Thank you all for the feedback! I understand the concern about adding more noise, and I agree that the current docs build CI provides sufficient feedback. I won't pursue the CI workflow approach.
2026년 5월 9일 (토) 오전 6:23, Vincent Beck <[email protected]>님이 작성: > Very much agree. I am quite against adding a new workflow and will yet add > more noise, more message to read. The rather keep the CI as is. > > On 2026/05/08 21:18:00 Jarek Potiuk wrote: > > Sorry - sent to fast. We have enough of those reminders already ... When > > the docs will be built on CI and fail in your PR - you already have > > sufficient feedback your doc build fails - no additional workflow is > > needed. > > > > On Fri, May 8, 2026 at 11:13 PM Jarek Potiuk <[email protected]> wrote: > > > > > I think prompts to check rendering are counter-productive. We do not > want > > > more messages or reminders. We have enough > > > > > > On Fri, May 8, 2026 at 8:10 AM gui <[email protected]> wrote: > > > > > >> Hi everyone, > > >> First of all, thank you all for sharing your thoughts and suggestions. > > >> The discussion really helped me investigate the options more > thoroughly. > > >> > > >> After testing both tools, I found some limitations: > > >> 1. **rstcheck** has false positives in our codebase > > >> 2. **sphinx-lint** didn't catch the heading level issue from PR #66252 > > >> > > >> Given these findings, I'd like to propose a lighter-weight > alternative: > > >> > > >> Instead of adding a linter, we could add a GitHub CI workflow that > posts > > >> a reminder comment on PRs that modify `.rst` files, prompting the > author > > >> to check RST rendering on GitHub. > > >> > > >> This approach: > > >> - Doesn't require fixing existing documentation issues > > >> - Avoids false positives from linters > > >> - Provides just-in-time reminders during code review > > >> > > >> If this direction isn't preferred, that's completely fine - I just > wanted > > >> to share what I found during the investigation. > > >> I'd love to hear your thoughts on this alternative approach. > > >> > > >> Best regards, > > >> Yunhui Chae > > >> > > >> 2026년 5월 3일 (일) 오전 12:57, Jens Scheffler <[email protected]>님이 작성: > > >> > > >> > I am also supportive but in my past (but worked ~2 years ago on > this) no > > >> > checker was really "good" and I had massive false-positives. Hope > the > > >> > static checkers have improved as early feedback can be helpful > before CI > > >> > runs for long and fails. Unfortunately the static check seems not > to be > > >> > easy. > > >> > > > >> > On 02.05.26 17:22, Jarek Potiuk wrote: > > >> > > Hi Yunhui, > > >> > > > > >> > > Please proceed with the PR. I agree with Shahar that documentation > > >> files > > >> > > should likely be excluded initially, as Sphinx verifies them and > they > > >> use > > >> > > extensions that might trigger false positives in basic checkers. > > >> > > > > >> > > However, if you find a tool that can run on the docs/ folder > without > > >> > > excessive noise or easily fixable failures - as Piyush mentioned, > it > > >> > would > > >> > > be a valuable addition. Flagging issues like missing empty lines > > >> before > > >> > > lists locally via pre-commit would be better than waiting for CI > > >> results. > > >> > > It may be difficult to keep it noise-free, but it is worth > > >> investigating. > > >> > > > > >> > > Thanks, > > >> > > Jarek Potiuk > > >> > > > > >> > > On Sat, May 2, 2026 at 2:47 PM Piyush Mudgal < > > >> > [email protected]> > > >> > > wrote: > > >> > > > > >> > >> I support this proposal. Adding an RST linter to pre-commit hooks > > >> will > > >> > help > > >> > >> contributors ensure documentation is correctly formatted before > > >> > submission. > > >> > >> > > >> > >> Best, > > >> > >> Piyush Mudgal > > >> > >> > > >> > >> On Sat, May 2, 2026 at 4:34 PM Shahar Epstein <[email protected] > > > > >> > wrote: > > >> > >> > > >> > >>> I support this idea, as long as it targets RST files intended > for > > >> > >>> GitHub reading (mostly development-facing docs). Automatically > > >> > >>> generated RST files should be excluded to avoid noisy failures > and > > >> > >>> keep the hook focused on files contributors edit directly. > Later, we > > >> > >>> could use such a linter to improve the templates used to > generate > > >> > >>> those files, but that requires some more research and can wait > for a > > >> > >>> later stage. > > >> > >>> > > >> > >>> > > >> > >>> Shahar > > >> > >>> > > >> > >>> On Sat, May 2, 2026 at 1:17 PM gui <[email protected]> wrote: > > >> > >>>> Hi everyone, > > >> > >>>> > > >> > >>>> I'd like to propose adding an RST linter to our pre-commit > hooks. > > >> > >>>> > > >> > >>>> ## Motivation > > >> > >>>> > > >> > >>>> Recently, PR #66252 [1] was submitted to fix an RST heading > level > > >> > error > > >> > >>>> that broke GitHub rendering. Currently, such syntax errors are > only > > >> > >>> caught > > >> > >>>> during the documentation build process, which delays feedback > for > > >> > >>>> contributors. By adding an RST linter, we can catch these > issues > > >> > >> locally > > >> > >>>> before the code is even pushed. > > >> > >>>> ## Current State > > >> > >>>> > > >> > >>>> We have `rst-backticks` hook but no RST syntax validation. > > >> > >>>> > > >> > >>>> ## Proposal > > >> > >>>> > > >> > >>>> Add either `rstcheck` [2] or `sphinx-lint` [3] to pre-commit: > > >> > >>>> > > >> > >>>> ```yaml > > >> > >>>> # rstcheck > > >> > >>>> - repo: https://github.com/rstcheck/rstcheck > > >> > >>>> rev: v6.2.5 > > >> > >>>> hooks: > > >> > >>>> - id: rstcheck > > >> > >>>> additional_dependencies: ['rstcheck[sphinx,toml]'] > > >> > >>>> ``` > > >> > >>>> > > >> > >>>> Both tools catch RST syntax errors early. `rstcheck` is more > > >> > >>> comprehensive; > > >> > >>>> `sphinx-lint` is lighter and Sphinx-focused. > > >> > >>>> > > >> > >>>> ## Note > > >> > >>>> > > >> > >>>> Pre-commit hooks only run on changed files by default, so > existing > > >> > >>>> documentation won't break. We can incrementally fix existing > issues > > >> > >> over > > >> > >>>> time rather than in one big bang. > > >> > >>>> > > >> > >>>> If there's interest, I can prepare a PR with the > implementation. > > >> > >>>> > > >> > >>>> Best regards, > > >> > >>>> Yunhui Chae > > >> > >>>> > > >> > >>>> [1] https://github.com/apache/airflow/pull/66252 > > >> > >>>> [2] https://github.com/rstcheck/rstcheck > > >> > >>>> [3] https://github.com/sphinx-contrib/sphinx-lint > > >> > >>> > > >> --------------------------------------------------------------------- > > >> > >>> To unsubscribe, e-mail: [email protected] > > >> > >>> For additional commands, e-mail: [email protected] > > >> > >>> > > >> > >>> > > >> > > > >> > > --------------------------------------------------------------------- > > >> > To unsubscribe, e-mail: [email protected] > > >> > For additional commands, e-mail: [email protected] > > >> > > > >> > > > >> > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
