Replacing a parameter declared as `mut a` with `let mut a = a;` causes program to no longer compile

[bstrie]

This may very well be a dupe of #31439, but it manifests somewhat differently and may shed light on the potential fix. Probably still worth having a test case for anyway.

Take this program, which compiles:

fn test_drain(mut a: Vec<u32>) -> Vec<u32> {
    a.drain(..).collect::<Vec<u32>>()
}

fn main() {}

Now, instead of declaring the parameter as mut a, declare it as immutable and rebind it as mutable in the body of the function, which is supposed to be semantically equivalent. This results in a compiler error:

fn test_drain(a: Vec<u32>) -> Vec<u32> {
    let mut a = a;
    a.drain(..).collect::<Vec<u32>>()
}

fn main() {}

<anon>:3:5: 3:6 error: `a` does not live long enough
<anon>:3     a.drain(..).collect::<Vec<u32>>()
             ^
<anon>:1:40: 4:2 note: reference must be valid for the destruction scope surrounding block at 1:39...
<anon>:1 fn test_drain(a: Vec<u32>) -> Vec<u32> {
<anon>:2     let mut a = a;
<anon>:3     a.drain(..).collect::<Vec<u32>>()
<anon>:4 }
<anon>:2:19: 4:2 note: ...but borrowed value is only valid for the block suffix following statement 0 at 2:18
<anon>:2     let mut a = a;
<anon>:3     a.drain(..).collect::<Vec<u32>>()
<anon>:4 }
error: aborting due to previous error
playpen: application terminated with error code 101

As in #31439, replacing the implicit return with an explicit return also allows the program to compile.

[arielb1]

This is the destruction order problem: the destructor of Vec runs before the destructor of drain, so if collect panics you will have trouble.

[Thiez]

@arielb1 I was under the impression that destructors ran in reverse order of declaration, so surely the destructor of drain would in fact run before the destructor of Vec?

[pnkfelix]

I would need to go review the temp r-value lifetime rules in a case like this, in terms of how the temporaries of the final expression in a block relate to the local variables in that block

(I know it's unintuitive for explicit return to work in cases like this that are supposedly equivalent ...)

[Thiez]

@pnkfelix If/when you review those rules, could you perhaps check whether the root cause is the same as the example here?

[Mark-Simulacrum]

Closing in favor of #21114. Please reopen if that's incorrect, but from what I can tell this is a duplicate of that issue.