Fix generated address returned by Substrate RPC runtime call

[castillax]

Description

When dry-running a contract deployment through the runtime API, the returned address does not match the actual address that will be used when the transaction is submitted. This inconsistency occurs because the address derivation logic doesn't properly account for the difference between transaction execution and dry-run execution contexts.

The issue stems from the create1 address derivation logic in exec.rs:

address::create1(
    &deployer,
    // the Nonce from the origin has been incremented pre-dispatch, so we
    // need to subtract 1 to get the nonce at the time of the call.
    if origin_is_caller {
        account_nonce.saturating_sub(1u32.into()).saturated_into()
    } else {
        account_nonce.saturated_into()
    },
)

The code correctly subtracts 1 from the account nonce during a transaction execution (because the nonce is incremented pre-dispatch), but doesn't account for execution context - whether it's a real transaction or a dry run through the RPC.

Review Notes

This PR adds a new condition to check for the IncrementOnce when calculating the nonce for address derivation:

address::create1(
    &deployer,
    // the Nonce from the origin has been incremented pre-dispatch, so we
    // need to subtract 1 to get the nonce at the time of the call.
    if origin_is_caller && matches!(exec_context, IncrementOnce::AlreadyIncremented) {
        account_nonce.saturating_sub(1u32.into()).saturated_into()
    } else {
        account_nonce.saturated_into()
    },
)

Before Fix

• Dry-run contract deployment returns address derived with nonce N
• Actual transaction deployment creates contract at address derived with nonce N-1
• Result: Inconsistent addresses between simulation and actual execution

After Fix

• Dry-run and actual transaction deployments both create contracts at the same address
• Result: Consistent contract addresses regardless of execution context
• Added test case to verify nonce handling in different execution contexts

This fix ensures that users can rely on the address returned by a dry run to match the actual address that will be used when the transaction is submitted.

Fixes paritytech/contract-issues#37

Checklist

• My PR includes a detailed description as outlined in the "Description" and its two subsections above.
• My PR follows the labeling requirements of this project (at minimum one label for T required)
  • External contributors: ask maintainers to put the right label on your PR.
• I have made corresponding changes to the documentation (if applicable)
• I have added tests that prove my fix is effective or that my feature works (if applicable)

[castillax]

/cmd prdoc --audience runtime_dev --bump patch

[pgherveou]

skimmed trough it quickly, will take another look in a bit

[athei]

/cmd fmt

[pgherveou]

can you cherry pick this eb1fe2b

I want to make sure that we are not breaking evm-test-suite here

[athei]

This is not good. Before the skip transfer was depending on whether we had a storage deposit limit. Now it is passed separately. We can have the situation where the limit is unchecked but the transfer is not skipped. I don't think we want that.

[athei]

This is dry_run. Same for the others.

[athei]

I don't think it makes sense to merge bump nonce logic with skip_transfer. Whether the transfer is skipped is already determined by the storage_deposit_limit argument. This also leads to the sitation that we need to add this arg to the call where it doesn't matter if we have to bump or not. We are also passing through this whole exec context even if we don't need the information later (as opposed to skip_transfer which we need for the whole stack duration).

I suggest just sticking to the original enum BumpNonce design. What do you think @pgherveou?

[pgherveou]

Yes you are right my bad, you can just pass the BumpNonce as an extra argument to run so we only carry it on the Stack
https://github.com/paritytech/polkadot-sdk/blob/master/substrate/frame/revive/src/exec.rs?plain=1#L1078-L1074

[castillax]

/cmd prdoc --audience runtime_dev --bump patch

[github-actions]

Command "prdoc --audience runtime_dev --bump patch" has failed ❌! See logs here

[athei]

I will have another look when the CI is green. The ones flagged as required.

[paritytech-workflow-stopper]

All GitHub workflows were cancelled due to failure one of the required jobs.
Failed workflow url: https://github.com/paritytech/polkadot-sdk/actions/runs/15137311658
Failed job name: test-linux-stable