[tamird]

Fixes #41701.

r? @arielb1

[tamird]

Note sure what to make of these test failures:

Testing libtest stage1 (x86_64-unknown-linux-gnu -> x86_64-unknown-linux-gnu)
   Compiling term v0.0.0 (file:///checkout/src/libterm)
   Compiling test v0.0.0 (file:///checkout/src/libtest)
error[E0308]: mismatched types
   --> /checkout/src/libtest/stats.rs:887:5
    |
887 | /     pub fn sum_three_items(b: &mut Bencher) {
888 | |         b.iter(|| {
889 | |             [1e20f64, 1.5f64, -1e20f64].sum();
890 | |         })
891 | |     }
    | |_____^ expected struct `__test::test::Bencher`, found struct `Bencher`
    |
    = note: expected type `fn(&mut __test::test::Bencher)`
               found type `fn(&mut Bencher) {stats::bench::sum_three_items}`

error[E0308]: mismatched types
   --> /checkout/src/libtest/stats.rs:893:5
    |
893 | /     pub fn sum_many_f64(b: &mut Bencher) {
894 | |         let nums = [-1e30f64, 1e60, 1e30, 1.0, -1e60];
895 | |         let v = (0..500).map(|i| nums[i % 5]).collect::<Vec<_>>();
896 | |
...   |
899 | |         })
900 | |     }
    | |_____^ expected struct `__test::test::Bencher`, found struct `Bencher`
    |
    = note: expected type `fn(&mut __test::test::Bencher)`
               found type `fn(&mut Bencher) {stats::bench::sum_many_f64}`

error[E0308]: mismatched types
   --> /checkout/src/libtest/stats.rs:903:5
    |
903 |     pub fn no_iter(_: &mut Bencher) {}
    |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected struct `__test::test::Bencher`, found struct `Bencher`
    |
    = note: expected type `fn(&mut __test::test::Bencher)`
               found type `fn(&mut Bencher) {stats::bench::no_iter}`

error: aborting due to 3 previous errors

error: Could not compile `test`.
warning: build failed, waiting for other jobs to finish...
error: build failed


command did not execute successfully: "/checkout/obj/build/x86_64-unknown-linux-gnu/stage0/bin/cargo" "test" "-j" "4" "--target" "x86_64-unknown-linux-gnu" "--release" "--locked" "--color" "always" "--manifest-path" "/checkout/src/libtest/Cargo.toml" "-p" "test:0.0.0" "-p" "term:0.0.0" "--" "--quiet"
expected success, got: exit code: 101


failed to run: /checkout/obj/build/bootstrap/debug/bootstrap test
Build completed unsuccessfully in 0:23:20
Makefile:54: recipe for target 'check' failed
make: *** [check] Error 1

[arielb1]

The problem is this function, which redirects tests of crates with a libtest attribute to their own test infrastructure, instead of the builtin libtest.

rust/src/libsyntax/test.rs

Lines 611 to 617 in 6f4ab94

	fn is_test_crate(krate: &ast::Crate) -> bool {
	match attr::find_crate_name(&krate.attrs) {
	Some(s) if "test" == s.as_str() => true,
	_ => false
	}
	}

[kennytm]

This line is too long.

[00:03:11] tidy error: /checkout/src/libsyntax/test.rs:88: line longer than 100 chars

[tamird]

Thank you, fixed.

[tamird]

Thanks @arielb1! This is green.

[arielb1]

These sort of changes are more of @alexcrichton's domain.

[arielb1]

r? @alexcrichton

[alexcrichton]

@bors: r+

[bors]

📌 Commit 07fc66c has been approved by alexcrichton

[bors]

⌛ Testing commit 07fc66c2583ded22f061ef7eb643e62edef7a707 with merge 86ca4b4773693d5f666a96fffefe1f0c95d226ff...

[bors]

💔 Test failed - status-travis

[kennytm]

x86_64-gnu-aux could not build stage2-cargo, the crate test not found.

[01:12:33] error[E0432]: unresolved import `test`
[01:12:33] 
[01:12:34] error: aborting due to previous error
[01:12:34] 
[01:12:34] error: Could not compile `cargo`.
[01:12:34] warning: build failed, waiting for other jobs to finish...
[01:12:58] error: build failed

[tamird]

I'm guessing that now that libtest doesn't have the name attribute, something's different in the resulting rlib (or dylib, on my mac). @alexcrichton any guidance? FWIW, it does seem like faa6769 is what broke this (and yet it is necessary to solve the earlier build problem).

[alexcrichton]

Looks like the bug here is that tests/test.rs is being compiled in Cargo which is failing to compile with the new handling. You may wish to just leave the test harness modification unchanged and leave the #![crate_name = "test"] annotation there. Definitely a hack, but should keep things working hopefully

[tamird]

@alexcrichton I don't follow: this failure occurs even without the mass removal of attributes (i.e. in the second commit).

[alexcrichton]

I'd recommend just backing out the modifications to libsyntax/test.rs. When doing that I'd add back #![crate_name] to src/libtest/lib.rs and that I believe should pass tests.

[tamird]

It'd be nice to understand the cause, though. I'll take another look at this later today.

…

On Aug 24, 2017 18:36, "Alex Crichton" ***@***.***> wrote: I'd recommend just backing out the modifications to libsyntax/test.rs. When doing that I'd add back #![crate_name] to src/libtest/lib.rs and that I believe should pass tests. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#43994 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABdsPNG6VmEkOVaw-r2NIkFeIdC9WmHuks5sbfr7gaJpZM4O8eXF> .

[alexcrichton]

The test harness acts differently based on its calculated value of is_test_crate. Today is_test_crate is true for libtest and is_test_crate is false for Cargo's unit test. With the logic you've done in this PR they both calculate to true. Presumably Cargo's unit test doens't work with is_test_crate being true.

[tamird]

@alexcrichton OK, done.

I don't quite understand this machinery that does specific-to-libtest stuff, but I've go ahead and documented some of this sorcery.

[alexcrichton]

@bors: r+

[bors]

📌 Commit b3f50ca has been approved by alexcrichton

[bors]

⌛ Testing commit b3f50ca with merge 83fcd4d...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: alexcrichton
Pushing 83fcd4d to master...