fix: validate the --contexts regexes up front - #2262

An invalid regex in --contexts reports a confusing error that blames the data file.

$ coverage report --contexts='foo('
Couldn't use data file '/tmp/covtest/.coverage': user-defined function raised exception

After:

Invalid context regex 'foo(': missing ), unterminated subpattern at position 3

The data file is fine. The regex goes straight to the SQLite REGEXP user function, where re.search raises inside sqlite, and sqlite reports only that some user function failed. Same for coverage json --contexts=, coverage html --contexts=, the [report] contexts config setting, and Coverage.report(contexts=[...]).

The fix

process_regexlist in config.py already does exactly this for every other regex-valued setting:

try:
    re.compile(value)
except re.error as e:
    raise ConfigError(f"Invalid [{name}].{option} value {value!r}: {e}") from e

contexts is the one that never got it, because it is not compiled at config time, it is handed to sqlite later. So the same check runs in set_query_contexts, which is the point every entry path funnels through.

Testing

test_set_query_contexts_bad_regex in tests/test_data.py.

Removing only the validation loop fails it, and the failure is the original symptom rather than a message mismatch:

        with pytest.raises(ConfigError, match=msg):
>           raise DataError(f"Couldn't use data file {self.filename!r}: {msg}") from exc
E           coverage.exceptions.DataError: Couldn't use data file '...': user-defined function raised exception

pytest tests/test_data.py is 4 passed for the selected tests.

Pre-existing failures

The full suite is 143 failed, 1450 passed, 23 skipped, 14 errors on a clean checkout here, and 143 failed, 1451 passed, 23 skipped, 14 errors with this change. A diff of the sorted FAILED lists is empty, so every one of those is pre-existing in this environment, which has no C extension built and lacks the venv/subprocess test dependencies.

Disclosure: written with AI assistance (Claude Code). I reproduced the confusing error at the CLI, checked all four entry paths, confirmed the pre-existing failures by diffing against a stashed tree, and ran the mutation check myself.