Sync mypy_primer_comment with typeshed

[sobolevn]

This PR adds a useful feature: now empty primer output will still add a comment: "According to mypy_primer, this change has no effect on the checked open source code."

Why is it useful?
Because in some cases it is not clear: is mypy_primer happy? Or is it broken?
For example, this PR #12118 does not have any comment.

I guess it would also benefit PR authors who can be sure that what they did is correct.

I will wait for mypy_primer to comment on this PR 🙂

[sobolevn]

It looks like it still runs the old version: https://github.com/python/mypy/runs/5064684428?check_suite_focus=true

[JelleZijlstra]

Yes, actions run based on what's on master, not on the PR.

I think this difference was intentional, perhaps because some mypy PRs (e.g., mypyc changes) don't affect the type checker and aren't expected to make a difference in mypy-primer. Then again, that's also true for some typeshed changes.

[sobolevn]

I guess if that's an issue we can add - files: spec to mypy_primer action.
So, mypyc only changes - will be ignored.

[hauntsaninja]

Yeah, the version of the action in the PR doesn't run to avoid security issues, since the action has write permissions on PRs.

I seem to remember some doubt about this being noisy, but I couldn't find where that was discussed and people on typeshed don't seem to mind, so let's go for it. Note that the real mypy-primer workflow already ignores mypyc, so we don't need to add files here.

The one change I would like is to minimise differences in the comment workflow between this and typeshed. E.g., I think typeshed acquired some nice change that uses a glob instead of manually listing the shards / the comment hider workflow runs in a different order due to not hiding on mypy. Let's update this PR to make the mypy version match typeshed as much as possible

[sobolevn]

I accidentally commited the requested changes to master: 26fdd45

Not quite used to the fact I can do that 😆
I hope it is fine, because the change in 26fdd45 is self-contained.

If needed, it can be reverted.

[sobolevn]

the comment hider workflow runs in a different order due to not hiding on mypy

I will address this in a separate PR, because I will need to investigate the changes there. They seem to be rather big 🤔

[sobolevn]

I am going to merge it and test with my upcoming tuple PR 🎉

[hauntsaninja]

Thanks, sounds good, I'm familiar with the comment hiding stuff, so can sync it myself :-)

[sobolevn]

@hauntsaninja I am working on right now 😊