Activity

1. webknjaz commented on May 4, 2026

   

   webknjaz

   on May 4, 2026

   Member

   More actions

   Copy of context @ #2035 (comment)

   I'm not aware of any tool or function that would support only blocking on new URLs, but we can add it separately. This PR is about not blocking working PRs on a broken linkcheck, including things such as getting changes from accepted PEPs linkcheck.

   So linkcheck produces JSONL like this: ansible/awx-plugins/actions/runs/24041509767#summary-70114297278. If we have that pre-downloaded into the job from artifacts/cache, we could have docs/conf.py read it and extract all the known URLs.

   Then, linkcheck_ignore could be populated with all those known URLs either right on the module level or from within the setup() function, but only in the context of PRs (and merge queues). This way, the PR runs would exclude checks against everything that used to be known prior to the PR merge, and new things would still be checked.

   ---

   Copy of context @ #2035 (comment)

   [..] I wanted to see where these reports are ending up now, in the current integration.

   I've found they are in the build dir:

   $ nox -s linkcheck
   nox > Running session linkcheck
   nox > Creating virtual environment (virtualenv) using python in .nox/linkcheck
   nox > python -m pip install -r requirements.txt
   nox > sphinx-build -b linkcheck -d .nox/.cache/.doctrees --color -n -W --keep-going source build
   Running Sphinx v7.2.6
   loading pickled environment... done
   [...]
   ^C
   Interrupted!
   nox > Interrupted...
   nox > Session linkcheck interrupted.
   $ ls -l build/output.txt build/output.json
   -rw-r--r-- 1 <username> <username> 161031 May  1 09:22 build/output.json
   -rw-r--r-- 1 <username> <username>  14242 May  1 09:22 build/output.txt
   

   There's an untested sketch for starters:

   $ git diff source/conf.py
   diff --git a/source/conf.py b/source/conf.py
   index 22b0e5e3..cdb3fe7c 100644
   --- a/source/conf.py
   +++ b/source/conf.py
   @@ -182,6 +182,12 @@ if _gh_token := os.getenv("GITHUB_TOKEN"):
            "https://api.github.com/": {"Authorization": f"Bearer {_gh_token}"},
        }
    
   +    if os.getenv("GITHUB_ACTIONS") == "true":
   +        output_dir = pathlib.Path("build/")
   +        output_jsonl_file = output_dir / "output.json"
   +        for json_object in output_jsonl_file.read_text().splitlines():
   +            linkcheck_ignore.append(json_object["uri"])  # FIXME: escape the regex
   +
    # -- Options for extlinks ----------------------------------------------------------
    # https://www.sphinx-doc.org/en/master/usage/extensions/extlinks.html#configuration

   Of course, this needs to be brushed up with the hardcoded dir replaced w/ a var + the GHA side would need to cache the file.
2. webknjaz

   added

   good first issue

   type: taskSomething that needs to be done that is not a bug or featureSomething that needs to be done that is not a bug or feature

   on May 4, 2026
3. webknjaz

   moved this to 📑 Ready 👌 in 📅 Procrastinating in public 😵‍💫on May 4, 2026
4. webknjaz

   added this to 📅 Procrastinating in public 😵‍💫on May 4, 2026
5. github-project-automation

   moved this from 📑 Ready 👌 to 🗒️ Backlog 📝 in 📅 Procrastinating in public 😵‍💫on May 20, 2026
6. harshitethic commented on Aug 30, 2026

   

   harshitethic

   on Aug 30, 2026 · Hidden as spam

   show commentMore actions
7. MugdhoAI commented on Oct 22, 2026

   

   MugdhoAI

   on Oct 22, 2026

   More actions

   I’m interested in contributing to this issue and have reviewed the proposed approach. Before I start, could a maintainer confirm whether this is still available to work on, given the existing expression of interest, and whether the intended direction is to reuse the previous linkcheck output in PR and merge queue runs so only newly introduced URLs are checked?

   I would verify the current workflow and add focused tests before opening a PR.
8. merwok commented on Oct 22, 2026

   

   merwok

   on Oct 22, 2026

   Contributor

   More actions

   We prefer contributions from people interested in becoming contributors, or even interesting in using their time, energy and intellect to solve a problem. A human interface to a robotic tool adds more burden than help to the few reviewers.

   Reacted by sinoroc and Ralf Gommers
9. rgommers

   removed

   good first issue

   on Nov 11, 2026
10. rgommers commented on Nov 11, 2026

    

    rgommers

    on Nov 11, 2026

    Contributor

    More actions

    This is very much not trivial (see 7 linked PRs in the sidebar), and it is quite annoying that CI is red all the time, so I had a go at this in gh-2182. The high-level approach follows what @webknjaz suggested.