feat: constructor support #31
Conversation
thanks Leo, will push something before eod |
| SECTIONS | ||
| { | ||
| /DISCARD/ : { | ||
| *(.eh_frame) | ||
| *(.eh_frame_hdr) | ||
| *(.eh_frame.*) | ||
| } | ||
| } | ||
|
|
There was a problem hiding this comment.
@leonardoalt apparently newer rustc versions create an llvm section for exception handling (.eh_frame) which they didn't use to.
this section caused me a lot of compilation issues because it was being allocated in a "forbidden" memory region by our interpreter.
it took me a while, but thanks to some AI help i ended up with this modification of the linker script, which maintains the same elf codegen as before the version bump (it simply discard the .eh_frame).
| "-C", "link-arg=-T../r5-rust-rt.x", | ||
| "-C", "inline-threshold=275" | ||
| "-C", "link-arg=-T../../r5-rust-rt.x", | ||
| "-C", "llvm-args=--inline-threshold=275" |
There was a problem hiding this comment.
fyi, i also updated the rustflags as i was getting deprecation warnings
|
and last but not least, i also moved all the contracts under an |
implements contract constructors with with the following design choices:
fn new(..constructor_args) -> Selfnomenclature. Note that returningSelfis irrelevant (therefore not enforced) and was simply used in the ERC20 example to make it feel more familiar for readers.fn deploy_contract(db: &mut InMemoryDB, bytecode: Bytes)to also takeencoded_args: Option<Vec<u8>>as an argument.Optioncause it feels more explicit, but we could also useVec<u8>and check whether it is empty or notdeployfeature flag, no function selector is needed, hence why i decided to only use the encoded args (to minimize initcode size)