[Yusaku01]

Description (required)

Purpose

The Japanese lint workflow currently reports findings in job logs and emits a generic failure annotation. Show each finding at the relevant line in the pull request so contributors can locate it without searching the logs.

With the GitHub formatter, printing the saved log again would emit each annotation command twice. Removing that reporting step avoids this repetition.

Why textlint 15.8.0?

The GitHub formatter was introduced in v15.5.0. v15.8.0 also adds per-file log groups, so this version provides both inline annotations and grouped CI logs.

Verification

• All 148 Japanese MDX files pass with both 14.7.2 and 15.8.0, with caching disabled. A test violation retains the same rule, location, and suggested fix.
• Draft demonstration PR uses the same workflow and dependency versions with one intentional violation. It shows one inline annotation, and the CI log confirms exit status 1 and one local-command hint. A passing fork run passed with the same lint command.
• Local checks with both formatters produce the same finding and exit status. The failure handler preserves exit statuses 0/1/2.

Screenshot from earlier fork verification; the demonstration PR above keeps a live example available.

One CLI behavior change is relevant: when the target pattern matches zero files, local checks returned 0 with 14.7.2 and 2 with 15.8.0. Failing in this case exposes missing lint coverage; the current documentation glob succeeds.

CI timing

Two earlier fork runs per version used the same input, including one intentional violation, and had no pnpm cache hits. GitHub's whole-second timings exclude queue time.

Measurement	14.7.2	15.8.0 + GitHub formatter
Japanese lint job	23s, 27s	27s, 24s
textlint step	2s, 2s	2s, 1s

No substantial increase appeared in this small sample. These measurements precede restoration of the local-command hint.

Sources: baseline attempts, updated PR run, updated manual run.

References

• textlint's GitHub formatter
• textlint v15 migration notes

[github-actions]

Preview deployment

✅ Deployment complete!

• Latest commit: cd6099f
• Preview: https://ci-textlint-github-annotations-upstream.previews.docs.astro.build
• Workflow run: View logs

[ArmandPhilippot]

Thanks, the changes make sense to me! However, we have some conflicts (probably because of #14782), and we have another ongoing PR updating deps (#14784) so we might end up with new conflicts. 😅

I'm pinging @morinokami to ensure this is what Japanese translators want, as textlint is a bit unusual in our translation process. And also to let him know he can merge this if the conflicts are resolved and I'm not around at that time! No need to wait! ✔️

[Yusaku01]

Thanks, the changes make sense to me! However, we have some conflicts (probably because of #14782), and we have another ongoing PR updating deps (#14784) so we might end up with new conflicts. 😅

I'm pinging @morinokami to ensure this is what Japanese translators want, as textlint is a bit unusual in our translation process. And also to let him know he can merge this if the conflicts are resolved and I'm not around at that time! No need to wait! ✔️

@ArmandPhilippot
Thank you for checking!! I've resolved the current conflicts for now!
I'll just wait and see how it goes. 🙏

[morinokami]

@ArmandPhilippot Thanks for the ping! I like this change, as it makes it much easier to spot problematic lines at a glance instead of digging through CI logs. (Maybe we could do the same for broken links detected by CI?)

@Yusaku01 Thanks for your contribution! LGTM!!

[ArmandPhilippot]

Thanks, Shinya!

Maybe we could do the same for broken links detected by CI?

Isn't that already the case? I can see the same kind of annotations in https://github.com/withastro/docs/pull/14452/changes#diff-64bc37c347d5fd63b00742ef646432bfe2bada3efd75618bfe7f7a384bc9b41f

The issue, I suppose, is when errors occur in an unchanged file? I'm not sure how we could improve that but, yeah, a more friendly way to discover them would be helpful. I'll try to explore ideas!

[morinokami]

@ArmandPhilippot

Isn't that already the case?

Haha, you're absolutely right! I've been looking at the logs this whole time and somehow missed that 😅 Thanks for pointing that out, and for looking into possible improvements!

[Yusaku01]

@ArmandPhilippot @morinokami

Thank you both for reviewing and merging!
I thought it would be better if it were easier to read, as long as the CI time and the actual generated logs remained the same, so this was really helpful!😌