Show the span of feature that is unstable when using a feature-list.

[hirschenberger]

It is hard to find the actual unstable feature which caused the error when using a list of stable and unstable features as the span marks the whole line

src/k8055.rs:22:1: 22:64 error: unstable feature
src/k8055.rs:22 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This PR spawns an error for each unstable feature in the list:

est.rs:1:12: 1:26 error: unstable feature [-D unstable-features]
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                     ^~~~~~~~~~~~~~
test.rs:1:28: 1:41 error: unstable feature [-D unstable-features]
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                                     ^~~~~~~~~~~~~
test.rs:1:43: 1:47 error: unstable feature [-D unstable-features]
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                                                    ^~~~
test.rs:1:49: 1:56 error: unstable feature [-D unstable-features]
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                                                          ^~~~~~~
test.rs:1:58: 1:62 error: unstable feature [-D unstable-features]
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                                                                   ^~~~

[rust-highfive]

r? @nrc

(rust_highfive has picked a reviewer for you, use r? to override)

[nrc]

The code looks fine, but I wonder if this is the best solution - it seems like it could increase the number of errors by an annoying amount. How would you feel about listing the features which are unstable instead? E.g,

src/k8055.rs:22:1: 22:64 error: unstable feature (slice_patterns, core)
src/k8055.rs:22 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Assumming that only slice_patterns and core are unstable for this example.

Or do you have any other ideas for giving better errors without giving more of them?

[hirschenberger]

Yes, first I wanted to implement it the way you suggested. But then I realized that warnings of unused features where realized this way. So I did the same for unstable features.

test.rs:1:49: 1:56 warning: unused or unknown feature, #[warn(unused_features)] on by default
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                                                          ^~~~~~~
test.rs:1:58: 1:62 warning: unused or unknown feature, #[warn(unused_features)] on by default
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                                                                   ^~~~
test.rs:1:28: 1:41 warning: unused or unknown feature, #[warn(unused_features)] on by default
test.rs:1 #![feature(slice_patterns, rustc_private, core, convert, libc)]
                                     ^~~~~~~~~~~~~

[nrc]

OK makes sense.

@bors: r+ rollup

[bors]

📌 Commit 70db73a has been approved by nrc

[nrc]

cc @alexcrichton Wondering if we should make unused and unstable features a single error when there are multiple features?

[alexcrichton]

This is probably fine for now, most crates don't have lots of features that aren't already on nightly.

[bors]

⌛ Testing commit 70db73a with merge d7185dc...

[bors]

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-win-gnu-32-nopt-t, auto-win-gnu-32-opt, auto-win-gnu-64-nopt-t, auto-win-gnu-64-opt