Regression in gitlab-ci: coverage regex not recognised

[quentin-ag]

Context

This follows from #436, about the wrong report 'coverage: [...] is not a regex' with the gitlab-ci schema. One solution was to use the following regex (minimal example of .gitlab-ci.yml file):

tests:
  coverage: /TOTAL.*? (100(?:\.0+)?\%|[1-9]?\d(?:\.\d+)?\%)$/

Problem

The solution from #436 worked until version 0.30.0 of check-jsonschema. This can be reproduced with the minimal example above. Assuming it is the content of the file min-gitlab-ci.yml:

python -m pip install --quiet 'check-jsonschema==0.30'
check-jsonschema --builtin-schema "gitlab-ci" min-gitlab-ci.yml

Result:

ok -- validation done

This now fails with version 0.31:

python -m pip install --quiet 'check-jsonschema==0.31'
check-jsonschema --builtin-schema "gitlab-ci" min-gitlab-ci.yml

Result:

Schema validation errors were encountered.
  min-gitlab-ci.yml::$.tests.coverage: '/TOTAL.*? (100(?:\\.0+)?\\%|[1-9]?\\d(?:\\.\\d+)?\\%)$/' is not a 'regex'

Is this a regression? Should the regex be changed?
Thanks in advance for any support on this.

Environment details

OS: Arch Linux
Python: 3.9.21

[sirosen]

This is a regression, but it's not in the schema! Thanks for the MRE, it made it easy to confirm.

In v0.31.0 the default regex behavior became ECMAScript regexes in unicode mode. However, for GitLab specifically that's probably the wrong choice, since GitLab's regexes are in the ruby engine (which we cannot perfectly support).

I was able to get your example to work by adding --regex-variant nonunicode.
Can you let me know if that works for your full use-case? It should be a viable workaround for now, on v0.31.0.

I plan to include a switch to this for the pre-commit hook in the next release, but confirmation that the setting works well would be good.

[quentin-ag]

Thanks for the quick reply 🙏

I can confirm that your workaround works for me on both the minimal example and on my full GitLab CI configuration 🎉

python -m pip install --quiet 'check-jsonschema==0.31'
check-jsonschema --builtin-schema --regex-variant nonunicode "gitlab-ci" min-gitlab-ci.yml
# ok -- validation done
check-jsonschema --builtin-schema --regex-variant nonunicode "gitlab-ci" .gitlab-ci.yml
# ok -- validation done

I'm happy with the workaround and I look forward to the switch in the next release. Feel free to close this ticket whenever you feel is appropriate. Thanks again!