TheKevJames / coveralls-python / f055a015-9742-4bab-b062-2fd917832bf1

93%

master: 96%

Pull #754

circleci

refactor(config)!: resolve every source through one uniform pipeline

Replace the ad-hoc per-service loaders (positional tuples, three different
pull-request parsers, the _coveralls_host/_token_required attrs, and call-time
os.environ reads for host/SSL) with a single resolve() that merges the CI
environment, COVERALLS_* env vars, the config file, and explicit overrides in
one documented precedence order. CI detection returns a uniform dict and every
service's pull-request value now goes through the one trailing-integer helper.
That also fixes TravisCI: a non-PR build sets TRAVIS_PULL_REQUEST=false, which
was previously submitted raw as service_pull_request='false'; the shared helper
drops the 'false' sentinel like the Buildkite branch already did.

token_required is not one of those merged config values: it is a calculated
guard against accidental tokenless uploads. resolve() takes it as an explicit
argument (the CLI derives it from --debug/--output) and waives it only on a CI
service that authenticates uploads itself (TOKENLESS_CI_SERVICES, e.g.
TravisCI). It is never read from the config file or env vars, so a committed
.coveralls.yml cannot silently disable the check.

changelog:
Configuration from every source -- CI detection, ``COVERALLS_*`` env vars, the
config file, and explicit overrides -- is now resolved through one pipeline with
a single documented precedence order (later wins), replacing the per-service
special cases. On TravisCI, non-pull-request builds no longer submit a bogus
``service_pull_request`` of ``"false"``. **Breaking (library API):**
``Coveralls()`` no longer accepts ``service_name`` as a dedicated argument; pass
it as a keyword override like any other field (e.g.
``Coveralls(service_name="travis-pro")``). Positional calls such as
``Coveralls(True, "travis-pro")`` must be updated.

129 of 140 branches covered (92.14%)

Branch coverage included in aggregate %.

114 of 118 new or added lines in 2 files covered. (96.61%)

414 of 442 relevant lines covered (93.67%)

5.62 hits per line