ci(github): loosen up file matching when selecting tests to run

[outSH]

• Current file match glob is incorrect and never matches a file.
• I didn't find a way to filter files by extension (as it was attempted now) so I propose to loose up the rules and run the test if any file in a package has changed.

More context

• I've noticed in another PR (chore(esm): change imports to esm compatible #3018) that most tests didn't run, even though this PR changes most ts files, from all packages I think.
• I've digged a little bit in dorny/paths-filter sources, found out the file matching string is passed to picomatch for actual filtering.
• picomatch is very simple (but fast), and does not support any advanced globbing techniques. I've done some tests locally, and the best solution I found is to use less strict matching rules.

Pull Request Requirements

• Rebased onto upstream/main branch and squashed into single commit to help maintainers review it more efficient and to avoid spaghetti git commit graphs that obfuscate which commit did exactly what change, when and, why.
• Have git sign off at the end of commit message to avoid being marked red. You can add -s flag when using git commit command. You may refer to this link for more information.
• Follow the Commit Linting specification. You may refer to this link for more information.

Character Limit

• Pull Request Title and Commit Subject must not exceed 72 characters (including spaces and special characters).
• Commit Message per line must not exceed 80 characters (including spaces and special characters).

A Must Read for Beginners
For rebasing and squashing, here's a must read guide for beginners.

[izuru0]

LGTM

[sandeepnRES]

Hi @outSH, I didn't get what you mean by never matches a file.
What I saw on PRs was, if there was a code change, the corresponding CI was running else it skipped.

[outSH]

Hi @outSH, I didn't get what you mean by never matches a file. What I saw on PRs was, if there was a code change, the corresponding CI was running else it skipped.

@sandeepnRES
Yeah, I've changed the original PR so it may not be clear, sorry. I've opened another one on my fork to make it more persitent here - outSH#11. As you can seet his PR changes all packages, but none are marked as changed by the CI:

This is because of invalid glob in config as I described above

[petermetz]

@outSH One more idea! Could you please double-check if the change detection is also broken if you submit the same diff (or something equivalent) to the upstream repo (e.g. hyperledger/cacti instead of outSH/cacti)?
This is just another wild guess but I'm thinking maybe it's something to do with the PR being submitted against your own fork.

This idea of mine was inspired by the fact that my benchmark result publishing PR of mine also seems to be broken in a way that is dependent on which repo a PR is being submitted against. 😅

[outSH]

@petermetz The problem was spotted on PR opened here, you can see this happening here - #3018

[sandeepnRES]

LGTM

[petermetz]

@outSH My theory right now is that this won't fix the issue because of the underlying cause explained here: #3018 (comment)

Is there any chance that you verify it over there on that PR?
I'm thinking that if my research is correct than we don't need this PR at all.

[petermetz]

@outSH LGTM - later once the path-filter action has the ability to support our more fine-grained filtering use-case I'll recommend to change the glob patterns once again, but in the meantime this is definitely a net positive! Thank you very much for taking the time to send this in!