Independent lifetimes still considered overly constrained

[philip-peterson]

I saw a confusing error message today which seems related to #60216, but not sure if it's the same.

error[E0495]: cannot infer an appropriate lifetime for lifetime parameter `'a` due to conflicting requirements
  --> cli/src/main.rs:27:14
   |
27 |             .arg(
   |              ^^^
   |
note: first, the lifetime cannot outlive the lifetime `'a` as defined on the function body at 24:20...
  --> cli/src/main.rs:24:20
   |
24 | fn with_build_args<'a, 'b, 'c, 'd>(app: Box<App<'a, 'b>>) -> Box<App<'c, 'd>> {
   |                    ^^
note: ...so that the types are compatible
  --> cli/src/main.rs:27:14
   |
27 |             .arg(
   |              ^^^
   = note: expected `clap::App<'_, '_>`
              found `clap::App<'a, 'b>`
note: but, the lifetime must be valid for the lifetime `'c` as defined on the function body at 24:28...
  --> cli/src/main.rs:24:28
   |
24 | fn with_build_args<'a, 'b, 'c, 'd>(app: Box<App<'a, 'b>>) -> Box<App<'c, 'd>> {
   |                            ^^
note: ...so that the expression is assignable
  --> cli/src/main.rs:25:5
   |
25 | /     Box::new(
26 | |         (*app)
27 | |             .arg(
28 | |                 Arg::with_name("release")
...  |
31 | |             )
32 | |     )
   | |_____^
   = note: expected `std::boxed::Box<clap::App<'c, 'd>>`
              found `std::boxed::Box<clap::App<'_, '_>>`

error: aborting due to previous error; 7 warnings emitted

For more information about this error, try `rustc --explain E0495`.

The function in question takes four independent lifetimes which (in my mind) should be free to grow as large as the compiler would like.

fn with_build_args<'a, 'b, 'c, 'd>(app: Box<App<'a, 'b>>) -> Box<App<'c, 'd>> {
    Box::new(
        (*app)
            .arg(
                Arg::with_name("release")
                    .help("Whether to invoke `cargo build` using the --release flag")
                    .long("release")
            )
    )
}

[Lonami]

The problem isn't about growing the lifetimes but the dependencies between them, as the compiler doesn't know the relationship between all of them four, and I don't think the compiler could infer it in this case.

[kadiwa4]

The diagnostic has got shorter and I think it's good now (nightly 2023-01-15, clap 2.34.0):

error: lifetime may not live long enough
  --> lib.rs:4:5
   |
3  |   fn with_build_args<'a, 'b, 'c, 'd>(app: Box<App<'a, 'b>>) -> Box<App<'c, 'd>> {
   |                      --      -- lifetime `'c` defined here
   |                      |
   |                      lifetime `'a` defined here
4  | /     Box::new(
5  | |         (*app)
6  | |             .arg(
7  | |                 Arg::with_name("release")
...  |
10 | |             )
11 | |     )
   | |_____^ function was supposed to return data with lifetime `'c` but it is returning data with lifetime `'a`
   |
   = help: consider adding the following bound: `'a: 'c`

error: lifetime may not live long enough
  --> lib.rs:4:5
   |
3  |   fn with_build_args<'a, 'b, 'c, 'd>(app: Box<App<'a, 'b>>) -> Box<App<'c, 'd>> {
   |                          --      -- lifetime `'d` defined here
   |                          |
   |                          lifetime `'b` defined here
4  | /     Box::new(
5  | |         (*app)
6  | |             .arg(
7  | |                 Arg::with_name("release")
...  |
10 | |             )
11 | |     )
   | |_____^ function was supposed to return data with lifetime `'d` but it is returning data with lifetime `'b`
   |
   = help: consider adding the following bound: `'b: 'd`

help: the following changes may resolve your lifetime errors
  |
  = help: add bound `'a: 'c`
  = help: add bound `'b: 'd`

but perhaps it should tell you to replace 'c and 'd with 'a and 'b instead

[estebank]

@philip-peterson feel free to reopen if there's more we could be doing here, but it does seem like the current output would help.