fuzz_targets/roundtrip_policy.rs fails with 6f7228706b282929 (~ "or(pk())")

[kiminuo]

Steps to reproduce:

1. Put 6f7228706b282929 to fuzz_targets/roundtrip_policy.rs :: duplicate_crash()
2. Run test

Output:

thread 'tests::duplicate_crash_3' panicked at 'assertion failed: `(left == right)`
  left: `"or(pk())"`,
 right: `"or(1@pk())"`', fuzz_targets/roundtrip_policy.rs:17:9

[apoelstra]

Hmm, gross. I don't really want to ban either of these forms.

[sanket1729]

This behaviour is expected, so I think we should either remove the test or modify the test so that it explicitly adds 1@ whenever missing.

[apoelstra]

We definitely want to keep this fuzztest. It's found a lot of bugs.

I think adding/removing 1@ (not sure which one is necessary) is the way to do.

[kiminuo]

@sanket1729 Is it a problem that both policies:

"6f72283140706b282929" ~ "or(1@pk())"
"6f7228706b282929" ~ "or(pk())"

are permitted at the moment? From maliability point of view

[sanket1729]

We do not aim for non-malleability in policies. We aim for non-malleability for Miniscript. Policies are not directly used in bitcoin transactions, but Miniscript it.

On another note, I think or() and and() with a single argument should be banned. As such both of those should be invalid policies.