Babe fetch epoch returns a complete epoch

[expenses]

This is to add the changes for paritytech/substrate#8072. I've tested it on a development substrate node and it works there, but on flaming fir I kept getting Err(BabeFetchEpoch(WasmStart(VirtualMachine(FunctionNotFound), HostVmPrototype))), despite the chain being fully synced. I'm unsure whether that was caused by this change or not.

[tomaka]

Oh I didn't realize that you changed the output of the runtime function.

[tomaka]

Just so you're aware: this will break GrandPa warp sync until we do a runtime upgrade?

[expenses]

I'm aware, I suggest we only merge after a runtime upgrade.

[expenses]

Polkadot has been updated so I think we can merge this right?

[expenses]

Ah, sample_decode in `babe_fetch_epoch failed. We should either remove this test or update the data.

[tomaka]

Only Westend has been updated. Polkadot and Kusama haven't.

Fortunately, since we (erroneously?) use decode instead of decode_all to decode the runtime output, the code in the main branch is still working with Westend, so there is no urgency to merge this.

[expenses]

Ah, okay.

[tomaka]

This is now ok to merge.

[tomaka]

So this unfortunately breaks GrandPa warp sync on Westend because the config has changed to VRF and we still receive headers with plain slot claims.

Asked André.

[tomaka]

So the story goes like this:

• A PR was made to Westend to change the type of slot from plain to VRF, but this unfortunately only changed the config that would be used to initially populate the genesis block, so it had no effect.
• When we implemented the Babe_epochConfig runtime API, the VRF enum was used (since that's what was written in the code), while in reality Westend continued using plain slots.

[tomaka]

André changed the config with a sudo call in block 5,567,609. It's working again.