cross-compilation doctests warning should not be behind verbose flag

[timfish]

^{Main tracking issue: #7040}

Problem

I just spent a lot longer than I'm willing to admit trying to work out why my doctests were not running. After working out I'd done everything correctly I finally tried --verbose and found the warning about cross-compilation not being supported.

cargo test completes successfully and I've not even known that the doctests have been skipped for months! Why is this warning hidden behind the verbose flag?

Steps

Create a src file with failing doc tests:
lib.rs

/// ```
/// panic("something')
/// ```
pub fn my_function() { }

Ensure you're building for another target. For me, this was on an M1 mac with the following .cargo/config to make this a cross platform build:

[build]
target = "x86_64-apple-darwin"

Run cargo test --doc

Results in no output after the build output.

Version

cargo 1.69.0

[weihanglo]

It was an intentional choice in #10132. Warning for every doctest target may be noisy. Could be improved by warning just once in normal mode if there is a cross-compile doctest, hinting people to use --verbose to see the full list of skipped doctest targets.

Relevant code is here:

cargo/src/cargo/ops/cargo_test.rs

Lines 190 to 205 in 0e93732

	CompileKind::Target(target) => {
	if target.short_name() != compilation.host {
	// Skip doctests, -Zdoctest-xcompile not enabled.
	config.shell().verbose(|shell| {
	shell.note(format!(
	"skipping doctests for {} ({}), \
	cross-compilation doctests are not yet supported\n\
	See https://doc.rust-lang.org/nightly/cargo/reference/unstable.html#doctest-xcompile \
	for more information.",
	unit.pkg,
	unit.target.description_named()
	))
	})?;
	continue;
	}
	}

[weihanglo]

Just a friendly reminder. It would be great for collaboration if people always strive to provide reproducible steps and version information for their issues. That may increase the chance for anyone wanting to fix gets started sooner.

[timfish]

provide reproducible steps and version information for their issues

Apologies. Just added these.

[weihanglo]

Labeled as P-low Priority: Low as we are considering to stabilize -Zdoctest-xcompile, so this warning may be unnecessary. There is no ETA though.

[1garo]

Would be okay if I work on this one? @weihanglo

[1garo]

I'm just having trouble building, I been having the message below in some crates:

error[E0658]: use of unstable library feature 'once_cell'

Any ideia?

[Rustin170506]

Try to use the latest stable version or nightly version rustc.

[1garo]

Try to use the latest stable version or nightly version rustc.

Thanks! Now I'm having some problem with linking error: linking with cc failed: exit status: 1 when cargo build, but cargo check is fine!

$ rustput -V
rustc 1.72.0-nightly

Mac M1 ventura 13.2

[weihanglo]

Labeled as P-low Priority: Low as we are considering to stabilize -Zdoctest-xcompile, so this warning may be unnecessary. There is no ETA though.

As it is labeled as P-low and S-needs-mentor, I'd suggest working on other issues if you're interested in contributing Cargo.

[progressive-galib]

Labeled as P-low Priority: Low as we are considering to stabilize -Zdoctest-xcompile, so this warning may be unnecessary. There is no ETA though.

has doctest cross complie stabilized ? if not i am going to work on this issue #12118 and warn when not verbose.

[progressive-galib]

nearly done, now i need some crate that needs doctest to check the warning is working.

[progressive-galib]

done

[progressive-galib]

merge please @weihanglo

[weihanglo]

@progressive-galib

Since @ehuss has stepped in #14920 and got some opinions, I would let them take over the review.

Per my early comment #12118 (comment), I still prefer we push forward the stabilization of #7040, though it is a cross-team effort and currently stuck with some incompatibility rust-lang/rust#64245 (comment).

[progressive-galib]

i think until we stabilize doctest cross compile we should merge it and remove it one that is stable.

[weihanglo]

I have no preference of merging now or not.

[progressive-galib]

@ehuss have you gotten any more second opinions and third opinions ?

[progressive-galib]

@weihanglo still no review ?

[weihanglo]

I've reviewed. See #12118 (comment).