Ah, good catch, thank you. When I spelled the functions out into their own functions instead of being generated in #2189, I defined the endian and sign arguments explicitly with IDE support in mind and didn't notice this change in expected behavior. The arguments weren't defined before and were thus picked up solely by the @LocalNoArchContext decorator when passing them as kwarg.

It's annoying that we have to pass those arguments through like this and I like the verbatism of having to use keyword args instead of allowing positional args, since you'd have to remember the order this way. To avoid confusing p64 calls with magic values like p64(x, 'big', True) removing the extra "unused" arguments from all the packing and unpacking functions would be the way to go imo.

This seems to be personal preference but pwntools scripts are designed to make exploits readable and using keyword arguments makes the function usage more explicit. What do you think?

BTW with py2 support deprecated we can finally use def f(a1, *, a2, a3): syntax for keyword-only arguments.

To avoid confusing p64 calls with magic values like p64(x, 'big', True) removing the extra "unused" arguments from all the packing and unpacking functions would be the way to go imo.

I've never written exploit code using IDE, only by raw editors like vim or sublime text. So the extra typing
endianness slows me down a little bit. Also, 'big' or 'little' hints me about endianness anyway so I don't feel it
much more clear to explicitly use keyword arguments. However, I agree that 'sign` should be explicitly spelled out
as I seldomly use them.

It's your decision, I'm happy with whatever.

Edit: I went ahead and removed sign from normal arguments.