Fix improper class.filename and package.name values on XML coverage report

[lormico]

Solves #862

[arcivanov]

flake8 warnings. @lormico

[lormico]

Whoooops, fixed it

[coveralls]

Pull Request Test Coverage Report for Build 2812061099

• 2 of 2 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+2.3%) to 84.672%

---

Totals
Change from base Build 2767987572:	2.3%
Covered Lines:	5241
Relevant Lines:	6140

---

💛 - Coveralls

[arcivanov]

@lormico failures on Windows.

https://github.com/pybuilder/pybuilder/runs/7585801888?check_suite_focus=true#step:3:1471

[lormico]

It looks like the Windows failures are not caused by my edit on coverage_plugin.py: I tried reverting my changes and disabling the assertions in the integration test, and it still fails when running the coverage step.

Maybe there's something wrong with GitHub's Windows testing platform? Should I make PyBuilder skip my test if "win" in sys.platform?

[arcivanov]

No, it is not. It's the issue with disks/locations.

ValueError: path is on mount 'c:', start on mount 'D:'

[arcivanov]

@lormico it's the change related to path manipulation. Current tree works fine:

https://github.com/pybuilder/pybuilder/runs/7593136238?check_suite_focus=true

[lormico]

Ok, I wanted to investigate deeper on this issue, so I created a branch on lormico/pybuilder from the latest commit on pybuilder/pybuilder/master and:

1. removed all integration tests (merely for build speedup)
2. added a new integration test which only sets up a simple project using the methods provided by itest_support.IntegrationTestSupport and then calls the coverage task (no assertions)

I still keep getting the same error, even if the src/main codebase is the same as the one on pybuilder/master, so I'd exclude that it's because of something that was introduced recently.

This particular error:

ValueError: path is on mount 'c:', start on mount 'D:'

seems to be due to how Python itself handles relative paths on Windows, when two different volumes are involved (here a very old ticket from 2009 which explains this better than I could do).

I'm still not completely sure about how this is connected to PyBuilder, but I'm more inclined to think that's because of how GitHub Actions work or how they are configured for a Windows VM. Alas, I have nearly zero experience with GitHub Actions, so I might be off track.

Do you have any pointers from here @arcivanov?

[arcivanov]

@lormico I took another dig at it and it looks like a bug in coverage

nedbat/coveragepy#1428

[arcivanov]

@lormico Please rebase on the the latest. The test also needs to be modified to not fail for unrelated reasons:

       File "d:\a\pybuilder\pybuilder\target\dist\pybuilder-0.13.7.dev\pybuilder\execution.py", line 95, in execute
        self.callable(*arguments)
        ^^^^^^^^^^^^^^^^^^^^^^^^^
      File "d:\a\pybuilder\pybuilder\target\dist\pybuilder-0.13.7.dev\pybuilder\plugins\python\coverage_plugin.py", line 157, in coverage
        raise failure
        ^^^^^^^^^^^^^
    pybuilder.errors.BuildFailedException: Test coverage for at least one module is below 70%
    
    ----------------------------------------------------------------------

Very close now 😄

[lormico]

I ran the Actions on my fork (I had to disable the upload step for pyb in order to make it run successfully, because of permissions), but during the build PyBuilder seems to have downloaded an old version of coverage, and failed on these two instances: [windows-latest, 3.11.0-beta.5, true, false] and [windows-latest, 3.11.0-beta.5, false, false].

The error reveals it is not using the latest version:

message(f"Combined data file {os.path.relpath(f)}")  # should be message(f"Combined data file {file_name}")

even though the dependency specification in pyb is coverage~=6.0, which should be ok.

I'm doing some tests to find out what is the cause.

[lormico]

Nevermind, it now doesn't seem to fail anymore.

Here is the Actions run on my fork.

@arcivanov can you please approve launching the workflow?

[arcivanov]

@lormico Thank you, merged.