Is `@pytest.mark.parallel_threads` intentionally deprecated?

[aeroyorch]

Hi, I noticed @pytest.mark.parallel_threads seems deprecated in favor of parallel_threads_limit.

Is this intended? I think parallel_threads is useful because it allows forcing some tests to always run in parallel.

[ngoldbaum]

It was intentionally deprecated, see #145 for the motivation.

I think parallel_threads is useful because it allows forcing some tests to always run in parallel.

Can you go into a little more detail into your usecase? Why doesn't parallel_threads_limit work?

[aeroyorch]

Our use case is that we need certain tests to always run with a specific number of parallel threads, rather than only limiting the number of threads that may be used.

For example, in pycurl/pycurl#999 we currently run certain tests explicitly with --parallel-threads=4. We initially considered using @pytest.mark.parallel_threads, but it appears to be deprecated in favor of parallel_threads_limit, which does not seem to be an exact replacement for this behavior.

Is there a recommended non-deprecated way to express “always run this test with N parallel threads” on a per-test basis?

[ngoldbaum]

Is there a recommended non-deprecated way to express “always run this test with N parallel threads” on a per-test basis?

Not right now, no. Our thinking with the deprecation was that the old naming and behavior was surprising and that most people want a limit rather than to force parallel testing.

I'd be open to adding a new mark (force_parallel_threads?) that makes it clear what it does.

[aeroyorch]

Is there a recommended non-deprecated way to express “always run this test with N parallel threads” on a per-test basis?

Not right now, no. Our thinking with the deprecation was that the old naming and behavior was surprising and that most people want a limit rather than to force parallel testing.

I'd be open to adding a new mark (force_parallel_threads?) that makes it clear what it does.

That solution makes sense to me :).

[ngoldbaum]

I opened #176, I'd appreciate it if you could give it a test in your project @aeroyorch.