test_that is the supported mechanism to run autotests against Chrome OS devices at your desk. test_that replaces an older script, run_remote_tests.


  • CTRL+C kills test_that and all its autoserv children. Orphaned processes are no longer left behind.
  • Tests that require binary autotest dependencies will just work, because test_that always runs from the sysroot location.
  • Running emerge after python-only test changes is no longer necessary. test_that uses autotest_quickmerge to copy your python changes to the sysroot.
  • Tests are generally specified to test_that by the NAME field of their control file. Matching tests by filename is supported using f:[file pattern]

Example uses (inside the chroot)

Run the test(s) named dummy_Pass:

$ test_that -b ${board} ${host} dummy_Pass

Run the test(s) named dummy_Pass.suspend:

$ test_that -b ${board} ${host} dummy_Pass.suspend

Run the smoke suite against dut:

$ test_that -b ${board} ${host} suite:smoke

Run all tests whose names match the regular expression ^login_.*$. Note that even though these tests have binary dependencies, there is no longer a need to specify extra flags.

$ test_that -b ${board} ${host} e:login_.*

Run all tests whose control file filename matches the regular expression ^.*control.dummy$:

$ test_that -b ${board} ${host} f:.*control.dummy

Running jobs in the lab

test_that now allows you to run jobs in the test lab. The usage is similar to running tests against a specified host. The keyword :lab: is used as test_that‘s REMOTE argument, and the -i/--build argument is required, and takes a trybot, paladin, or canary build number. To learn how to build a trybot image with a new test that you’re iterating on, see “dynamic suite” codelab.

For instance:

$ test_that -b lumpy -i lumpy-paladin/R38-6009.0.0-rc4 :lab: dummy_Pass

This will kick off a suite in the lab that consists of just 1 job, dummy_Pass, to run in this case on board lumpy using the image lumpy-paladin/R38-6009.0.0-rc4. The lab's scheduler will take responsibility for finding a suitable set of hosts, provisioning them to the correct image, and running the tests. test_that will return after the suite finishes running, with a suite run report.

You can specify multiple tests or test-matching expressions in the same way as before:

$ test_that -b lumpy -i ${latest_image} :lab: dummy_Pass dummy_Fail
$ test_that -b lumpy -i ${latest_image} :lab: e:login_.*

Kicking off a run in the lab should be useful whenever you need to run a particular test on a board or image that you do not have readily available locally.For occasional runs of ad-hoc suites in the lab, this will also avoid the need to create a suite control file and wait for it to end up in an image.

You can also kick off a suite, for example with:

test_that -b peach_pit :lab: suite:pyauto_perf -i 'peach_pit-release/R32-4763.0.0'

That told me that my job ID was 5196037. I could follow along by going to http://cautotest/afe/#tab_id=view_job&object_id=5195962.

Things to note:

This will only work with images newer than Sept 20, 2013 (specifically, builds that contain Ifa73d7de7aac9c6efebd5f559708623804ad3691). Jobs will be scheduled in the pool:try-bot machine pool.