data_suffix being appened to data_file name when it's None / False.

[amanagr]

When we set data_suffix to False / None, data_file's is appended with a suffix like .coverage.ubuntu-bionic.2681.658565. The documentation (https://coverage.readthedocs.io/en/coverage-5.1/api_coverage.html?highlight=data_suffix) says, suffix will only be appended when it's value is True, which is not the case here.
Setting data_suffix to "" (empty string) appears to work though.

coverage version = 5.1

[nedbat]

From your linked commit, you have concurrency='multiprocessing', which is what forces the suffix. Don't you need a suffix on the data file? Aren't you recording coverage from a number of processes?

[amanagr]

@nedbat I am sorry, but I don't get your point here. I don't want the final data file to have any suffix. I want it named as .coverage so that codecov can easily find it and upload the coverage results. We are recording coverage from a number of processes.

Am I correct in saying that each process generates a different .coverage.SUFFIX file and we can merge them using .combine on data? If so, then either the .combine or .save should have an argument to specify the file name / data_suffix.

[amanagr]

Also, can you please tell if setting "data_suffix=""` will pose any problems here? (Or, if there is a better solution for what I want to do there.)

[nedbat]

When running with multiple processes, each process needs to write its data to a different file. The data suffix is what keeps the files separate. Then when the tests are done, "coverage combine" is used to combine all the files together into a suffix-less file.

It's possible that some of your data is not being collected properly if you are not using suffixes on your data files. Are you using "coverage combine"?

[nedbat]

Sorry, I re-read your comment. If you are already using .combine, then everything should work properly if you simply don't specify data_suffix at all. Your "concurrency=multiprocessing" argument will create the correct data files (with a suffix), and then .combine will combine them into a suffix-less file, ready for codecov.

[amanagr]

Even withcov.combine, I see the final file named .coverage.ubuntu-bionic.20843.916684
where cov = coverage.Coverage(config_file="tools/coveragerc", concurrency='multiprocessing').

My coveragerc:

[report]
# Regexes for lines to exclude from consideration
exclude_lines =
    # Re-enable the standard coverage pragma
    nocoverage
    # Don't complain if non-runnable code isn't run:
    if False:
    # Don't require coverage for base class NotImplementedErrors
    raise NotImplementedError()
    # Don't require coverage for test suite AssertionError -- they're usually for clarity
    raise AssertionError
    # Don't require coverage for __str__ statements just used for printing
    def __str__[(]self[)] -> .*:
    # Don't require coverage for errors about unsupported webhook event types
    raise UnexpectedWebhookEventType
    # Don't require coverage for blocks only run when type-checking
    if TYPE_CHECKING:
    # Don't require coverage for abstract methods; they're never called.
    @abstractmethod
    # Don't require coverage for the settings.LOG_API_EVENT_TYPES code paths
    # These are only run in a special testing mode, so will fail normal coverage.
    if settings.LOG_API_EVENT_TYPES:
    # PEP 484 overloading syntax
    ^\s*\.\.\.

[run]
omit =
    */zulip-venv-cache/*
    */migrations/*
    */management/commands/*
    # Parts of the test runner infrastructure
    tools/test-backend
    zerver/lib/test_fixtures.py
    zerver/lib/test_runner.py
    # Has its own independent test suite
    zerver/openapi/python_examples.py
    zerver/lib/parallel.py
    # Debugging tools that don't lend themselves well to unit tests
    zerver/lib/debug.py
    # Part of provisioning/populate_db
    zerver/lib/generate_test_data.py
    # Excluded because its coverage state is flaky.
    zerver/tornado/ioloop_logging.py
    # Zulip's library for use in scripts
    scripts/lib/zulip_tools.py
    # Used only for the legacy Zephyr integration, and unlikely to ever be unit-tested
    zerver/lib/ccache.py
    # Settings.py files are hard to test
    zproject/*settings.py
    # https://github.com/davidhalter/jedi/issues/1122
    blub

[nedbat]

Can you make a small example that duplicates how you are using the coverage API that shows the problem?

[amanagr]

test-backend has the complete code.

Code which saves data file with suffix:

cov = coverage.Coverage(config_file="tools/coveragerc", concurrency='multiprocessing')
cov.start()
cov.stop()
cov.combine()
cov.save()

You can take coveragerc same as my last comment.
Setting cov = coverage.Coverage(data_suffix="", config_file="tools/coveragerc", concurrency='multiprocessing') removes the suffix.