Proof size for chains without PoV causes runtime panic

[nazar-pc]

Is there an existing issue?

• I have searched the existing issues

Experiencing problems? Have you tried our Stack Exchange first?

• This is not a support question.

Description of bug

#12383 in particular claims that chains without PoV should not care about proof size, but it doesn't seem to be the case.

The runtime I'm working on after Substrate upgrade was panicking with following:

Thread 'tokio-runtime-worker' panicked at 'Sensible defaults are tested to be valid; qed: ValidationErrors { has_errors: true, errors: ["[DispatchClass::Normal] Weight { ref_time: 1500000000000, proof_size: 0 } (total) has to be greater than Weight { ref_time: 5346284000, proof_size: 0 } (base block) & Weight { ref_time: 86298000, proof_size: 0 } (base extrinsic)", "[DispatchClass::Normal] Some(Weight { ref_time: 1299913702000, proof_size: 0 }) (max_extrinsic) must not be 0. Check base cost and average initialization cost.", "[DispatchClass::Normal] Weight { ref_time: 2000000000000, proof_size: 0 } (max block) must fit at least one extrinsic Weight { ref_time: 5432582000, proof_size: 0 } (base weight)", "[DispatchClass::Operational] Weight { ref_time: 2000000000000, proof_size: 0 } (total) has to be greater than Weight { ref_time: 5346284000, proof_size: 0 } (base block) & Weight { ref_time: 86298000, proof_size: 0 } (base extrinsic)", "[DispatchClass::Operational] Some(Weight { ref_time: 1799913702000, proof_size: 0 }) (max_extrinsic) must not be 0. Check base cost and average initialization cost.", "[DispatchClass::Operational] Weight { ref_time: 500000000000, proof_size: 0 } (reserved) has to be greater than Weight { ref_time: 86298000, proof_size: 0 } (base extrinsic) if set", "[DispatchClass::Operational] Weight { ref_time: 2000000000000, proof_size: 0 } (max block) must fit at least one extrinsic Weight { ref_time: 5432582000, proof_size: 0 } (base weight)", "[DispatchClass::Mandatory] Weight { ref_time: 2000000000000, proof_size: 0 } (max block) must fit at least one extrinsic Weight { ref_time: 5432582000, proof_size: 0 } (base weight)"] }', /home/nazar-pc/.cargo/git/checkouts/substrate-7bc20b373ca3e834/4fa79c1/frame/system/src/limits.rs:335

I have changed this:

BlockWeights::with_sensible_defaults(WEIGHT_PER_SECOND.saturating_mul(2), NORMAL_DISPATCH_RATIO)

To this:

BlockWeights::with_sensible_defaults(WEIGHT_PER_SECOND.saturating_mul(2).set_proof_size(1), NORMAL_DISPATCH_RATIO)

Now I'm getting another panic:

2022-11-09 03:43:28 [PrimaryChain] Proposing failed: Import failed: Error at calling runtime api: Execution failed: Runtime panicked: Sensible defaults are tested to be valid; qed: ValidationErrors { has_errors: true, errors: ["[DispatchClass::Operational] Weight { ref_time: 500000000000, proof_size: 0 } (reserved) has to be greater than Weight { ref_time: 86298000, proof_size: 0 } (base extrinsic) if set"] }    

There must be something wrong with Substrate if chains without PoV are required to be aware of such things.

Steps to reproduce

No response

[nazar-pc]

u64::MAX instead of 1 worked, despite I don't see how from the error. Weird...

[bkchr]

CC @KiChjang

[ggwpez]

There was a SE question about it https://substrate.stackexchange.com/questions/5557/construct-runtime-integrity-test-failing
since some of the invariants dont hold anymore with pov_size = 0.

PS: Just saw that you already found that 😆

[bkchr]

But he said that he used 1 and not 0?

[ggwpez]

But he said that he used 1 and not 0?

Looks like only for the NORMAL and not for the OPERATIONAL class.

[nazar-pc]

Looks like only for the NORMAL and not for the OPERATIONAL class.

Yeah, I think because NORMAL_DISPATCH_RATIO was 75% and 75% of 1 is 0 🤷‍♂️