Anvil + Cast does not show failed tx

[Pongch]

I have deployed a contract through low level calls via cast send, the transaction gets mined successfully on Anvil with 'Deployed Contract:

' log and 'cast tx' shows the transaction receipt properly

However, the contract does not exist, due to the transaction gas estimate being underpriced

How to replicate

• make a transaction that requires N amount of gas
• underprice transaction with gasEstimate at N-1

expected behavior

• show transaction failure and reason on Anvil log or via Cast tx

[mattsse]

would appreciate it if you could provide a minimal example.

does cast tx list the transaction as successful or failed?

anvil does log failed transactions and the reason string if available: (require(.., reason)

[mds1]

This issue is old, but we actually just reproduced this today. Unfortunately I can't provide our exact case to reproduce, but it was very similar to what @Pongch said. We were simulating a Gnosis Safe transaction and provided all the signatures in calldata (which is why I can't share a repro). Details:

First we used anvil

• It estimated a gas limit of ~117k, but this was insufficient because ~138k gas is is needed
• The transaction receipt showed success, but the full transaction did not actually execute (the expected storage variable change did not occur)
• We looked at the trace with cast run. This indicated a top-level revert because of a wrong ecrecover (the recovered address was not a signer on the safe). Only one ecrecover was shown, i.e. it failed on the first signature, and it recovered an unexpected address (the address was not a signer, but all sigs were valid)
• When we lowered the gas limit to ~110k, the results were as expected: failed status, longer trace showing more ecrecovers occuring.
• When specifying a high enough gas limit, everything works as expected

Then we tried hardhat (via jackhammer):

• It chose a better default gas limit so worked out of the box
• We hardcoded the ~117k gas limit anvil used, and saw multiple ecrecovers for all signers before the out of gas (as expected)
• We hardcoded the same ~110k test gas limit, and similarly saw multiple ecrecovers for all signers before the out of gas

To summarize, there were three issues with anvil:

• Bad gas estimate
• Wrong success value reported
• Wrong trace reported with the 117k gas case