Rebindable syntax: unintentional?

[brianhuffman]

It turns out that you can change the meaning of _ # _ patterns by defining a function called splitAt in a where clause.

Cryptol> (\(x # y) -> (y, x)) (True, False) where splitAt a = a
[warning] at <interactive>:1:42--1:49
    This binding for splitAt shadows the existing binding from
    (at /var/folders/hf/g8kjlj3x4cz05819qsk684_c0000gp/T/Cryptol98291-0.cry:347:11--347:18, splitAt)
(False, True)

I don't think this is intentional. We should probably make it so that# patterns always mean the same thing.

This would probably be easier to fix if we do the NoPat pass after renaming, instead of before. (See #567.)

[yav]

Agreed :)

[brianhuffman]

It looks like the unary operators - and ~ are also rebindable:

Cryptol> - 0x4 where negate x = x
[warning] at <interactive>:1:13--1:19
    This binding for negate shadows the existing binding from
    (at /var/folders/hf/g8kjlj3x4cz05819qsk684_c0000gp/T/Cryptol99252-0.cry:103:11--103:17, negate)
0x4
Cryptol> ~ 0x4 where complement x = x
[warning] at <interactive>:1:13--1:23
    This binding for complement shadows the existing binding from
    (at /var/folders/hf/g8kjlj3x4cz05819qsk684_c0000gp/T/Cryptol99252-0.cry:109:11--109:21, complement)
0x4

We should probably fix these too.