tests: Wait for transaction to be in a finalized block

[lexnv]

The test fetch_block_and_decode_extrinsic_details failed on this CI job:

Rpc(ClientError(Call(ErrorObject { code: ServerError(2001), message: "Invalid block hash", data: None })))

The test is submitting an extrinsic and waiting for inclusion in any block.

Submitting an extrinsic with the unstable backend (rpc v2) implies:

• subscribing to chainHead for related block hashes
• calling transaction rpc method

It is entirely possible that the block hash reported by the transaction method was not reported yet by the chainHead class yet:

subxt/subxt/src/backend/unstable/mod.rs

Lines 513 to 524 in ff3a086

	rpc_methods::TransactionStatus::BestChainBlockIncluded {
	block: Some(block),
	} => {
	// Look up a pinned block ref if we can, else return a non-pinned
	// block that likely isn't accessible. We have no guarantee that a best
	// block on the node a tx was sent to will ever be known about on the
	// chainHead_follow subscription.
	let block_ref = match seen_blocks.get(&block.hash).cloned() {
	Some(block_ref) => block_ref.into(),
	None => BlockRef::from_hash(block.hash),
	};
	TransactionStatus::InBestBlock { hash: block_ref }

Given that the test was waiting for any block inclusion (wait_for_in_block), the test attempted to call chainHead_header before the block hash was reported.

To mitigate this, the wait_for_in_block is replaced by wait_for_finalized. The submit_transaction will return TransactionStatus::InFinalizedBlock only after the block hash has been reported by chainHead.
The downside of this approach is that the test may wait for a longer time (until the block becomes finalized + overhead of reporting the block via chainHead).

I was unable to reproduce the CI error in ~900 runs.

[jsdw]

Looks good! Yup, mostly wait_for_in_block should be avoided for exactly this reason :)

[niklasad1]

I was unable to reproduce the CI error in ~900 runs.

🤣