N runner running multiple runnable at the same time

[marioaag]

Could anybody explain me or point me to an example of how to run multiple runnable at the same time?

I can see that if the switch --test-runner='nrunner' is used avocado execute all the test in parallel but in someway the executions get isolated and looks like there is no way to specify the runnable values like the kind or the args.

In the other hand if I directly do it with the Python API like this:
runnable = nrunner.Runnable('python-unittest', 'test') runner_klass = runnable.pick_runner_class() runner = runner_klass(runnable) results = [status for status in runner.run()]

What is needed to do to get multiple runnable running at the same time? is it possible to pass multiple runnable to the same runner?

[richtja]

Hi @marioaag the runnables for each test are determinate by resolvers and avocado runs different kinds of tests based on the information from resolvers. Avocado will decide which runnable will be used for each test. So you can run multiple types of tests just by running them like this:
avocado run --test-runner=nrunner /bin/true examples/tests/passtest.py

You will find more information in https://avocado-framework.readthedocs.io/en/latest/guides/contributor/chapters/runners.html#finding-tests

For having bigger power over test execution you can use Avocado API, but with API you have to define everything by yourself. You can find usage in our selftest https://github.com/avocado-framework/avocado/blob/master/selftests/check.py

Small example of setting runnable by Avocado API:

#!/usr/bin/env python3

import sys

from avocado.core.job import Job
from avocado.core.suite import TestSuite
from avocado.core.nrunner import Runnable

config = {
    'run.test_runner': 'nrunner',
    'core.show': ['app']
}

suite = TestSuite(name="sleep", config=config, tests=[Runnable('exec-test', '/bin/sleep', "3")])

with Job(config, [suite]) as j:
    sys.exit(j.run())

[marioaag]

Thanks this is a really good example!! it shows exactly what I want to do...

Now I know that there are couple of runnable kinds (noop, exec, exec-test, python-unittest) but not sure about the purpose of each one of them and if anyone of them could be used to run avocado tests written with Python created for the actual runner?

[clebergnu]

Hi @marioaag,

You should find some information in their docstrings, for instance:

avocado/avocado/core/nrunner.py

Line 355 in 9ea1274

Sample runner that performs no action before reporting FINISHED status

avocado/avocado/core/nrunner.py

Line 374 in 9ea1274

Runner for standalone executables with or without arguments

avocado/avocado/core/nrunner.py

Line 433 in 9ea1274

Runner for standalone executables treated as tests

avocado/avocado/core/nrunner.py

Line 465 in 9ea1274

Runner for Python unittests

Now, when you say "tests written with Python created for the actual runner", do you mean Avocado's own instrumented tests? If so, the matching kind is avocado-instrumented:

avocado/avocado/core/runners/avocado_instrumented.py

Line 12 in 9ea1274

Runner for Avocado INSTRUMENTED tests

Let me know if that helps.

[marioaag]

Thanks! that helps

Now the question is, that is for the lates version GitHub right? if I'm running in Avocado 86.0 that runner is not listed as part of the capabilities

marioalvarado@MacBook-Pro avocado % avocado-runner capabilities
{"runnables": ["noop", "exec", "exec-test", "python-unittest"], "commands": ["capabilities", "runnable-run", "runnable-run-recipe", "task-run", "task-run-recipe"]}

is this a known issue?

Because looking a the source code I found that it is included but in a different place:

marioalvarado@MacBook-Pro avocado % ll core/nrunner_avocado_instrumented.py
-rw-r--r-- 1 marioalvarado staff 4514 Mar 25 13:32 core/nrunner_avocado_instrumented.py

And reading the code of core/nrunner.py looks like it will find it if I define it as part of the runnable kind

[clebergnu]

That's because one runner script may implement support for more than one runnable kind, but it's not required that all runners be built into one script (such as avocado-runner).

For instance, on my machine I have a number of runner-scripts such as:

 ls -1 ~/.local/bin/avocado-runner*
/home/cleber/.local/bin/avocado-runner
/home/cleber/.local/bin/avocado-runner-avocado-instrumented
/home/cleber/.local/bin/avocado-runner-avocado-vt
/home/cleber/.local/bin/avocado-runner-exec
/home/cleber/.local/bin/avocado-runner-exec-test
/home/cleber/.local/bin/avocado-runner-golang
/home/cleber/.local/bin/avocado-runner-magic
/home/cleber/.local/bin/avocado-runner-noop
/home/cleber/.local/bin/avocado-runner-python-unittest
/home/cleber/.local/bin/avocado-runner-robot
/home/cleber/.local/bin/avocado-runner-tap

You may have fewer, but the general point here is that the nrunner architecture can cope with standalone runners for specific kinds. They do not even need to be written in Python.

For an Avocado job to use one of those, it can either exist in the PATH or be registered as an entrypoint, such as here:

avocado/setup.py

Line 179 in 9ea1274

'avocado.plugins.runnable.runner': [