Infer targets from subdirectories

[rwakulszowa]

Fixes #4086

I still have a few questions:

• should I add some tests for the old behaviour? It isn't really tested at the moment (no tests failed when I broke the implementation); I could refactor the tests to check for both single file and subdirectory inference
• I moved things around, mostly reusing the code from inferred_bins - hopefully I didn't break anything, but it won't hurt to double check :)
• Do we have something like servo's tidy check for coding style? I'm open for suggestions if something isn't formatted correctly
• Just a general one - should I rebase + squash commits every time I make subsequent changes to cargo?

[rust-highfive]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @matklad (or someone else) soon.

If any changes to this PR are deemed necessary, please add them as extra commits. This ensures that the reviewer can see what has changed since they last reviewed the code. Due to the way GitHub handles out-of-date commits, this should also make it reasonably obvious what issues have or haven't been addressed. Large or tricky changes may require several passes of review and changes.

Please see the contribution instructions for more information.

[matklad]

should I add some tests for the old behaviour? It isn't really tested at the moment (no tests failed when I broke the implementation); I could refactor the tests to check for both single file and subdirectory inference

Adding test for the old behavior is very important, because backwards compatibility is one of the major constraints of Cargo. Refactoring tests to check for all flavors of conventions seems like a great idea!

Do we have something like servo's tidy check for coding style? I'm open for suggestions if something isn't formatted correctly

The only hard rule we have is that line length should be under 100 (it's checked on CI). Otherwise, it's a good idea to stick to an existing code style (which unfortunately may vary depending on which file you are editing now). We are not ready to use rustfmt yet.

Just a general one - should I rebase + squash commits every time I make subsequent changes to cargo?

It does not matter a lot, but an ideal appraoch I think is to do rebasing/squshing and other history rewrites, if necessary, and then ping the reviewer by the comment, because GH does not deliver notifications about force pushes.

I moved things around, mostly reusing the code from inferred_bins - hopefully I didn't break anything, but it won't hurt to double check :)

Will take a look shortly! In general, I feel rather confident about our tests :)

[matklad]

Changes look good to me! Perhaps we should get rid of inferred_(examples|benches|tests) one-liners and call infer_from_directory directly?

Also we need docs for this at https://github.com/rwakulszowa/cargo/blob/master/src/doc/manifest.md#the-project-layout.

And more tests are always welcome, especially if they test long-preexisting, but not currently covered functionality :)

[rwakulszowa]

I'm using the cargo test --test bar --test baz to check for existence of multiple tests currently. It seems that the logic is different for benchmarks - cargo bench allows me to pass only a single --bench=<name> argument (i.e.

cargo bench --bench=a --bench=b

throws an exception. Is this the expected behavior or is it a bug?

[matklad]

I would think it's a bug

[rwakulszowa]

I've added the requested changes. I've also created an issue for multiple --bench arguments (#4504) and marked the affected code with a #FIXME.

I also think I could clean up the tests for bin a little bit - don't you think this test already covers all cases for bin inference, so this one can be removed? Also, doesn't it have some duplicate assertions?

One more thing - currently my fake bench and test files only have a fn main(...) - is this good enough for these tests, since they are only supposed to test the build system, while benchmark specific tests are in bench.rs, or should I add some simple, but real code? This would complicate things a bit for benchmarks, since they only work with nightly cargo afaik.

[matklad]

don't you think this test already covers all cases for bin inference, so this one can be removed?

I don't think so. The second test is different in that it has an explicit [[bin]] section in Cargo.toml, so it tests that we can map from a binary name to its path corretly. It would be fine to merge these tests into one though.

Also, doesn't it have some duplicate assertions?

Yeah, some assetions indeed are twins.

One more thing - currently my fake bench and test files only have a fn main(...) - is this good enough for these tests, since they are only supposed to test the build system, while benchmark specific tests are in bench.rs, or should I add some simple, but real code?

It's ok to use fake code.

Looks like you've studied the relevant tests deeply, so, if you think you can organize them better than they are orginized currently, go ahead :) The current situation is pretty much historical, and I expect that the test suite can be made shorter and easier to undestand without compromizing test coverage.

So let's r+ this right away, and you can create a separate PR for tests cleanup then!

@bors r+

Thanks, @rwakulszowa !

[bors]

📌 Commit 6b94ed2 has been approved by matklad

[bors]

⌛ Testing commit 6b94ed2 with merge 7619c50...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: matklad
Pushing 7619c50 to master...