The behaviour of pg_regress is to exit with
- 0 if all tests passed,
- 1 if some tests failed,
- 2 if the tests could not be run for some reason
I think cargo pgrx regress should follow the same principle.
Currently when the regress command is run inside a CI job, the only way to catch an error is to parse the ouput which is ironic since cargo pgrx regress itself parse the psql output :)
Anyway I'm willing to fix this, but the change will (rigthfullty) break some CI jobs that were previously passing and break some pipelines so before going any further, I'd like your advice on this :)