When running several tests simply having the list of failed test names
is not very convenient. The actual error may be thousands of lines north
and worse, meson only prints the last 100 lines of a test log by default.
So let's print the full test data including backtrace etc. at the end
instead.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1189>
Require the type to be added in the litest_test_params_fetch() so we can
easily detect a mismatch. And add some type-safe getters that are much
easier to use for all the tests that only have a single parameter to
fetch anyway.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1139>
litest supports ranged tests but they are not enough, doubly so with
tests where we want to parametrize across multiple options.
This patch adds support for just that, in clunky C style.
The typical invocation for a test is by giving the test parameter
a name, a number of values and then the values themselves:
struct litest_parameters *params = litest_parameters_new("axis", 's', 2, "ABS_X", "ABS_Y",
"enabled", 'b', '2', true, false,
"number", 'u', '2', 10, 11,
NULL);
litest_add_parametrized(sometest, LITEST_ANY, LITEST_ANY, params);
litest_parameters_unref(params);
Currently supported are u (uint32), i (int32), d (double), b (bool),
c (char) and s (string).
In the test itself, the `test_env->params` variable is available and
retrieval of the parameters works like this:
const char *axis;
uint32_t number;
bool enabled;
litest_test_param_fetch(test_env->params,
"axis", &axis,
"enabled", &enabled,
"number", &number,
NULL);
Note that since this is an effectively internal test-suite only
functionality we don't do type-checking here, it's assumed that if you
write the code to pass parameters into a test you remember the type
of said params when you write the test code.
Because we don't have hashmaps or anything useful other than lists the
implementation is a bit clunky: we copy the parameter into the test
during litest_add_*, permutate it for our test list which gives us yet
another linked list C struct, and finally copy the actual value into
the test and test environment as it's executed. Not pretty, but it
works.
A few tests are switched as simple demonstration. The name of the
test has the parameters with their names and values appended now, e.g.:
"pointer:pointer_scroll_wheel_hires_send_only_lores:ms-surface-cover:axis:ABS_X"
"pointer:pointer_motion_relative_min_decel:mouse-roccat:direction:NW"
Filtering by parameters can be done via globs of their string
representation:
libinput-test-suite --filter-params="axis:ABS_*,enabled:true,number:10*"
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1109>
This replaces check. The code is a copy of pwtest which I wrote years
ago for pipewire but adjusted for us here the last few days.
There are a few advantages over check:
- Ability to SKIP tests or mark them as NOT_APPLICABLE, the latter
of which is used for early checks if a device doesn't meet
requirements.
- it captures stdout/stderr separately
- colors!
- YAML output format makes it a lot easier to read the results and
eventually parse them for e.g. "restart failed tests"
Less abstraction: we set up the tests, pass them to the runner and run
them with the given number of forks. This is an improvement over before
where we forked into N test suites which each called check which then
forked again. Since we're now keeping track of those processes
ourselves we can also write tests that are expected to fail with
signals.
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>