Test failure for 0.7.0 on Arch Linux

[orhun]

Hey, I'm getting the following test failure while building the latest version for Arch Linux (in a clean chroot):

failures:

---- test_ignore_target stdout ----
thread 'test_ignore_target' panicked at src/main.rs:310:5:
assertion failed: entries.unwrap().is_empty()
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

failures:
    test_ignore_target

test result: FAILED. 12 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.98s

error: test failed, to rerun pass `--bin cargo-machete`

Do I need any preparation before running the tests? Any ideas about it? 🤔

As a workaround I currently --skip this test in the package.

[bnjbvr]

Thanks for the report. That's weird, I can't repro on my (MacOS) machine; are you using the version from a github release, or installed from cargo, or building from a specific git hash?

[orhun]

I'm pulling the release from GitHub (0.7.0) and building in an isolated environment. See the PKGBUILD for build steps.

Let me know if there is any other information is needed!

[bnjbvr]

Thanks for the details; unfortunately I can't reproduce on my machine. If anyone can reproduce and track the test failure, that would be greatly appreciated 🙏

[orhun]

Maybe you can tell me what that test is doing, how is it being run and why it might fail. Then I can dig into it a bit :)

[bnjbvr]

The test checks collect_paths, which has options to ignore target directories (first assertion), or to ignore files mentioned in ignore files (second assertion, where there's /target in ./integration-tests/with-target/target/program/.gitignore), or that when none of these is enabled, we do find some entries.

If the .gitignore file is missing from the archive you're downloading, I could see that test failing. Otherwise it's not clear why it's not failing: is the .gitignore file found? is it then not ignored, or interpreting /target as a different path than ./integration-tests/with-target/target/?

[orhun]

Okay I found the culprit. The downloaded archive from GitHub is not a git repository (does not contain .git folder) so I'm guessing the underlying library does not take .gitignore into account in that case.

The test passes when I rename .gitignore to .ignore 🫠

[bnjbvr]

Thanks for the investigation! I think we could rename .gitignore to .ignore and not lose any testing coverage, since this is testing that any kind of ignore files is taken into account. Thoughts?

[orhun]

Yup, sounds good to me 👍🏼