restructuredtext_lint.test.test.TestRestructuredtextLintCLI.test_level_fail fails

[mcepl]

When packaging this package (1.4.0 using the tarball from PyPI) this test failed:

[   22s] ======================================================================
[   22s] FAIL: test_level_fail (restructuredtext_lint.test.test.TestRestructuredtextLintCLI.test_level_fail)
[   22s] Confirm low --level threshold fails file with warnings only
[   22s] ----------------------------------------------------------------------
[   22s] Traceback (most recent call last):
[   22s]   File "/home/abuild/rpmbuild/BUILD/python-restructuredtext_lint-1.4.0-build/restructuredtext_lint-1.4.0/restructuredtext_lint/test/test.py", line 210, in test_level_fail
[   22s]     self.assertEqual(output.count('\n'), 2, output)
[   22s] AssertionError: 1 != 2 : WARNING /home/abuild/rpmbuild/BUILD/python-restructuredtext_lint-1.4.0-build/restructuredtext_lint-1.4.0/restructuredtext_lint/test/test_files/second_short_heading.rst:6 Title underline too short.

Complete build log with all packages used listed and all steps required for reproduction included.

[twolfson]

Thanks for reporting this! I suspect it's tied to #64 and its associated deprecation error.

I'm currently low on bandwidth so marking this as "help wanted".

If you want to take a shot at patching this, feel free to do so =)

[twolfson]

I finally had some time to triage this, though not fully patch it (it seems a number of things are broken in the development setup, as typically happens with time/age -- e.g. needed to switch to nose-py3 instead of nose)

The issue looks like it previously was expecting a double output which I guess is now resolved -- was:

https://github.com/twolfson/restructuredtext-lint/blob/1.4.0/restructuredtext_lint/test/test.py#L202-L204

        # $ rst-lint --level warning second_short_heading.rst ; echo "Return code $?"
        # WARNING second_short_heading.rst:6 Title underline too short.
        # WARNING second_short_heading.rst:6 Title underline too short.

Now is:

$ rst-lint --level warning second_short_heading.rst
WARNING second_short_heading.rst:6 Title underline too short.

So the good news is rst-lint is still behaving in a good fashion, the bad news is the test suite needs to be corrected (which I haven't touch in a long time + need to see if need to support multiple Python versions or not)

As I mentioned, I don't have much free time to patch this, so I'll hopefully come back some time soon (could be a few days, could be a few months) -- preferably sooner obv

[mcepl]

e.g. needed to switch to nose-py3 instead of nose

No, you don’t. I think you are doing just fine with the plain unittest (and python -munittest [discover] for runner), and if you ever really need to use different testing framework, please, please, please, use pytest. nose is dead and I spent hundreds of hours in openSUSE (and I know my colleagues at Fedora and Debian did as well) to eliminate it everywhere from our packages. Please.

[twolfson]

It feels like you might've misread "needed to switch [nose versions as part of triage]" as "need to add nose to this repo". This repo is already using nose. It's been using it for 12 years.

Thank you for the feedback and alternatives. I'll need to weigh that when doing maintenance here (usually try to not swap something that's working "good enough" since behaviors can change when doing so, and no idea if nose had some special behavior that was preferable then)

[mcepl]

On Thu Oct 30, 2025 at 6:35 PM CET, Todd Wolfson wrote: It feels like you might've misread "needed to switch [nose versions as part of triage]" as "need to add nose to this repo". This repo is already using `nose`. [It's been using it for 12 years](https://github.com/twolfson/restructuredtext-lint/blame/f943e89ea7544ffc6d73cfc3832d4fb36cd1e399/requirements-dev.txt#L1).

I understood you perfectly correctly, and I understand that you have been using `nosetests` test runner. And I don’t care about that much at all. What I meant when I was saying is that you don’t depend on nose is that your code is so far 100% clean of imports from nose packages, and so the test suite works perfectly well without any `nose` on our systems. I wanted to encourage you to keep your code still nose-free, and if you see anything which you think the standard library and `unittest` doesn’t offer, don’t look for it in `nose` (I don’t think you would find it there anyway, `unittest` these days is on parity or even better than `nose` ever was), but either use special package for `unittest` or switch to `pytest`.

[twolfson]

Thank you for restating your points and attempting to come to a better place of understanding among dependencies/packages.

Let's keep the issue on track for what you originally reported (i.e. test suite assertions failing), as I feel we're spending a lot of unnecessary time on a tangential topic.

I'm still low on bandwidth to fully understand and fix the double output issue, so I'll hopefully come back some time soon (could be a few days, could be a few months) -- preferably sooner is better obv

[mcepl]

Let's keep the issue on track for what you originally reported (i.e. test suite assertions failing), as I feel we're spending a lot of unnecessary time on a tangential topic.

I understood you to claim that the problem was just that docutils fixed something, so it didn’t double the reply any more, and so that something like this (which is what I use in the openSUSE package at the moment) is appropriate:

---
 restructuredtext_lint/test/test.py |    7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

Index: restructuredtext_lint-1.4.0/restructuredtext_lint/test/test.py
===================================================================
--- restructuredtext_lint-1.4.0.orig/restructuredtext_lint/test/test.py 2022-02-24 06:50:23.000000000 +0100
+++ restructuredtext_lint-1.4.0/restructuredtext_lint/test/test.py      2025-10-30 11:37:31.488257443 +0100
@@ -201,14 +201,13 @@
         # This is the expected behaviour we are checking:
         # $ rst-lint --level warning second_short_heading.rst ; echo "Return code $?"
         # WARNING second_short_heading.rst:6 Title underline too short.
-        # WARNING second_short_heading.rst:6 Title underline too short.
         # Return code 2
         with self.assertRaises(subprocess.CalledProcessError) as e:
             subprocess.check_output((sys.executable, rst_lint_path, '--level', 'warning', warning_rst),
                                     universal_newlines=True)
         output = str(e.exception.output)
-        self.assertEqual(output.count('\n'), 2, output)
-        self.assertEqual(output.count('WARNING'), 2, output)
+        self.assertEqual(output.count('\n'), 1, output)
+        self.assertEqual(output.count('WARNING'), 1, output)
         # The expected 2 warnings should be treated as failing
         self.assertEqual(e.exception.returncode, 2)

[twolfson]

Yep, that's correct. The externalities I'd want to spend more time on understanding:

• When did docutils implement this?
• Does the test suite run for multiple docutils versions or just one?
• Does the test suite even still integrate with CI or does that need an overhaul? (since maintenance here is infrequent)
• And likely many more thoughts when I start to dig in

I can prob remove the "help wanted" flag since the actual patch seems trivial, but all the other context is something I'd want to vet directly

[twolfson]

Picking this up finally, it should be released with a migration and update to our CI suite (#67)