memory_opt calcs wrong memory word range

[lightsing]

current we calc the word by this approach:

let start_slot = offset - offset % 32;
let end = offset + length;
let end_slot = end - end % 32;

the range is bytes[start_slot..end_slot+32].

consider we are coping memory of first 32 bytes, aka offset = 0, length = 32.
the words bytes will be bytes[0..63], there comes an extra word.
expect words bytes to be bytes[0..31], only one word.

This should be fixed, with a low priority.

@DreamWuGit @lispc

[DreamWuGit]

it is known edge case like in https:/scroll-tech/zkevm-circuits/blob/memory_opt/bus-mapping/src/evm/opcodes/mload.rs#L62, very low priority for now.

[naure]

Can you point to the code to review specifically?

[naure]

This should be fixed here with the function align_range. See also the unittest below. In particular: "When the range ends exactly at a word boundary, it is still ONE WORD."

Please let me know whether we agree on the behavior. (@lightsing @DreamWuGit )

[lightsing]

This should be fixed here with the function align_range. See also the unittest below. In particular: "When the range ends exactly at a word boundary, it is still ONE WORD."

Please let me know whether we agree on the behavior. (@lightsing @DreamWuGit )

I aggree with your fix.