ABI/"Switch statement" for Contracts about sway HOT 5 CLOSED

Ah, great, yeah I misinterpreted that as bits and thought they were word aligned. Perfect. So, then, let's decide that this can be our convention and move forward with it.

from sway.

Ideally it should simply return with RET and let the caller handle the error, however there's no downside at this time of having a revert at the end of the switch if no cases are taken. You can just make it revert with $zero for now. Eventually we can standardize some error codes.

from sway.

cc @adlerjohn for confirmation -- the "input" to the selector comes from the first four bytes of in (addr, size)s in the call frame, so it can be accessed with:

addi $rA $fp i7;
lw $rB $rA i0;

The above follows my aforementioned assumption that the selector is in the first four bytes of the input data, which, if my counting is correct, is word 7.

(citing here)

from sway.

It depends on our convention. But using the first parameter exclusively as the selector is probably the most reasonable approach.

You can fetch it as 4 + 4 + 1*64 + 1 + 1 + 1 = 75 words offset from $fp, not 7, if my math is right. (Each row in the call frame table in the specs isn't one word.)

from sway.

Should I put a revert opcode after the contract ABI switch statement? It seems correct to me, but I just want to double check.
And, if yes, what should the RVRT return?
cc @adlerjohn

from sway.