Scheduler: remove empty agenda on cancel

[muharem]

Fixes #12977 #12851

With the small change to the script provided by @ggwpez, you can see there are only agendas for the future blocks with all None items.
Its possible because we do not remove such agendas on a cancelation, and they get removed when serviced.

From the implementation and tests it looks like it was expected to have empty future agendas, nevertheless it seems to me valid, to remove them on the cancelations and address the raised issues.

// DONE fix benches
// DONE migration to remove empty agendas

[kianenigma]

It seems like the original author intended to be like this, but seeing the data I also agree that we better not have all of these empty agendas.

I don't think we need a migration, the existing agendas would eventually be cleared up when they are in the past, right?

[muharem]

It seems like the original author intended to be like this, but seeing the data I also agree that we better not have all of these empty agendas.

I don't think we need a migration, the existing agendas would eventually be cleared up when they are in the past, right?

Yes, looks same to me.
And yes, they will be eventually removed.

[ggwpez]

With the small change to the script provided by @ggwpez, you can see there are only agendas for the future blocks with all None items.

Hm can you share the modification? For me it looks like about 1/3 of them are in the past. So either a manual cleanup call or a migration will still be needed.
PS: Yes you are right.

[muharem]

With the small change to the script provided by @ggwpez, you can see there are only agendas for the future blocks with all None items.

Hm can you share the modification? For me it looks like about 1/3 of them are in the past. So either a manual cleanup call or a migration will still be needed. PS: I will add one.

Only extended the condition with if block < block_number and agenda.value.count(None) == len(agenda.value):

[ggwpez]

Only extended the condition with if block < block_number and agenda.value.count(None) == len(agenda.value):

But the current block is last_block = substrate.get_block_header()['header']['number'].
The block_number is the the snapshot position at which it requests the state, not the agendas block.

[gavofyork]

Looks good, but CI needs to pass still.

[muharem]

Bot merge

[Polkadot-Forum]

This pull request has been mentioned on Polkadot Forum. There might be relevant details there:

https://forum.polkadot.network/t/polkadot-release-analysis-v0-9-37/1736/1