Fixes manual_flatten removes the useless if let

[donkomura]

Closes #14692

The suggestion of manual_flatten does not includes the replacement of if let so far despite of .flatten() suggestion.
This PR eliminates a redundant if let.

changelog: [manual_flatten] the suggestion removes if let

[donkomura]

The suggestion always have multiple lines, so we can become the help message simplify.
(or should we suggest the one-liner and separate the suggestion that includes removing if let?)

[rustbot]

r? @samueltardieu

rustbot has assigned @samueltardieu.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[samueltardieu]

If this is supposed to now be auto-fixable, then //@no-rustfix should be removed from the test file.

However, the current suggestion don't appear to be valid unless the binding from the Some() or Ok() is a single identifier with the same name as the identifier from the for loop.

For example, given this code:

    for x in [Some(3)] {
        if let Some(y)  = x {
            println!("{y}");
        }
    }

the suggestion will not compile:

    for x in [Some(3)].into_iter().flatten() {
        println!("{y}");
    }

[donkomura]

I'd fixed it to out valid code.

[samueltardieu]

This is getting good.

//@no-rustfix is still in place and should be removed, so that autofixes can be tested as well.

Also, the following code makes the lint fail to give a good error message and fix the problem:

    macro_rules! inner {
        ($id:ident / $new:pat => $action:expr) => { if let Some($new) = $id { $action; } }
    }

     for ab in [Some((1, 2)), Some((3, 4))] {
         inner!(ab / (c, d) => println!("{c}-{d}"));
    }

because it gets confused by the macro. The lint should probably not trigger if the for loop, the if let, or its test, don't all come from the same context.

[samueltardieu]

Shouldn't you initialize applicability with MachineApplicable if the suggestion is complete?

[donkomura]

Thank you for pointing that out. I made corrections to prevent the macro from triggering a lint. I also adjusted the Applicability.

[rustbot]

☔ The latest upstream changes (possibly 3927a61) made this pull request unmergeable. Please resolve the merge conflicts.