@pocmatos did you mean to remove “fetch-depth: 0”?

It’s still there in the commit that you pushed

i push forced it … you sure?

omg - right. I push forced an empty commit apparently and left the file unstaged locally. sigh

thanks

Crazy idea: Instead of diffing the files, is it possible for GitHub Action to overwrite the existing file on a push action? This is similar to how people use git pre-commit hook to cleanup trailing whitespaces.

Though I guess there’s a benefit of seeing the error

Not a good idea to push to your own repo from CI, I think.

It is definitely possible, I think if you give secret to the CI to push to the repo. Then you need to ensure (and there’s currently no good way) that a push to the repo does not itself trigger CI.

so you have a CI infinite loop.

I looked into it in the past and talked to a few expert CI people and the agreement seems to be that the CI for project X should not itself push to X.

Therefore I took the easy road and instead fail the CI.

yep, makes sense

The user will be notified of this on PR or Push and fix it.

The other possibility is to have a pre-commit hook that runs the checks before the commit through a make check or something.

But we can’t enforce the pre-commit hook for everyone, right?

That’s my biggest letdown for Git hooks. It looks like people need to individually set it up.

right.

I think as part of a process, we can add pre-commit hooks that people should use…

or…

suggest that people should use …

(forgot the name of the program)

lol

one sec

suggest that people should use act locally before committing.

https://github.com/nektos/act

although unfortunately act doesn’t work well a lot of the time, at least for our actions.

would pushing to pull requests be more reasonable?

Yes - that would be cool indeed.

“Hey, you forgot to generate the build.md file, here’s a PR that does it for you.” :slightly_smiling_face:

I like that.

I have been asking this for years but I will ask again. Who wants to work on a github api pkg for racket?

:slightly_smiling_face:

The one I know of looks unmaintained and has a pagination issue: https://github.com/eu90h/racket-github-api/issues/4

I guess then we would need a github user to represent a racket bot so that CI could create the PRs as him.

The racket bot should be called rackot?

racoon?

:wink:

ohhhhhhhhhhhhhh noooooooooooooo

the set-env command is disabled in CI - a lot of our workflows will fail… :disappointed:

wow they moved quickly from deprecating them and disabling them.

OK - I removed all instances of set-env. It should be back working now.