Inline pallets 2

[illuzen]

Inline several pallets

• transaction-payment
• assets and assets-holder
• referenda
• conviction-voting
• ranked-collective

---

Note

Medium Risk
Large vendored copies of fee, asset, and governance pallets increase maintenance and merge risk, but behavior should match upstream unless later patched locally.

Overview
Vendors several Polkadot SDK FRAME pallets into pallets/ and wires the workspace to use local path dependencies instead of crates.io version pins for pallet-assets, pallet-assets-holder, pallet-conviction-voting, pallet-ranked-collective, pallet-referenda, and the pallet-transaction-payment family (including RPC crates).

The workspace member list and root Cargo.toml / Cargo.lock are updated accordingly (e.g. impl-trait-for-tuples, sp-rpc, and extra dev-dependencies such as pallet-balances / sp-io on vendored crates). Vendored crate manifests note that upstream pallet macros still expand to expect/unwrap, so workspace restriction lints are not applied there.

For assets, this PR adds the full in-tree pallet-assets implementation plus pallet-assets-holder, which implements BalanceOnHold and fungibles hold traits so other pallets can reserve asset balances; pallet-assets is configured to use Holder for hold/freeze-aware transfers and account teardown.

^{Reviewed by Cursor Bugbot for commit 5bb6f43. Configure here.}

[n13]

I verified the inlined pallets against the actual upstream releases pulled from crates.io (the published polkadot-sdk source at each stated version). Here's the result.

Verdict: ✅ No production code was changed

All non-test source files are byte-for-byte identical to upstream. The only src/ deviations are 3 test/mock files, each matching a documented "fix tests" commit.

Methodology

Downloaded each crate at its declared version from crates.io and ran a recursive diff against the inlined src/ trees:

Crate	Version	src/ result
pallet-transaction-payment	45.0.0	identical
pallet-transaction-payment-rpc	48.0.0	identical
pallet-transaction-payment-rpc-runtime-api	45.0.0	identical
pallet-assets	48.1.0	identical (lib, functions, impl_fungibles, types, weights, benchmarking, migration, mock, tests…)
pallet-assets-holder	0.8.0	only tests.rs differs
pallet-referenda	45.0.0	only mock.rs + tests.rs differ
pallet-conviction-voting	45.0.0	identical
pallet-ranked-collective	45.0.0	identical

File sets also match exactly (no added/dropped source files; upstream's packaged Cargo.lock is correctly dropped).

The only source changes (test-only, all justified)

1. assets-holder/src/tests.rs — splits the died_fails_if_holds_exist #[should_panic] test by build profile, because defensive_assert! only panics under debug_assertions but CI runs --release. Matches commit 5bb6f43.

2. referenda/src/mock.rs — rewires the pallet_scheduler::Config mock to the in-tree timestamp-aware scheduler (Moment / TimeProvider / TimestampBucketSize instead of RuntimeEvent / BlockNumberProvider). Matches commit b765f02.

3. referenda/src/tests.rs — imports qp_scheduler::BlockNumberOrTimestamp and updates two alarm assertions from (5, 0) to BlockNumberOrTimestamp::BlockNumber(5). Also drops one stray println!("Agenda #6: …") debug line (harmless, slightly beyond the strict scope of the commit message).

These are unavoidable: the upstream tests target the vanilla scheduler, but this repo already vendors a timestamp-aware scheduler.

Manifest / workspace changes (non-functional)

• Each pallet Cargo.toml: crates.io/monorepo deps → workspace/path; [lints] workspace = true dropped (replaced with a comment, since upstream macros expand to expect/unwrap); cosmetic field/feature reordering. Feature lists (std/runtime-benchmarks/try-runtime) are content-identical to upstream. referenda adds a qp-scheduler dev-dep for the test fix.
• Root Cargo.toml: adds the 8 crates as members, switches their deps from version pins to path, and adds impl-trait-for-tuples + sp-rpc workspace deps. Declared versions match the inlined versions exactly.
• Cargo.lock: consistent (drops registry source for the now-local crates, adds dev-deps).

Minor notes (not blockers)

• assets-holder dev-deps flip default-features = false → true for pallet-balances/sp-core/sp-io. Dev-deps only, so it can't affect runtime/WASM behavior.
• The removed println! in referenda/tests.rs isn't mentioned in the commit message, but it's just debug noise.

Bottom line: the inlining is clean — production logic is an exact copy of the upstream releases, and the only edits are the minimum test/mock adaptations needed to compile against this repo's existing in-tree scheduler and --release CI.

[n13]

LGTM!