[AlphaSparrow]

Description

Fixes #2286.

TxGraph::insert_tx previously skipped every null previous output (OutPoint::null()) unconditionally when indexing spends, under the assumption that any null prevout implied the transaction was a coinbase.

However, under Bitcoin consensus rules, a transaction is only a coinbase if it has exactly one input and that input has a null outpoint (tx.is_coinbase()). A malformed transaction with a null prevout and another input is not a coinbase, yet its null input was skipped. Two such transactions sharing only the null prevout were therefore not recorded in spends, missed by direct_conflicts, and both erroneously admitted to CanonicalView.

This PR:

• Updates TxGraph::insert_tx to skip indexing spends only if tx.is_coinbase(). Non-coinbase transactions have all their inputs indexed, including null prevouts.
• Updates TxGraph::direct_conflicts to ignore coinbases, as valid coinbases have no previous outputs and cannot conflict on inputs.
• Adds regression tests in crates/chain/tests/test_null_outpoint.rs verifying conflict detection on null prevouts, eviction in CanonicalView, and isolation from valid coinbases.
• Includes a CI commit resolving the recent MSRV dependency breakage (yoke-derive pin drop) and check-wasm runner pin (ubuntu-22.04).

Notes to the reviewers

Valid coinbases remain strictly isolated from conflict tracking. Non-coinbase transactions (even malformed consensus-invalid ones) now have their spends indexed consistently so direct_conflicts and CanonicalView correctly detect conflicts.

Changelog notice

Fixed

• bdk_chain: Track spends and conflicts on null prevouts for non-coinbase transactions so conflicting consensus-invalid transactions are detected and pruned from CanonicalView. TxGraph doesn't index null-prevout spends of non-coinbase txs, so their conflicts go undetected #2286

Checklists

All Submissions:

• I followed the contribution guidelines

Bugfixes:

• This pull request breaks the existing API
• I've added tests to reproduce the issue which are now passing
• I'm linking the issue being fixed by this PR

[evanlinjin]

@AlphaSparrow can you explain how a valid non-coinbase tx will have a null prevout?

[AlphaSparrow]

@evanlinjin yeah you're right, it can't, as per the bitcoin consensus rules (bad-txns-prevout-null), any non-coinbase txn with a null prevout is invalid and would never be accepted by a node, let alone make it into the block/mempool.

I picked up #2286 to fix the internal conflict tracking inside TxGraph without properly taking into account the bitcoin consensus rules on the actual network. I noticed calculate_fee checks tx.is_coinbase(), whereas insert_tx was using prevout.is_null(), so I thought aligning them and resolving the edge case in #2286 made sense from a pure data structure standpoint, but i overlooked that the consensus already prevents these transactions from ever reaching a wallet in practice.

Given it's redundant in the real world, I'm completely fine closing this PR and #2286 .
(Happy to see the MSRV fix in the 2nd commit already helping CI fixes in #2438!)