Fix subscribe blocks

[0xc0ffeebabeeth]

Motivation

Some RPC Providers (Alchemy and Reth were tested) don't have the "uncles" field in their newHeads subscription, this breaks the subscribe_blocks() stream.

Solution

Providing the serde(default) flag on the uncles field allows graceful handling of both existing and missing uncle fields in the subscription response.

PR Checklist

• Added Tests
• Added Documentation
• Breaking changes

[DaniPopes]

Thanks! Ref https://t.me/ethers_rs/36155
Pending @mattsse or @prestwich, should this also be skip_serializing_if?

[prestwich]

Thanks! Ref https://t.me/ethers_rs/36155 Pending @mattsse or @prestwich, should this also be skip_serializing_if?

Given that we are post-merge, does it now ALWAYS omit that field in the newHeads objects? The field is present in getBlockByNumber even for recent blocks, so the newHeads notification object is NOT a valid block object according to rpc spec? cc @mattsse what's reth do?

[prestwich]

my initial take is that we should NOT do skip_serializing_if, and just leave it as default

[DaniPopes]

Actually should this be Header instead of block?

@prestwich https://github.com/paradigmxyz/reth/blob/4e49aa9df66dfc19907c7e788f1817e83f2f61c9/crates/rpc/rpc/src/eth/pubsub.rs#L292

[prestwich]

Actually should this be Header instead of block?

this is quite likely

[0xc0ffeebabeeth]

The newHeads object does seem to contain withdrawals though, which aren't included in the Header object

[prestwich]

so its a separate object? 😮‍💨

[DaniPopes]

most consistent eth rpc api