The pipeline jobs (LAVA v2) have retained compatibility with respect to the content of Lava-Test-Shell Test Definitions although the submission format has changed:
deploy block is encountered.There are 3 types of test actions:
definitions) are used for POSIX compliant operating
systems on the DUT. The deployed system is expected to support
a POSIX shell environment (/bin/ash, /bin/dash or /bin/bash
are the most common) so that LAVA can execute the LAVA Test Shell Helper
scripts.monitors) are used for devices which have no POSIX shell and start
the test (and corresponding output) immediately after booting, for
example microcontroller/IoT boards.interactive) are further extension of “monitor” tests idea, allowing
not just matching some output from a device, but also feeding some input.
They are useful for non-POSIX shells like bootloaders (u-boot for instance)
and other interactive command-line applications.See also
Contents
A publicly readable repository location.
The type of the repository is not guessed, it must be specified
explicitly. Support is available for git. Support is planned
for url and tar.
A simple test definition present in the same file as the job submission,
instead of from a separate file or VCS repository. This allows tests to be run
based on a single file. When combined with file:// URLs to the deploy
parameters, this allows tests to run without needing external access. See
Inline test definition example.
(optional): Pass parameters to the Lava Test Shell Definition. The format is a YAML dictionary - the key is the name of the variable to be made available to the test shell, the value is the value of that variable.
- test:
definitions:
- repository: https://git.linaro.org/lava-team/hacking-session.git
from: git
path: hacking-session-debian.yaml
name: hacking
params:
IRC_USER: ""
PUB_KEY: ""
- test:
definitions:
- repository: git://git.linaro.org/lava-team/lava-functional-tests.git
from: git
path: lava-test-shell/smoke-tests-basic.yaml
name: smoke-tests
- repository: https://git.linaro.org/lava-team/lava-functional-tests.git
from: git
path: lava-test-shell/single-node/singlenode03.yaml
name: singlenode-advanced
When a single test definition is to be used across multiple deployment types
(e.g. Debian and OpenEmbedded), it may become necessary to only perform certain
actions within that definition in specific jobs. The skip_install support
has been migrated from V1 for compatibility. Other methods of optimizing test
definitions for specific deployments may be implemented in V2 later.
The available steps which can be (individually) skipped are:
skip running lava-install-packages for the deps: list of the
install: section of the definition.
skip running lava-add-sources for the sources: list of the install:
section of the definition.
identical to ['deps', 'keys', 'sources', 'steps']
Example syntax:
- test:
failure_retry: 3
name: kvm-basic-singlenode
timeout:
minutes: 5
definitions:
- repository: git://git.linaro.org/lava-team/lava-functional-tests.git
from: git
path: lava-test-shell/smoke-tests-basic.yaml
name: smoke-tests
- repository: http://git.linaro.org/lava-team/lava-functional-tests.git
skip_install:
- all
from: git
path: lava-test-shell/single-node/singlenode03.yaml
name: singlenode-advanced
The following will skip dependency installation and key addition in the same definition:
- test:
failure_retry: 3
name: kvm-basic-singlenode
timeout:
minutes: 5
definitions:
- repository: git://git.linaro.org/lava-team/lava-functional-tests.git
from: git
path: lava-test-shell/smoke-tests-basic.yaml
name: smoke-tests
- repository: http://git.linaro.org/lava-team/lava-functional-tests.git
skip_install:
- deps
- keys
from: git
path: lava-test-shell/single-node/singlenode03.yaml
name: singlenode-advanced
https://git.lavasoftware.org/lava/lava/blob/master/lava_dispatcher/tests/sample_jobs/kvm-inline.yaml
- test:
failure_retry: 3
definitions:
- repository:
metadata:
format: Lava-Test Test Definition 1.0
name: smoke-tests-basic
description: "Basic system test command for Linaro Ubuntu images"
os:
- ubuntu
scope:
- functional
devices:
- panda
- panda-es
- arndale
- vexpress-a9
- vexpress-tc2
run:
steps:
- lava-test-case linux-INLINE-pwd --shell pwd
- lava-test-case linux-INLINE-uname --shell uname -a
- lava-test-case linux-INLINE-vmstat --shell vmstat
- lava-test-case linux-INLINE-ifconfig --shell ifconfig -a
- lava-test-case linux-INLINE-lscpu --shell lscpu
- lava-test-case linux-INLINE-lsusb --shell lsusb
- lava-test-case linux-INLINE-lsb_release --shell lsb_release -a
from: inline
name: smoke-tests-inline
path: inline/smoke-tests-basic.yaml
The V2 dispatcher supports some additional elements in Lava Test Shell which will not be supported in the older V1 dispatcher.
LAVA collects results from internal operations, these form the lava test
suite results as well as from the submitted test definitions. The full set of
results for a job are available at:
results/1234
LAVA records when a submitted test definition starts execution on the test
device. If the number of test definitions which started is not the same as the
number of test definitions submitted (allowing for the lava test suite
results), a warning will be displayed on this page.
A TestSet is a group of lava test cases which will be collated within the LAVA Results. This allows queries to look at a set of related test cases within a single definition.
- test:
definitions:
- repository:
run:
steps:
- lava-test-set start first_set
- lava-test-case date --shell ntpdate-debian
- ls /
- lava-test-case mount --shell mount
- lava-test-set stop
- lava-test-case uname --shell uname -a
This results in the date and mount test cases being included into a
first_set TestSet, independent of other test cases. The TestSet is
concluded with the lava-test-set stop command, meaning that the uname
test case has no test set, providing a structure like:
results:
first_set:
date: pass
mount: pass
uname: pass
{'results': {'first_set': {'date': 'pass', 'mount': 'pass'}, 'uname': 'pass'}}
Each TestSet name must be valid as a URL, which is consistent with the requirements for test definition names and test case names in the V1 dispatcher.
For TestJob 1234, the uname test case would appear as:
results/1234/testset-def/uname
The date and mount test cases are referenced via the TestSet:
results/1234/testset-def/first_set/date
results/1234/testset-def/first_set/mount
A single test definition can start and stop different TestSets in sequence, as long as the name of each TestSet is unique for that test definition.
An interactive test action allows to interact with a non-POSIX shell or just arbitrary interactive application. For instance, the shell of u-boot bootloader.
The workflow of the interactive test action is:
command to the DUT, unless emptyecho: discard is specified, discard next output line (assumed to be
an echo of the command)prompts, successes or failuresname is defined, log the result for this command (as soon as a prompt or a message is matched)successes or failures was matched, wait for the promptsNote
The interactive test action expects the prompt to be already matched
before it starts. If this is not the case, then wait for the prompt by
adding an empty command directive as described below.
A u-boot interactive test might look like:
- test:
interactive:
- name: network
prompts: ["=>", "/ # "]
echo: discard
script:
- name: dhcp
command: dhcp
successes:
- message: "DHCP client bound to address"
failures:
- message: "TIMEOUT"
exception: InfrastructureError
error: "dhcp failed"
- name: setenv
command: "setenv serverip {SERVER_IP}"
- name: wait for the prompt
command:
The name of the test suite.
The list of possible prompts for the interactive session. In many cases, there is just one prompt, but if shell has different prompts for different states, it can be accommodated. (Prompts can also include regexps, as any other match strings).
If set to discard, after each sent command of a script, discard
the next output line (assumed to be an echo of the command). This option
should be set when interacting with shell (like u-boot shell) that will echo
the command, to avoid false positive matches. Note that this options applies
to every command in the script. If you need different value of this
option for different commands, you would need to group them in different
script’s.
The list of commands to send and what kind of output to expect:
name: If present, log the result (pass/fail) of this command
under the given name (as a testcase). If not present, and the command
fails, the entire test will fail (with TestError).command: The command (string) to send to device, followed by newline.
The command can use variables that will be substituted with live data,
like {SERVER_IP}. If value is empty (command: in YAML), nothing
is sent, but output matching (prompts/successes/failures) will be
performed as usual. (Note that empty command: is different from empty
string command: "". In the latter case, just a newline will be sent
to device.)failures and successes: Each optional. If present, check the
device output for the given patterns.successes should be a list of dictionaries with just one key:
message: The string (or regexp) to match. Substring match is
performed, so care should be taken to reliably encode the match pattern.
(E.g. message: 4 would match “4” appearing anywhere in the output,
e.g. “14” or “41”).failures should be a list of dictionaries with:
message: The string (or regexp) to match. Substring match is performed.exception (optional): If the message indicates a fatal problem,
an exception can be raised, one of:
InfrastructureError,
JobError,
TestError. If not present, the error
is not fatal and will be recorded just as a failed testcase in test
results. (If this is a named command; as mentioned above, failure of
unnamed (“not a testcase”) command leads to implicit TestError).error: if defined, the exception message which will appear in the job logIf successes is defined, but LAVA matches one of the prompts
instead, an error will be recorded (following the logic that the lack
of expected success output is an error). This means that in many cases
you don’t need to specify failures - any output but the successes
will be recorded as an error.
However, if successes is not defined, then matching a prompt will
generate a passing result (this is useful for interactive commands
which don’t generate any output on success; of course, in this case
you would need to specify failures to catch them).
See also
Test jobs using Monitors must:
start string:end string:start and end strings will match part of a line but make sure
that each string is long enough that it can only match once per boot.
If start does not match, the job will timeout with no results.
If end does not match, the job will timeout but the results (of
the current boot) will already have been reported.
The name of the test suite.
- test:
monitors:
- name: tests
start: BOOTING ZEPHYR
end: PROJECT EXECUTION SUCCESSFUL
pattern: '(?P<test_case_id>\d+ *- [^-]+) (?P<measurement>\d+) tcs = [0-9]+ nsec'
fixupdict:
PASS: pass
FAIL: fail
If the device output is of the form:
***** BOOTING ZEPHYR OS v1.7.99 - BUILD: Apr 18 2018 10:00:55 *****
|-----------------------------------------------------------------------------|
| Latency Benchmark |
|-----------------------------------------------------------------------------|
| tcs = timer clock cycles: 1 tcs is 12 nsec |
|-----------------------------------------------------------------------------|
| 1 - Measure time to switch from ISR back to interrupted thread |
| switching time is 107 tcs = 1337 nsec |
|-----------------------------------------------------------------------------|
...
PROJECT EXECUTION SUCCESSFUL
The above regular expression can result in test case names like:
1_measure_time_to_switch_from_isr_back_to_interrupted_thread_switching_time_is
The raw data will be logged as:
test_case_id: 1 - Measure time to switch from ISR back to interrupted thread |
| switching time is
Caution
Notice how the regular expression has not closed the match at the end of the “line” but has continued on to the first non-matching character. The test case name then concatenates all whitespace and invalid characters to a single underscore. LAVA uses pexpect to perform output parsing. pexpect docs explain how to find line ending strings: https://pexpect.readthedocs.io/en/stable/overview.html#find-the-end-of-line-cr-lf-conventions
r'(?P<test_case_id>\d+ *- [^-]+) (?P<measurement>\d+) tcs = [0-9]+ nsec'
The test_case_id will be formed from the match of the expression \d+
*- [^-]+ followed by a single space - but only if the rest of the
expression matches as well.
The measurement will be taken from the match of the expression \d+
preceded by a single space and followed by the exact string tcs =
`` which itself must be followed by a number of digits, then a single
space and finally the **exact** string ``nsec - but only if the rest
of the expression also matches.
See also