The HAL test suite
~~~~~~~~~~~~~~~~~~~
The tests in these directories serve to test the behavior of HAL components.
Each subdirectory of this directory may contain a test item. The runtests
script recurses through the directory structure, so multiple tests could
be structured as
tests/
xyz.0
xyz.1
xyz.2
or
tests/
xyz/
0
1
2
Two types of tests are supported: Regression tests, in which the output is
tested against a "known good" output, and functional tests, in which the
output is fed to a program that can determine whether it is correct or not
Running the tests
~~~~~~~~~~~~~~~~~
Currently, tests only work with the "run in place" configuration. They
can be run by executing (from the top emc2 directory)
scripts/runtests tests
A subset of the tests can also be run:
scripts/runtests tests/xyz tests/a*
The directories named on the commandline are searched recursively for
'test.hal' or 'test.sh' files, and a directory with such a file is
assumed to contain a regression test or a functional test.
Tests may contain files other than the ones specified below. For instance,
when using 'streamer' data as test input, a shell script with
"halstreamer<<EOF" and a "here document" will generally be present.
(see and-or-not-mux.0/runstreamer for an example)
Regression Tests
~~~~~~~~~~~~~~~~
A regression test should consist of these three files:
README
A human-readable file describing the test
test.hal *or* test.sh *or* test
The test script to execute. test.hal is executed with
'halrun -f', test.sh is executed with 'bash -x', and
test is executed as ./test
expected
A file whose contents are compared with the stdout of
halrun -f test.hal
A typical regression test will load several components, usually including
'threads' and 'sampler', and often including 'streamer', then collect samples
from some number of calls to the realtime thread, then exit.
Regression test "test.hal" files will almost always include the line
setexact_for_test_suite_only
which causes HAL to act as though the requested base_period was available.
Otherwise, results will differ slightly depending on the actual base_period
and regression tests will fail.
The test passes if the expected and actual output are identical.
Otherwise, the test fails.
Functional Tests
~~~~~~~~~~~~~~~~
A functional test should consist of three files:
README
A human-readable file describing the test
test.hal *or* test.sh *or* test
The test script to execute. test.hal is executed with
'halrun -f', test.sh is executed with 'bash -x', and
test is executed as ./test
checkresult
An executable file (such as a shell or python script)
which determines if the stdout of
halrun -f test.hal
indicates success or failure
Regression test "test.hal" files will often include the line
setexact_for_test_suite_only
which causes HAL to act as though the requested base_period was available.
Otherwise, results will differ slightly depending on the actual base_period,
which could affect whether 'checkresult' gives an accurate result.
A typical regression test will load several components, usually including
'threads' and 'sample', then collect samples from some number of calls
to the realtime thread, then exit. 'checkresult' will look at the output
and see if it indicates success.
The test passes if the command "checkresult actual" returns a shell
success value (exit code 0). Otherwise, the test fails.
