testutils (re-export, SHIELDED_FUNDING_POOL, default_mining_pool, the mine_to_pool parameters) and the five e2e call sites. The proto-side PoolType enum in zaino-proto is unrelated and untouched. Only the three infrastructure-sourced crates move in the lock; zcash_protocol drops out of zcash_local_net's dependency list. Verified with cargo check --tests on zaino-testutils/e2e/clientless under both default features and zcashd_support; fmt clean; clippy shows only pre-existing warnings. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Follow zingolabs/infrastructure#276 to its current head (029e93eb) by switching the git dependency from an exact rev to branch tracking; the committed Cargo.lock still pins the resolved commit, so builds stay deterministic until the next cargo update. The PR empties zcash_local_net's dependency tables of the zebra and librustzcash stacks, which drops an entire duplicate zebra stack from our lock: zebra-chain 9.0.0, zebra-rpc 9.0.0, zebra-node-services 7.0.0, zebra-consensus/network/script/state 8.x, and tower-batch-control/tower-fallback. No zaino code changes were needed — the MinerPool rename in the previous commit was the only API break. Verified with cargo check --tests on zaino-testutils/e2e/clientless under both default features and zcashd_support; fmt clean; clippy shows only pre-existing warnings. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Pure lock bump following new pushes to zingolabs/infrastructure#276: getset replaced by hand-written accessors (dropping getset from zcash_local_net's dependency list), the zebra-rpc oracle dev-dependency deleted, and an external-types allowlist fix. No zaino API impact. Verified with cargo check --tests on zaino-testutils/e2e/clientless under both default features and zcashd_support. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Runs `cargo publish --dry-run` for all publishable crates in dependency order. Advisory (warn-only) on regular PRs; hard-fails on rc/* and stable branches to catch unpublishable patch overrides before release.
remove zebra and zcash* from infrastructure
refactor(state): apply ironwood-migration polish from review of #1379
The new rev (85502cd0) picks up the twelve commits on the bump_to_NU6.3 branch since the previous pin, including deterministic listener-port allocation with an Indexer-convergence barrier, the generic Wallet abstraction, and the front proxies published as the canonical endpoints. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
DRY zaino-common
The interesting resolutions all follow from dev's unification of
FetchService/StateService into NodeBackedIndexerService meeting this
branch's rule that the validator is the sole source of activation-height
truth (zaino#1076):
The network-adoption logic that lived in the deleted backends module
moves into ValidatorConnector::{spawn_fetch,spawn_state}, which now
return the adopted runtime network alongside the connector, and
activation_heights_from_upgrades moves (with its tests) into
validator_connector.rs. source::State carries the adopted
zebra_chain::parameters::Network together with dev's new
chain_tip_change and sync_task_handle fields.
The v1.3.0 migration keeps dev's PoolActivationHeights::resolve shape,
resolved against the runtime network the config now carries directly.
Test-only construction sites replace the removed payload-carrying
Network::Regtest kind with ActivationHeights::to_regtest_network().
The generic live-test suites (ironwood_activation, validator_heights)
move from FetchService/StateService type parameters to the Rpc/Direct
ValidatorConnectionMarker types, matching dev's ported devtool suite.
dev's dry_zaino_common refactor (#1371) touched the payload-carrying Network type this branch deletes; the resolution keeps the branch's payload-free network kind in zaino-common.
The dev merge took the incoming Cargo.lock, which predates this branch keeping rustls in zaino-common (ADR-0006); cargo regenerates the entry on the first build. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The NetworkUpgrade-variant ↔ ActivationHeights-field correspondence is recorded exactly once, in zaino-common's activation_heights_mirror! macro, but activation_heights_from_upgrades restated it as a hand-written thirteen-arm match and both it and its tests restated the all-None schedule as field-by-field literals. The macro now generates the two missing derived items — the NEVER_ACTIVATED schedule and the per-upgrade slot_mut accessor — and the adoption function collapses onto them. The generated match stays exhaustive, so a new zebra upgrade variant still fails the build until its pair is added to the mirror. Also remove the dead ChainIndexConfig::new: it had no callers and was the one constructor shape that invited supplying a network from somewhere other than validator adoption; from_backend_config is now the sole named production constructor (zaino#1076). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Public-network activation heights come only from checked-in, ecosystem-shared consensus code; regtest parameters are configured on the validator at scenario launch and every other component adopts them from it. adopt_network previously enforced the regtest half but took the compiled Mainnet/Testnet parameters on faith: a validator running a *configured* testnet (zebra supports custom testnet activation heights) would silently drift from the index — the exact bug class this branch exists to kill, on a different network kind. adopt_network now fetches getblockchaininfo on every network kind and, on Mainnet and Testnet, verifies each reported (upgrade, height) pair against the compiled parameters before trusting them, failing loud on the first disagreement. Only reported entries are checked, not report completeness: the upgrades map is branch-ID-keyed, so unconfigured upgrades are simply absent and an older validator may not know the newest ones. Also document two acceptance decisions the spec made but the code did not state: adoption happens exactly once at spawn (a validator restarted mid-session with a different schedule is out of scope), and before_overwinter is always None when adopted (BeforeOverwinter has no consensus branch ID, so it can never appear in the report; zebra's new_regtest supplies its handling). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The public shared-consensus test network and a hermetic regtest chain are categorically different things, and the bare word "testnet" has named both. Where the intention is the public network, the canonical vocabulary is now: "The Public Testnet" in prose, `the_pub_testnet` in identifiers, and `PubTestnet` in types. The CONTEXT.md glossary entry records the rule. The config kind renames to `Network::PubTestnet`. Its canonical config spelling becomes `"PubTestnet"`, with `#[serde(alias = "Testnet")]` keeping existing config files parsing; a zainod config test pins the legacy spelling. The Public-Testnet chain-cache static renames to ZEBRAD_THE_PUB_TESTNET_CACHE_DIR, and the boundary-walk suite renames to the_pub_testnet_ironwood_boundary (the Makefile ironwood filter follows). References to zebra's own types and config format (parameters::Network::Testnet, NetworkType::Testnet, zebrad.toml) are foreign vocabulary and keep their spelling. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
ChainWork::to_u256 takes self by value (the type is Copy, and clippy's wrong_self_convention expects to_* methods on Copy types to consume self), and the address-deltas test fixture uses i.is_multiple_of(2) in place of the manual modulo test. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
feat!: enforce the Validator as the sole source of activation-height truth
An audit of every zebra* and zcash* dependency edge (cargo machete, the compiler's unused_crate_dependencies lint, and inspection of each remaining single-reference hit) found exactly one dead edge: zainod declared zebra-chain but never referenced it. Remove it. Every other declared edge is genuinely used; zebra-chain remains the type backbone of the workspace, with some 350 references in zaino-state alone. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Advance ZEBRA_VERSION in .env.testing-artifacts from 6.0.0-rc.0 to the final 6.0.0 release. Zebrad 6.0.0 depends on zebra-chain ^11.1.0, the same crate generation the workspace now pins, so the zebrad binary the live tests run against and the zebra libraries zaino links are once again from a single release. The makers check-matching-zebras guard passes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A workspace-wide unused-dependency audit (the compiler's unused_crate_dependencies lint analyzed per target, cross-checked with cargo machete and a text search of every candidate's package, including feature-gated code) found twelve dead edges. Remove them: prost, bs58, nonempty, derive_more, and once_cell from zaino-state; tracing-subscriber from zainod; and tracing, corez, serde_json, tempfile, tower, and zip32 from clientless. The bs58, nonempty, and zip32 entries in [workspace.dependencies] lost their last inheritors and are removed with them. One edge that looks dead is kept deliberately: clientless does not import zainod in any test source, but its transparent_address_history_experimental feature forwards to zainod/transparent_address_history_experimental, and cargo rejects a feature forward to a non-dependency. A comment in the manifest now records that. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The same audit that removed the dead edges found three dependencies that are real but used exclusively under #[cfg(test)]: primitive-types in zaino-state (the legacy_chainwork_reference round-trip tests), tempfile in zainod (the config tests), and tonic in zaino-testutils (the cfg(test)-gated build_client helper). Move each from [dependencies] to [dev-dependencies] so shipped builds stop linking them. The lockfile is unchanged: dependency kind does not affect resolution. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…mary `makers test all` previously ran the packages set outside the summary runner, so the closing table covered only the two live partitions. The `all` branch now delegates to `live-summary --all`, which runs `makers container-test` first and prints its tallies as a `packages:` row above `clientless:`, `e2e:`, and `TOTAL:`, under a "test summary" header. `makers test live` is unchanged apart from the new column. The old contract is preserved: every set runs even when an earlier one fails, and the runner exits non-zero if any set did. Each row also gains a wall-clock time column, parsed from the nextest summary line's `[ 510.718s]` stamp. Durations are held as integer milliseconds in a u64 — whole seconds and a right-padded three-digit fraction parsed digit-by-digit, no floating point — and the TOTAL row sums them, which reads as true wall time because the suites run sequentially. Six new unit tests cover the parser (padding, short and absent fractions, and propagation through parse_summary), and the pre-existing `filter(..).last()` on a double-ended iterator is tightened to `rfind(..)` per clippy. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The test-environment image rebuild (first since the ZEBRA_VERSION bump changed the image tag) failed because Debian trixie point updates replaced two pinned package versions, and the archive does not keep superseded versions: protobuf-compiler 3.21.12-11 -> 3.21.12-11+deb13u1 (all three build stages) and curl 8.14.1-2+deb13u3 -> 8.14.1-2+deb13u4 (both stages). Every other pin in the Containerfile still matches the archive; verified against qa.debian.org/madison and a live trixie-slim container. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ut parsing Replace the duration_millis parser (which scraped nextest's human-readable summary line) with Instant::now()/elapsed() around each run_partition call. The nextest display format is not a stable contract; the runner already owns the process lifecycle so it can time it directly. This also captures build time, which the nextest stamp did not include.
PRs that target rc/** or stable carry unbumped dev code by design — version bumps happen on the RC branch after merge. The blocking mode was failing these PRs for an expected, non-actionable condition. Only direct pushes to rc/** or stable are blocking now; PRs targeting them are advisory.
…for-prs fix(ci): publish dry-run advisory for PRs targeting rc/stable
Advance rc/0.6.0 to dev for 0.6.0-rc.3
zaino-common 0.2.0 → 0.4.0-rc.3 zaino-fetch 0.2.1 → 0.4.0-rc.3 zaino-proto 0.1.3 → 0.2.0-rc.3 zaino-serve 0.3.1 → 0.5.0-rc.3 zaino-state 0.3.1 → 0.5.0-rc.3 zainod 0.4.3-ironwood.1 → 0.6.0-rc.3 Workspace dependency version pins updated to match.
build: bump crate versions for 0.6.0-rc.3
Merge rc/0.6.0 into stable and bump all crate versions to their final stable form: zaino-common 0.4.0-rc.3 → 0.4.0 zaino-fetch 0.4.0-rc.3 → 0.4.0 zaino-proto 0.2.0-rc.3 → 0.2.0 zaino-serve 0.5.0-rc.3 → 0.5.0 zaino-state 0.5.0-rc.3 → 0.5.0 zainod 0.6.0-rc.3 → 0.6.0 Resolves stable/rc divergence from the 0.5.1 hotfix commits that were never backported to dev (zebra git patch removal, version bumps).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Zaino 0.6.0 — Ironwood/NU6.3 release.
rc/0.6.0intostable, resolving the divergence from the 0.5.1 hotfixNonFinalizedBlocksListenerstruct variant regression from the mergeCrate versions
Highlights since 0.5.1
After merge, tag
0.6.0on the merge commit to trigger the release workflow.