Support a string argument to "ignore" describing the reason

[ijc]

Since Rust 1.61.0 the argument to an ignore attribute has been printed as part of the test output. So:

#[test]
#[ignore = "Some reason"]
fn something() {...}

Will result in:

$ cargo test 
...
test something ... ignored, Some reason

I believe the syntax has been valid for some releases prior to 1.61.0 (I'm afraid I don't know since when) but was ignored.

It would be great if test_case's ignore modifier (and inconclusive I suppose) could similarly take an option reason and propagate it so that cargo test will show it.

I'm not sure that #[test_case(INPUT => ignore = "Some reason" OUTPUT)] will fit the current syntax but perhaps #[test_case(INPUT => ignore("Some reason") OUTPUT)] could work?

[luke-biel]

Hmm, yeah, it's definitelly something missing. I remember ignore comment syntax was present since 1.3x or earlier, so it shouldn't be a breaking change, we have MSRV 1.49 anyway. It just wasn't thought of.

I like your second proposal, if you have resources and time PR is welcome.

[ijc]

I assumed we want the reason to be optional, for backward compat so that existing cases do not all need updating and can still use a bare inconclusive. Given that then using () to surround the reason argument was ambiguous -- since => inconclusive () should mean a test with output () but is treated as one with an empty reason and no output specified. Similar problems exist with => inconclusive ("this is the result") where this is the result ended up as the reason not the expected result...

I went with [] instead and will open a PR in a moment, {} would be an easy change if preferred. Other things (like <>) would need more custom handling.

[frondeus]

What about => inconclusive = "this is reason" ("this is result")? @luke-biel would that be ambiguous or/and better?

[ijc]

I think that would be unambiguous? At least I don't think there are any output matchers which start with an = (nor an == FWIW` although I don't think that would be ambiguous even if there was).

I guess it's a question of taste/style and perhaps future direction, happy to defer to you both.

[luke-biel]

I think I'm leaning towards [] as proposed right now. I don't think that any of these is better, but
=> ignore = "reason" "result" may be a bit confusing.