[Buildroot] [next] listing of tests is broken

[Thomas Petazzoni]

Hello,

On Sun, 11 Feb 2018 23:44:02 -0200, Ricardo Martincoski wrote:

> The correct fix IMO is to change the Rust test case.

Agreed.

> With the current code, nose2 will collect 3 test cases from the file:
> TestRustBase, TestRustBin and TestRust. It happens because all three have a
> method called test_something. But TestRustBase is not a test case by itself.
> 
> So we should move the test_* method from the base class to the 2 derived classes
> (yes, duplicating the code), like it is already done by the 'init' test cases.
> 
> An alternative solution is to rename the method in the base class to not start
> with test_ and create a test_run method in each derived class calling it. The
> method is short, so I don't think we need to do this.

Yes, duplicating would be fine for now.

> With the listing of tests broken, also 'make .gitlab-ci.yml' is broken. Based
> only on this I would say we should fix that on master branch. But see below...
> And also the update to .gitlab-ci.yml must be committed. This way we fix
> check-gitlab-ci.yml in GitLab:
> https://gitlab.com/buildroot.org/buildroot/-/jobs/52000035
> 
> But after fixing those, the TestRustBin still fails at GitLab:
> https://gitlab.com/RicardoMartincoski/buildroot/-/jobs/52092045
> from test_rust.py, line 39, like this:
> >SystemError: Cargo init failed  
> The build.log ends in these messages:
> >error: Failed to create project `hello-world` at `/tmp/tmp_US3dA-br2-testing-rust/hello-world`
> >
> >Caused by:
> >  could not determine the current user, please set $USER  
> The same test case does not fail for me when it runs outside the Docker image,
> so I guess we will need to tweak the Docker image.
> 
> 
> I don't have the result yet for TestRust in GitLab. It is still running (105
> minutes and counting) by the time I write this.
> https://gitlab.com/RicardoMartincoski/buildroot/-/jobs/52092044
> But running outside the Docker image it also passes.
> 
> ... Would it be the case to revert only the rust test case only from the master
> branch? If yes, extra care must be taken when merging the branches to not remove
> it from the next branch too.

I'd say, we fix the test enumeration problem, and then let the Rust
test case fail. Hopefully Eric Le Bihan (added in Cc) will chime in and
fix the Rust test failure.

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
http://bootlin.com