fix(ci): multiline regex in change detection

[villebro]

SUMMARY

Currently the CI test skipping logic applies a regex on a variable containing all modified files. As these checks are checking for start of line (^) and the regular =~ doesn't support multiline matching, the script has been changed to loop through each row in the changed files variable, and flag the first row where changes were detected.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

TEST PLAN

Tested with a PR where changes weren't being picked up correctly:

$ GITHUB_REPO=apache/superset PR_NUMBER=13051 ./ci_check_no_file_changes.sh python
Searching for changes in python files

CHANGED FILES:
.github/workflows/docker_build_push.sh
superset/config.py
superset/connectors/sqla/views.py
superset/views/core.py
superset/views/datasource.py

Detected changes in following file: superset/config.py
Exiting with FAILURE code

Checking this PR:

$ GITHUB_REPO=apache/superset PR_NUMBER=13075 ./ci_check_no_file_changes.sh python frontend
Searching for changes in python files
Searching for changes in frontend files

CHANGED FILES:
scripts/ci_check_no_file_changes.sh

No changes detected... Exiting with SUCCESS code

ADDITIONAL INFORMATION

• Has associated issue:
• Changes UI
• Requires DB Migration.
• Confirm DB Migration upgrade and downgrade tested.
• Introduces new feature or API
• Removes existing feature or API

[dpgaspar]

LGTM, just some shellcheck nit picking. What do you think about adding shellcheck to our CI?

[dpgaspar]

nit: shellcheck complains about:

  REGEXES=(${REGEXES[@]} ${REGEX})
           ^-----------^ SC2206: Quote to prevent word splitting/globbing, or split robustly with mapfile or read -a.
                         ^------^ SC2206: Quote to prevent word splitting/globbing, or split robustly with mapfile or read -a.

[dpgaspar]

nit:

In 1.sh line 54:
  for REGEX in ${REGEXES[@]}
               ^-----------^ SC2068: Double quote array expansions to avoid re-splitting elements.

[villebro]

@dpgaspar fixed - I'm wondering if we should lint our scripts with ShellCheck? Seems like a good idea IMO