Google Git
Sign in
fuchsia / fuchsia / refs/heads/main / . / src / testing / end_to_end / honeydew / tests / functional_tests
tree: d37838d5b90d6430fc62a8276e5008fdcc8d104d [path history] [tgz]
  1. affordance_tests/
  2. auxiliary_device_tests/
  3. fuchsia_device_tests/
  4. transport_tests/
  5. __init__.py
  6. BUILD.gn
  7. README.md
src/testing/end_to_end/honeydew/tests/functional_tests/README.md

Honeydew functional tests

Every Honeydew host target interaction API must have at least one functional test case that ensures the API works as intended. For example, a functional test checks that <device>.reboot() actually reboots the Fuchsia device.

Which PRODUCT and BOARD to use

Running a functional test case requires a Fuchsia device. There can be many different Fuchsia <BOARD>s that support multiple <PRODUCT> configurations. Decide which <PRODUCT> and <BOARD> to use following this approach:

  • By default, use the lowest possible <PRODUCT> configuration that the corresponding functional test requires.
  • For example,
    • to run trace affordance test cases, use core
    • to run session affordance test cases, use terminal (as core does not support it)

<BOARD>

  • By default, run the test on femu unless the test can’t be verified using femu (because femu does not yet support the functionality that corresponding functional test needs). For example, bluetooth functional tests can’t be run on femu.
  • Else, run on NUC hardware if that test can be verified using NUC.
  • Else, run on VIM3 hardware if that test can be verified using VIM3.
  • Run on actual Fuchsia product only if:
    • None of the above conditions work.
    • If a specific Host-(Fuchsia)Target interaction API (say <device>.reboot()) has a special implementation for this particular product that requires explicit verification.
  • Special cases - If a Host-(Fuchsia)Target interaction API has multiple implementations then ensure all the different implementations have been verified using whatever device type is needed.

How to run a functional test locally

  1. Check that the device type that the test runs on (will be listed in “device_type” field in test case's BUILD.gn file) is connected to the host and detectable by FFX:

    $ ffx target list
NAME                SERIAL       TYPE             STATE      ADDRS/IP                           RCS
fuchsia-emulator*   <unknown>    core.x64         Product    [fe80::1a1c:ebd2:2db:6104%qemu]    Y

    Learn more about setting up these devices:

  2. Determine if your local testbed requires a manual local config. For more information, refer to Lacewing Mobly Config YAML file) Tip: You can most likely skip this step, if your test needs only one fuchsia device and if your host has only that one device connected.

  3. Run appropriate fx set command along with --with //src/testing/end_to_end/honeydew/tests/functional_tests param.

    • If test case requires SL4f transport, then make sure to do the following:
      • Use a <PRODUCT> that supports SL4F (such as core)
      • Include --args=core_realm_shards+="[\"//src/testing/sl4f:sl4f_core_shard\"]"
        • Alternatively, run fx set without the --args option, and then run fx args, and add the line directly to the args.gn file in the editor that opens.

  4. If test needs SL4F, then run a package server in a separate terminal

    $ fx -d <device_name> serve

  5. Run fx test command and specify the target along with --e2e --output args

Here is an example to run a functional test on emulator

$ fx set core.x64 --with //src/testing/end_to_end/honeydew/tests/functional_tests:tests

# start the emulator with networking enabled
$ ffx emu stop ; ffx emu start -H --net tap

$ fx test //src/testing/end_to_end/honeydew/tests/functional_tests/fuchsia_device_tests/test_fuchsia_device:x64_emu_test --e2e --output

How to add a new test to run in infra

  • Refer to CQ VS CI vs FYI section to decide which test case build group the new test belongs to.
  • Once decided, you can add it accordingly to that group by updating Honeydew Infra Test Groups
  • If the desired group is not present in Honeydew Infra Test Groups, then in addition to creating a new group here, you may also need to:
    • Add the same group in Lacewing Infra Test Groups (follow the already available test groups in this file for reference)
    • Update the corresponding Lacewing builder configuration file (maintained by Foundation Infra team) to include this newly created group in Lacewing Infra Test Groups. If your test depends on target side packages, make sure to include the appropriate *_packages group. Please reach out to Lacewing team if you need help with this one. And also, please include Lacewing team as one of the reviewer in this CL.
  • Update the test case's python_mobly_test rule in BUILD.gn to include appropriate BOARDS (based on what all the boards this test need to be run in infra) in environments field (example)

CQ VS CI vs FYI

Lacewing test case can be run in Infra in of the following builder types:

  • CQ builder - Builds and tests pending CLs before they get checked in, to make it less likely that breakages will make their way to main (and thus break the CI).
  • CI builder - Builds and tests the code that has been checked in. Any failure in this builder will be reported Fuchsia Build gardeners and need to be triaged ASAP.
  • FYI builder - Build and and tests the code that has been checked in, but unlike CI these builders are not expected to be green at all times.

Use the following approach in deciding whether to run the test case in CQ/CI/FYI:

  • Anytime a new Lacewing test needs to be run in infra, always start in FYI.
  • After 200 consecutive successful runs, test case can then be promoted to CQ/CI based on below criteria:
    • If infra has enough testbed capacity (that is needed to run this test) then it can be promoted to CQ.
    • If infra has limited testbeds (that is needed to run this test) to run this test then it can be promoted to CI.

Once your test case is ready to be promoted to CI/CQ:

  • If a corresponding stable Honeydew Infra Test Groups exists, and that test group is already enabled in a CQ builder, simply move your test target from unstable to stable.
  • Otherwise, reach out to the Lacewing team for further assistance.

Accordingly, we have created the following:

  • Test groups:
    • Test group naming scheme: <PRODUCT>_<BOARD>_<STABILITY>[ |_sl4f]_tests
      • <PRODUCT> - The product that the tests require to run on - e.g. “core”, “workbench”.
      • <BOARD> - The board that the tests require to run on - e.g. “emulator”, “nuc”, “vim3”.
      • <STABILITY> - Whether tests are stable or flaky - “stable” or “unstable”. All newly added tests must be added to the “unstable” groups until 200 passing runs in infra FYI builder have been observed, after which they may be promoted to “stable” groups.
      • [ |_sl4f] - If tests require SL4F server then include “_sl4f”. Otherwise, leave it empty.
      • Examples:
        • core_emulator_stable_tests
        • workbench_vim3_unstable_tests
    • This naming scheme is chosen to facilitate infra builder configuration.
      • <STABILITY> informs whether a group is potential to run in CI/CQ (as it depends on <BOARD> also).
      • [ |_sl4f] informs whether a test group can be run on certain products.
      • <BOARD> informs whether a test group can be run on certain boards.
  • Package groups:
    • Package group naming scheme: <PRODUCT>_<BOARD>_<STABILITY>[ |_sl4f]_packages
  • Builders:
    • Package group naming scheme: <PRODUCT>.<BOARD>-[debug|release]-lacewing-[ |staging|ci]
      • If builder is for “CQ” then no postfix is needed but for other stages, postfix is necessary (-staging or -ci)
    • Examples
      • core.x64-debug-lacewing - CQ builder to run stable emulator and NUC tests
      • core.x64-debug-lacewing-staging - FYI builder to run unstable emulator and NUC tests
      • core.vim3-debug-lacewing-staging - FYI builder to run unstable VIM3 tests