(sorry for the revert, I erronously included the commit from PR #8 - the revert is to fix that)

Uh I'm a little confused about this, so correct me if I have this all wrong @leny.

I see two separate issues here:

1. The current action appears to error out when there is nothing to commit.

This does not, on the face of it, make much sense.
I can think of any number of scenarios where, for some commits to the source do not cause a diff in the build asset dir/.
Erroring it clearly isn't the right response here, because there's nothing wrong per se.
I also tend to think that 78 status might not be quite right, because as per the GitHub actions spec:

When an action returns this exit status, GitHub terminates all concurrently running actions and prevents any future actions from starting.

This, again, in some scenarios isn't what users might want or expect.

Also, I think this problem should only occur in the first place for deploys from . where there already (qua github actions provisioning of the workspace) is a git repo. In any subdir/, there won't be a git repo, and so there'd by definition always be something to commit.
Thus just as an aside.

2. If deploying from ., this might cause an "infinite loop" of sorts (if 1 were fixed)..
That's also bad.
I wonder though, whether the correct response to this wouldn't be to:

1. Commit sources to one branch, and deploy from some other branch (gh-pages).
2. Simply add a filter action before the deploy, and have the workflow continue only for !gh-pages or even just master, or whatever the plan may be.

Sorry this got more complicated than expected.

So my (preliminary) suggestion would be to just exit 0 instead of exit 78 for empty commits and rely on filter actions to avoid infinite loops.

@leny Does that make sense? Does that meet user expectations and keep things simple to reason about?

Set me straight if I got this wrong.

The more I think of it, the more I think the exit 0 is what makes more sense, and the other cases should be addressed with a filter action, as you suggest.

perfect, will merge, test and ship this then.
Thanks for the quick turnaround and great contributions @leny!