Fix archive sync panic

[orriin]

• Fixes the archive sync getting stuck at block 2585477 due to panic by adding a codeSubstitute that patches the buggy code with one that doesn't panic.
• Updates node/src/chain_spec/finney.rs so Finney chainspec always builds with the codeSubstitues field.

[ales-otf]

It would be better to have the substitute in the ChainSpec builder or in the function producing the chain spec instead. It's more reliable and doesn't require patching. If it's possible.

[ales-otf]

It would be better to have the substitute in the ChainSpec builder or in the function producing the chain spec instead. It's more reliable and doesn't require patching. If it's possible.

Probably this could help: https://paritytech.github.io/polkadot-sdk/master/sc_chain_spec/fn.set_code_substitute_in_json_chain_spec.html

[orriin]

It would be better to have the substitute in the ChainSpec builder or in the function producing the chain spec instead. It's more reliable and doesn't require patching. If it's possible.

@ales-otf I tried this but couldn't figure it out. It seems the ChainSpec Builder doesn't support setting code_substitutes, also the fn you linked seems to operate on a JSON object, whereas we have a ChainSpec. Let me know if you have any ideas tho! Would be great to get it working that way.

[ales-otf]

It would be better to have the substitute in the ChainSpec builder or in the function producing the chain spec instead. It's more reliable and doesn't require patching. If it's possible.

@ales-otf I tried this but couldn't figure it out. It seems the ChainSpec Builder doesn't support setting code_substitutes, also the fn you linked seems to operate on a JSON object, whereas we have a ChainSpec. Let me know if you have any ideas tho! Would be great to get it working that way.

I believe you can modify this JSON: https://github.com/opentensor/subtensor/blob/devnet-ready/node/src/chain_spec/finney.rs#L198

Oh, it's just genesis. Then yeah. Looks like you have to patch the json the way you do. As an option, you can get the json from the ChainSpec (https://paritytech.github.io/polkadot-sdk/master/sc_chain_spec/struct.GenericChainSpec.html#method.as_json) modify it using the function and build the ChainSpec from the modified JSON. It's ugly, but it's more safe doing it in rust, than in a shell script.

BTW, it looks like we can omit using the script and just hardcode everything, as there is with_code method.

[orriin]

@ales-otf that is the genesis JSON not the chain spec JSON

[ales-otf]

@ales-otf that is the genesis JSON not the chain spec JSON

yeah, I've updated my comment

[orriin]

As an option, you can get the json from the ChainSpec (https://paritytech.github.io/polkadot-sdk/master/sc_chain_spec/struct.GenericChainSpec.html#method.as_json) modify it using the function and build the ChainSpec from the modified JSON. It's ugly, but it's more safe doing it in rust, than in a shell script.

Good spot! I will have a go at this.

BTW, it looks like we can omit using the script and just hardcode everything, as there is with_code method.

I don't think with_code is what we want, I believe it sets the genesis code rather than a code substitute for a specific block number

[ales-otf]

I don't think with_code is what we want, I believe it sets the genesis code rather than a code substitute for a specific block number

Yeah, that's exactly what I meant. We now need to use this script to keep the genesis mostly because of the genesis code. The rest comes frome that generate_genesis function. So if we put the code value into ChainSpec builder and if we also add substitute there, we can not only omit the shell script, but we can even connect to the running mainnet using cargo run -- --chain finney or something like this - without requiring to have the chainspec file at all. Still, it would be good to track changes in the chainspecs. Some warning system at least. Also, it would be better to extract the code value into a separate file and include it into the source from that file. It would be easier to edit the file and the git diffs would be smaller.

Opened an issue for this.

[ales-otf]

I don't think with_code is what we want, I believe it sets the genesis code rather than a code substitute for a specific block number

Yeah, that's exactly what I meant. We now need to use this script to keep the genesis mostly because of the genesis code. The rest comes frome that generate_genesis function. So if we put the code value into ChainSpec builder and if we also add substitute there, we can not only omit the shell script, but we can even connect to the running mainnet using cargo run -- --chain finney or something like this - without requiring to have the chainspec file at all. Still, it would be good to track changes in the chainspecs. Some warning system at least. Also, it would be better to extract the code value into a separate file and include it into the source from that file. It would be easier to edit the file and the git diffs would be smaller.

This is definitely out of scope of this PR. I've just spotted the method and this idea came to mind.

[orriin]

A bit hacky but figured out a solution for setting codeSubstitutes at the node/src/chain_spec/finney.rs level, thanks @ales-otf !

[ales-otf]

Please, look the comment - this change is important.

And just an additional not much important styling thing, which would improve readability and debugging — to encapsulate this whole part related to setting the substitute into a separate function.

[ales-otf]

Thanks!