Understanding the test discovery (Avocado Loaders)¶
In this section you can learn how tests are being discovered and how to customize this process.
Some definitions here may be out of date. The current runner can still be using some of these definitions in its design, however, we are working on an improved version of the runner, the NextRunner that will use an alternative strategy.
A Test Loader is an Avocado component that is responsible for discovering tests that Avocado can run. In the process, Avocado gathers enough information to allow the test to be run. Additionally, Avocado collects extra information available within the test, such as tags that can be used to filter out tests from actual execution.
This whole process is, unless otherwise stated or manually configured, safe, in the sense that no test code will be executed.
How Loaders discover tests¶
Avocado will apply ordering to the discovery process, so loaders that run earlier, will have higher precedence in discovering tests.
A loader implementation is free to implement whatever logic it needs to discover tests. The important fact about how a loader discover tests is that it should return one or more “test factory”, an internal data structure that, as stated before, contains enough information to allow the test to be executed.
The order of test loaders¶
As described in previous sections, Avocado supports different types of test starting with SIMPLE tests, which are simply executable files, the basic Python unittest and tests called INSTRUMENTED.
With additional plugins new test types can be supported, like the avocado-vt ones, which uses complex matrix of tests from config files that don’t directly map to existing files.
Given the number of loaders, the mapping from test names on the command line to executed tests might not always be unique. Additionally some people might always (or for given run) want to execute only tests of a single type.
To adjust this behavior you can either tweak
plugins.loaders in avocado
/etc/avocado/), or temporarily using
--loaders (option of
avocado run) option.
This option allows you to specify order and some params of the available test
loaders. You can specify either loader_name (
file), loader_name + TEST_TYPE
file.SIMPLE) and for some loaders even additional params passed after
external:/bin/echo -e. You can also supply
injects into that position all the remaining unused loaders.
Example of how
--loaders affects the produced tests (manually gathered as
some of them result in error):
$ avocado run passtest.py boot this_does_not_exist /bin/echo > INSTRUMENTED passtest.py:PassTest.test > VT io-github-autotest-qemu.boot > MISSING this_does_not_exist > SIMPLE /bin/echo $ avocado run passtest.py boot this_does_not_exist /bin/echo --loaders @DEFAULT "external:/bin/echo -e" > INSTRUMENTED passtest.py:PassTest.test > VT io-github-autotest-qemu.boot > EXTERNAL this_does_not_exist > SIMPLE /bin/echo $ avocado run passtest.py boot this_does_not_exist /bin/echo --loaders file.SIMPLE file.INSTRUMENTED @DEFAULT external.EXTERNAL:/bin/echo > INSTRUMENTED passtest.py:PassTest.test > VT io-github-autotest-qemu.boot > EXTERNAL this_does_not_exist > SIMPLE /bin/echo
A Test Reference is a string that can be resolved into (interpreted as) one or more tests by the Avocado Test Resolver.
Each resolver (a.k.a. loader) can handle the Test References differently. For example, External Loader will use the Test Reference as an argument for the external command, while the File Loader will expect a file path.
If you don’t specify the loader that you want to use, all of the available loaders will be used to resolve the provided Test References. One by one, the Test References will be resolved by the first loader able to create a test list out of that reference.
Basic Avocado Loaders¶
Below you can find some extra details about the specific builtin Avocado loaders. For Loaders introduced to Avocado via plugins (VT, Robot, …), please refer to the corresponding loader/plugin documentation.
For the File Loader, the loader responsible for discovering INSTRUMENTED, PyUNITTEST (classic python unittests) and SIMPLE tests.
If the file corresponds to an INSTRUMENTED or PyUNITTEST test, you can filter
the Test IDs by adding to the Test Reference a
: followed by a regular
For instance, if you want to list all tests that are present in the
gdbtest.py file, you can use the list command below:
$ avocado list examples/tests/gdbtest.py INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_start_exit INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_existing_commands_raw INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_existing_commands INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_load_set_breakpoint_run_exit_raw INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_load_set_breakpoint_run_exit INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_generate_core INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_set_multiple_break INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_disconnect_raw INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_disconnect INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_remote_exec INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_stream_messages INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_connect_multiple_clients INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_server_exit INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_multiple_servers INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_server_stderr INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_server_stdout INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_remote
To filter the results, listing only the tests that have
their test method names, you can execute:
$ avocado list examples/tests/gdbtest.py:test_disconnect INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_disconnect_raw INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_disconnect
As the string after the
: is a regular expression, two tests were
filtered in. You can manipulate the regular expression to have only the
test with that exact name:
$ avocado list examples/tests/gdbtest.py:test_disconnect$ INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_disconnect
The regular expression enables you to have more complex filters. Example:
$ avocado list examples/tests/gdbtest.py:GdbTest.test_[le].*raw INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_existing_commands_raw INSTRUMENTED examples/tests/gdbtest.py:GdbTest.test_load_set_breakpoint_run_exit_raw
Once the test reference is providing you the expected outcome, you can
list subcommand with the
run subcommand to execute your
$ avocado run examples/tests/gdbtest.py:GdbTest.test_[le].*raw JOB ID : 333912fb02698ed5339a400b832795a80757b8af JOB LOG : $HOME/avocado/job-results/job-2017-06-14T14.54-333912f/job.log (1/2) examples/tests/gdbtest.py:GdbTest.test_existing_commands_raw: PASS (0.59 s) (2/2) examples/tests/gdbtest.py:GdbTest.test_load_set_breakpoint_run_exit_raw: PASS (0.42 s) RESULTS : PASS 2 | ERROR 0 | FAIL 0 | SKIP 0 | WARN 0 | INTERRUPT 0 | CANCEL 0 JOB TIME : 1.15 s JOB HTML : $HOME/avocado/job-results/job-2017-06-14T14.54-333912f/html/results.html
Specially when using regular expressions, it’s recommended
to individually enclose your Test References in quotes to avoid bash
of corrupting them. In that case, the command from the example above
avocado run "examples/tests/gdbtest.py:GdbTest.test_[le].*raw"
Using the External Loader, Avocado will consider that and External Runner will be in place and so Avocado doesn’t really need to resolve the references. Instead, Avocado will pass the references as parameters to the External Runner. Example:
$ avocado run 20 Unable to resolve reference(s) '20' with plugins(s) 'file', 'robot', 'vt', 'external', try running 'avocado -V list 20' to see the details.
In the command above, no loaders can resolve
20 as a test. But running
the command above with the External Runner
/bin/sleep will make Avocado
to actually execute
/bin/sleep 20 and check for its return code:
$ avocado run 20 --loaders external:/bin/sleep JOB ID : 42215ece2894134fb9379ee564aa00f1d1d6cb91 JOB LOG : $HOME/avocado/job-results/job-2017-06-19T11.17-42215ec/job.log (1/1) 20: PASS (20.03 s) RESULTS : PASS 1 | ERROR 0 | FAIL 0 | SKIP 0 | WARN 0 | INTERRUPT 0 | CANCEL 0 JOB TIME : 20.13 s JOB HTML : $HOME/avocado/job-results/job-2017-06-19T11.17-42215ec/html/results.html
It’s safer to put your Test References at the end of the
command line, after a –. That will avoid argument vs. Test
References clashes. In that case, everything after the – will
be considered positional arguments, therefore Test References.
Considering that syntax, the command for the example above would be:
avocado run --loaders external:/bin/sleep -- 20
This loader enables Avocado to execute binaries or scripts and parse their Test Anything Protocol output.
The tests can be run as usual:
$ avocado run --loaders tap -- ./mytaptest
Notice that you have to be explicit about the test loader you’re
using, otherwise, since the test files are executable binaries, the
FileLoader will detect the file as a
SIMPLE test, making the
whole test suite to be executed as one test only from the Avocado
perspective. Because TAP test programs should exit with a zero exit
status, this will cause the test to pass even if there are failures.