Dealing with errors

[secona]

cargo-plumbing/tests/testsuite/locate_manifest.rs

Lines 106 to 118 in 1c9fa58

	#[cargo_test]
	fn errors_when_no_manifest() {
	let p = project().no_manifest().build();
	p.cargo_plumbing("plumbing locate-manifest")
	.with_stderr_data(str![[r#"
	[ERROR] could not find `Cargo.toml` in `[ROOT]/foo` or any parent directory
	"#]])
	.with_stdout_data("")
	.with_status(101)
	.run();
	}

Currently, we have one test for errors and that error is in a form of plain strings outputted to stderr. In the future, should we keep doing this or should we opt in for an all jsonlines option? Perhaps something like this.

{
  "reason": "locate-manifest-error",
  "msg": "could not find `Cargo.toml` in `[ROOT]/foo` or any parent directory"
}

I don't think outputting to stderr is a problem, its just that we currently don't have a way for users to understand errors programmatically.

I remember we had a conversation about handling errors in cargo-plumbing-schemas, though I don't remember where exactly.

[epage]

I was thinking we could hold off on this until we better understand how we want to deal with programmatic cargo errors more generally.

Granted, this could be an opportunity to experiment with it, so I guess I can go either way.

its just that we currently don't have a way for users to understand errors programmatically.

One of the big questions would be how much programmatic information would people want on errors and how do we not break people in providing it as errors are fairly fluid.