Errors when running bats test locally

[Stratus3D]

Steps to reproduce

On OSX run:

bats test

Expected behavior

Tests should pass as they were before.

Actual behavior

Many failing tests are encountered. And some tests hang indefinitely. An example error I captured from a single test:

/Users/trevorbrown/.asdf/bin/asdf: line 82: /var/folders/7m/9bq62tmj60zf14d788w0lk8c0000gp/T/asdf.XXXX.ed4MwuLg/home/.asdf/lib/commands/command-help: No such file or directory

Environment

OS: 10.15.1
asdf version: v0.7.6-bde7ae5

[Stratus3D]

@vic do you know what is going on? I've been running bats test for years and I'm not sure why I'm seeing so many failures now.

[Stratus3D]

The asdf update is a noop for non-git repos test hangs indefinitely.

[klane]

I am having the same issue, which makes debugging #633 challenging. Manually running asdf with my changes seems to work as expected.

[Stratus3D]

On tag v0.7.5 I only get one failure:

 ✗ get_config returns default when config file does not exist
   (in test file test/get_asdf_config_value.bats, line 29)
     `[ "$result" = "no" ]' failed

On tag v0.7.6 I get 65 failures before it reaches the asdf update is a noop for non-git repos test and it hangs indefinitely.

[Stratus3D]

I get the exact same failures and hanging test on my Ubuntu laptop as well.

[vic]

Try using https://github.com/bats-core/bats-core (just like our CI is doing) and not https://github.com/sstephenson/bats, the later has not been updated in years.

[Stratus3D]

@vic I am using bats-core version 1.1.0. I don't think the bats version is the issue.

[Stratus3D]

Do test pass for you @vic ? What OS are you on?

[vic]

Yep, all green here. 212 test, 0 failures.

Also green in CI macos catalina / ubuntu.
https://github.com/asdf-vm/asdf/runs/375715344#step:7:219

I'm using Mac OS Catalina.
bats-core 1.2.0-dev (commit 87b16eb8ec381beca7f05d028e4ad7cf3f647a0f)

git --version
git version 2.21.0 (Apple Git-122)

[Stratus3D]

I ran git bisect with the failing tests (git bisect run bats test/current_command.bats) and got:

e968fe6142d121d1da11aeada21b0338b62b9b16 is the first bad commit
commit e968fe6142d121d1da11aeada21b0338b62b9b16
Author: Victor Hugo Borja <vborja@apache.org>
Date:   Wed Nov 27 02:24:29 2019 -0600

    Subcommand-aware command line interface.

This is a very large commit, and it's not immediately clear what the problem is, but I am continuing to dig.

[lahwran]

CI seems to be failing as well, and has been since November. Any idea what's going on there? Is that a separate issue or the same one?

[vic]

@lahwran The only failing check is the Appveyor one, it's still a work in progress (we have never been successful in running tests under Windows). Maybe we should remove the appveyor check until it's stable.

[Stratus3D]

I found the problem! Because I'm running asdf on my computer ASDF_DIR was already set and was interfering with the tests. By unsetting the variable all tests pass:

unset ASDF_DIR
bats test

The asdf update is a noop for non-git repos test is still hanging for me. I'm guessing it is caused by a similar issue. Will investigate later.

[Stratus3D]

I created #644 to fix that hanging test.