Lint for superfluous let/match

[glaebhoerl]

Following the conversation in #10488, now that let _ = $foo; will be equivalent to just $foo;, we should have a lint to protect against people writing it, because anyone who does is likely doing so based on mistaken assumptions (i.e. that it does anything at all).

Going further, we should flag any let or match whose meaning is equivalent to just its inner/RHS expression.

My thinking w.r.t. the precise criteria is that we should flag a lint [what's the right verbage here?] unless either of the following are true:

• it's a match with more than one arm, in which case it's presumably doing a branch (if not, there are other lints for that)
• the pattern recursively contains a PatIdent (one which is actually an identifier and not a nullary enum) at any point, in which case the following code presumably uses the binding (if not, again there's another lint for that)

My question is whether this list should also include the presence of PatBox, PatUniq, and/or PatRegion. With the new semantics for _, will e.g. let ~_ = foo; "do anything", such as move out of foo?

As far as I can tell the change in _ semantics hasn't happened yet, but I'm thinking the lint could go in either before or after, with the potential difference that if it goes in before, the lint message should perhaps be in future tense.

[pnkfelix]

Yes, this should probably wait until #10488 is actually implemented.

[pnkfelix]

(and that is an interesting question about what the semantics of let ~_ = <expr>; should be. Not immediately obvious to me.)

[glaebhoerl]

Currently let ~ref x = foo; is allowed and does not move out of foo, so I suspect that let ~_ = ... would also be a no-op.

[glaebhoerl]

I just noticed this is in tension with #2974.

[steveklabnik]

Triage bump: nothing has changed.

[tbu-]

I believe this shouldn't be linted as described, let _ = expr; silences the unused result lint.

[steveklabnik]

Yup, this seems really bad. Since I'm the only one who's commented since 2013, I'm going to give this a close. A lot of the semantics here have changed since then.