more address truncation fixes #399
Conversation
Signed-off-by: Ignacio Hagopian <[email protected]>
| address := slot.Bytes20() | ||
| if evm.chainRules.IsPrague { | ||
| usedGas += evm.TxContext.Accesses.TouchAddressOnReadAndComputeGas(slot.Bytes(), uint256.Int{}, trieUtils.CodeSizeLeafKey) | ||
| usedGas += evm.TxContext.Accesses.TouchAddressOnReadAndComputeGas(address[:], uint256.Int{}, trieUtils.CodeSizeLeafKey) |
There was a problem hiding this comment.
This was the offending line that caused the new panic.
But I was wondering if L105-107 are really needed. This method is the one that calculates the dynamic gas. This touch on CodeSizeLeafKey also happens in opExtCodeSize (which I also fixed, since if not it would panic there too).
The dynamic gas method is called before the actual opcode execution, so that's why it panicked here. But I think we can avoid the touching here, and just rely on the opcode execution method (i.e: opExtCodeSize)?
There was a problem hiding this comment.
yes, and in fact this is fixed in my access witness rebase but since it's going to take a bit of time, better merge this asap.
| address := slot.Bytes20() | ||
| cs := uint64(interpreter.evm.StateDB.GetCodeSize(address)) | ||
| if interpreter.evm.chainRules.IsPrague { | ||
| statelessGas := interpreter.evm.Accesses.TouchAddressOnReadAndComputeGas(slot.Bytes(), uint256.Int{}, utils.CodeSizeLeafKey) | ||
| statelessGas := interpreter.evm.Accesses.TouchAddressOnReadAndComputeGas(address[:], uint256.Int{}, utils.CodeSizeLeafKey) |
There was a problem hiding this comment.
The replay panicked just before here, but this also requires a fix.
There was a problem hiding this comment.
ah ok so you're not actually removing it. nvm, it'll be removed soon enough.
| statelessGas := interpreter.evm.Accesses.TouchAddressOnReadAndComputeGas(slot.Bytes(), uint256.Int{}, utils.CodeKeccakLeafKey) | ||
| statelessGas := interpreter.evm.Accesses.TouchAddressOnReadAndComputeGas(address[:], uint256.Int{}, utils.CodeKeccakLeafKey) |
There was a problem hiding this comment.
This is other case I found by code exploring.
Signed-off-by: Ignacio Hagopian <[email protected]>
Signed-off-by: Ignacio Hagopian <[email protected]>
Signed-off-by: Ignacio Hagopian <[email protected]>
The replay found another case similar to #397. I scanned for more similar cases and found one extra.
Leaving some comments.