Initialize beacon light client from checkpoint

[yrong]

Refactoring including:

• rename InitialUpdate to CheckPointUpdate
• rename HeaderUpdate to ExecutionHeaderUpdate

Extrinsic added:

• check_point_update
• deactivate_bridge

Essentially InitialUpdate is just a CheckPoint for initialization, it's possible we do checkpoint any time at runtime(e.g. checkpoint after a routine maintenance or emergency which cause bridge stopped for a long time).

Snowfork/cumulus#23 required

Fixes: SNO-429, SNO-430

[yrong]

Seems there is no Sudo or Democracy pallet in cumlus bridge, so would assume check_point_update which requires sudo authority can only be triggered from relaychain by XCM.

To make e2e test easy, we keep the GenesisConfig.

[alistair-singh]

requires sudo authority can only be triggered from relaychain by XCM

Working on a script for this. Will PR shortly.

[alistair-singh]

Working on a script for this. Will PR shortly.

#829

[yrong]

@alistair-singh Thanks, that's exactly what I need.
https://github.com/Snowfork/snowbridge/blob/feceb0aef2fd0143ec5db68fc6f52eae15603766/core/packages/test/scripts/configure-statemine.sh#L30C25-L50

[yrong]

E2E setup seems a bit too slow require almost 4 minutes even skip building steps, so speedup in this commit decreased to almost 90 seconds.
5a1263c

[vgeddes]

E2E setup seems a bit too slow require almost 4 minutes even skip building steps, so speedup in this commit decreased to almost 90 seconds.
5a1263c

I would rather not reduce the slot time. Clara reverted that optimization for several reasons I can't remember.

I don't think waiting 4 minutes for sync is too bad.

[yrong]

I would rather not reduce the slot time. Clara reverted that optimization for several reasons I can't remember.

It's different. Last time what we change is SLOTS_PER_EPOCH from 32 to 8 which will impact the generation of ancestry proof whereas change SECONDS_PER_SLOT won't.

[vgeddes]

I would rather not reduce the slot time. Clara reverted that optimization for several reasons I can't remember.

It's different. Last time what we change is SLOTS_PER_EPOCH from 32 to 8 which will impact the generation of ancestry proof whereas change SECONDS_PER_SLOT won't.

Does this mean geth will produce an execution header every 2 seconds? If so, this could cause the light client to fall behind importing execution headers every 2 seconds (due to the overhead of BLS signature verification).

[yrong]

Does this mean geth will produce an execution header every 2 seconds? If so, this could cause the light client to fall behind importing execution headers every 2 seconds (due to the overhead of BLS signature verification).

Yes, that's true. But the hack is for minimal spec and dev usage only. It did help for fast checking basic functions like beacon/beefy sync work as expected so I would prefer keep it.

Just add an env config here dc488f0 will not enable fast mode by default.

[vgeddes]

+1

[claravanstaden]

@vgeddes @yrong Just mentioning, I would rather we not deviate from the official spec minimal config, since it a standard spec that all the client implementation uses. Speeding up those configs may save time with testing, but we lose time with development since it has wasted some time figuring out why some proofs fail, only to find because the spec is different. I'd wager that more cases like this will occur if we deviate from the standard spec in the future.

[vgeddes]

@vgeddes @yrong Just mentioning, I would rather we not deviate from the official spec minimal config, since it a standard spec that all the client implementation uses. Speeding up those configs may save time with testing, but we lose time with development since it has wasted some time figuring out why some proofs fail, only to find because the spec is different. I'd wager that more cases like this will occur if we deviate from the standard spec in the future.

That's a good point @claravanstaden. I see that these config parameters are all tightly related, and decreasing the slot time may interfere with ancestry proofs and the SLOTS_PER_HISTORICAL_ROOT setting.

@yrong can you revert this slot change please. Since it is disabled by default, I think no one will actually use it anyway.