test: filter out duplicate name/device combos in --list

This doesn't have an effect yet because ranged tests are handled inside
check. But in the future these tests will be split up individually so
de-duplicating is useful here.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>

test: make litest_backtrace() available from outside litest.c

Prep-work for adding a new runner

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>

test: append, not insert the suites

list_append() came later than list_insert() and there's an argument to
be made that tests added later should be run first since they're less
likely to succeed. But it's a lot harder to read test logs when they are
in reverse order, and with the TEST_COLLECTION() macro the order of the
test suite is not obvious anyway.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>

util: add a bunch more time conversions

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>

util: move libinput_now() into a utility function

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>

util: add xclose(fd) to close-and-reset an fd

This prevents accidentally leaving the fd set after closing.

And it includes the -1 check so we don't need this everywhere ourselves
(not that we use it right now but valgrind likes to complain about
this).

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1067>

pre-commit: drop black, use ruff-format instead

And switch to the current-ish version of the ruff pre-commit hook,
including updating the repo.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1072>

tools: change direct type check to isinstance

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1072>

util: make a float to int conversion explicit

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1071>

util: add the backtrace printing function as separate util

Might as well make this easier to re-use since it doesn't do anything
specific to litest.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1066>

pad: default a pad group to mode 0 if all LEDs are off

Since our API doesn't accomodate for "dunno" we need to pick a mode that
we are actually in. This happens on the Intuos Pro 2 (PTH-660) which has
all LEDs on brightness zero, resulting in a failure to set up the modes
and we're left without a mode button.

Fix it by just picking zero as the default mode until specified
otherwise.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1069>

pad: move a function to save on an ifdef

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1069>

pad: don't print a clobbered errno

is_litest_device() may change errno since it calls into libudev.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1069>

tools: print full pad mode group info in libinput-list-devices

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1068>

tools: add rel dial support to libinput-list-devices

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1068>

test: split the touchpad tests into a palm and a dwt test collection

dwt and palm tests have a lot of timeouts to wait for so let's split
those out.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

test: split the tablet left-handed tests out into a separate collection

These aren't complicated but there's a lot of them so let's run them
separately to make the overall tablet test shorter.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

CI: move the "device" test suite to the "misc" set of suites

That one is still with the gestures but only takes 1-2 minutes, the
gestures takes 13 minutes. Let's move it to misc which currently has an
overall runtime of only 2 min anyway.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

test: split out the touchpad tap tests into multiple collections

These take a long time and have a reasonable high chance of failure due
to the timing constraints. Let's split them up so they don't hog the
runners for that long and in case they fail, we only need to re-run a
short test.

Before: one test running approx 21 min, now 3 tests running approx 7 +
11 + 4 min.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

meson: break out the test suites into manual list

If we start having multiple TEST_COLLECTION() in the same file we can
no longer use the file name - so a manually maintained list it is.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

test: make the TEST_COLLECTION() macro re-usable in the same file

Concat the line number to the generated variable names, this way we can
have more than one TEST_COLLECTION() in the same file.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

test: detach the suite handling from the file names

Instead of extracting the suite name from the test's file name use the
current suite that is being parsed. This way we pave the way for
multiple suites in the same file.

This uses a global because otherwise we'd have to redo all the
litest_add() functions but it does the job here.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

test: match the tablet's pad test collection with the file name

This is currently a requirement, so let's match this before we run
into issues here re-structuring stuff.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1065>

tablet: fix a memory leak caused by an early return

Previously we'd return early, failing to libwacom_destroy() the device.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1064>

CI: set the LITEST_JOBS in the template, not the script

Fixes: 70c57e964450 ("CI: drop the job count for the valgrind test suite to 2")
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1063>

test: use a litest helper instead of return code comparison

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: rename a variable to indicate suites, not tests

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: rename all_tests to all_test_suites

This is a a list of struct suite with the various tests inside that
suite.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: check for verbosity immediately in main()

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: check for empty tests immediately, not later when we're running

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: init the global devices/test lists immediately

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: add restore_tty helper and use it

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: init the device list via argument

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: add a helper for deleting a created file

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: switch a bunch of tests from abort() to one with a message

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: fix litest_abort_msg() not starting a new line

Most callers of litest_abort_msg() don't add '\n' so the output was
mangled.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

doc: correct that the test suite is indeed run in the CI

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1062>

test: switch the tests to use enum comparisons

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: add macros to compare enum values

This requires switching a lot of int_eq/int_ne over to enum_eq/enum_ne
because the compiler doesn't infer the right type from a harcoded enum
value - it just defaults to int.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: switch a bunch of integer comparisons to right sizes

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: switch ck_assert over to litest_assert

This is the first step in switching away from the check framework.
Our litest macros already do almost exactly the same anyway so most of
this is a simple sed with a few compiler fixes where things mismatch
(nonnull -> notnull) and (_tol -> _epsilon).

This now generates a whole bunch of integer mismatch warnings: check
casts everything to intmax_t whereas we use typeof, so lots of warnings
especially for enums.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: add litest string comparison macros

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: replace manual "almost equal" with the new double epsilon

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: fix some checks using ints to compare doubles

Theoretically we should be using ck_assert_double_eq here for
consistency but this patch  is part of a series eventually
replacing those calls, so let's jump to litest_assert_double
directly to avoid further rebase conflicts.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: fix a touch test comparing ints to doubles

We were checking doubles for integers but better to check that we're
close to the maximum range without actually being over.

This worked because check typecasts to uint_max_t but let's be explicit
here.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: add a macro to compare two doubles with a specific epsilon

Same as ck_assert_double_tol

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: initialize a variable to make scan-build happy

Assuming safe_atoi works as expected, `fuzz` cannot be
uninitialized by the time we get here. But let's init it anyway to make
scan-build happy.

[202/249] Compiling C object libinput-test-suite.p/test_test-touch.c.o
../../../test/test-touch.c:964:2: warning: Assigned value is garbage or undefined [core.uninitialized.Assign]
  964 |         litest_assert_int_eq(fuzz, 10); /* device-specific */
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Note that this error message is the result of a follow-up commit,
this commit is shuffled before so we have bisectable build.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: add litest_assert_event_type_not_one_of

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: replace ck_abort with litest_abort_msg()

Weirdly, that also required initializing two variables to NULL to stop a
compiler warning.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: fix missing END_TEST for strv_for_each_test

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: fix double parsing test

We should check that we actually have a double here...

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: fix a data type in the utils test

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: add a missing va_end

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1059>

test: fix uinput creation for the slotted devices with too many slots

Kernel commit 206f533a0a7c
       "Input: uinput - reject requests with unreasonable number of slots"
limits the number of slots to 99 - let's manually adjust that so we can
keep creating uinput devices.

Since these are just a test device and we don't use the slots here
anyway (they're all fake MT devices) we can manually work around this.
The real devices won't be affected by this since this is a limitation
in uinput, not the input subsystem.

Also move the comment one line up in the ms-surface device, the previous
comment referred to the wrong event code.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1061>

test: make litest-test-suite --list YAML-compatible

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1060>

test: move the test devices into the test_device_section

Better naming this way

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1060>

test: switch unused argc/argv to the list of tests

This means we don't rely on the "all_tests" global here though it also
means we need to move the cleanup into the caller.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1060>

quirks: add Logitech ERGO M575(S)

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1058>

gestures: change a boolean to an enum for more expressive code

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: add debugging for state transitions

Our state machine allows falling through multiple transitions, so let's
make it possible to debug this.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: get rid of tp_gestures_start

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: rename gesture_notify_hold to gesture_notify_hold_begin()

HOLD gestures only have begin/end so let's rename this so it's
immediately obvious which one we're sending.

notify_swipe/pinch pass the event type in so it's clear that it's a
BEGIN vs an UPDATE.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: rename most events to FOO_START

These events aren't used to signal scroll/swipe/pinch/..., merely to
signal the start of that gesture. So let's rename it to make the code
clearer (e.g. why do we log a bug when for a FOO event when we're in
state FOO?).

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: push tp_gesture_start() for hold gestures into the state handling

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: push tp_gesture_end() into the actual state machine

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: log a bug if we're trying to reset an actual gesture

Now that end/cancel are separate states, we shouldn't get here.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: distinguish between a gesture reset and a gesture end/cancel

All current states treat them the same way but let's distinguish those
two so we can have different code depending on whether a gesture was
cancelled/reset or ended properly.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: replace switch-timeout based gesture_cancel calls with a state

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: rename tp_gesture_handle_state and tp_gesture_post_gesture

Confusingly, tp_gesture_handle_state() would do almost nothing with our
state machine and the various tp_gesture_handle_state_foo() were called
later from tp_gesture_post_gesture().

Rename those functions into so that we have
tp_gesture_update_finger_state() first followed by
tp_gesture_handle_state() which is responsible for dealing with the
state machine.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: make a error message more informative

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

gestures: log the finger count together with the gesture state

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1049>

quirks: add a quirk for the HONOR MagicBook Art 14 touchpad

The touchpad need ModelPressurePad=1 to enable moving cursor,
and need AttrEventCode=-BTN_RIGHT to use right key by touchpad
two-fingers press..

Closes: #1040
Link: https://gitlab.freedesktop.org/libinput/libinput/-/issues/1040
Signed-off-by: Wentao Guan <guanwentao@uniontech.com>
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1053>

quirks: Add quirk for Google Chromebook Cret

Signed-off-by: Brady Norander <bradyn127@protonmail.com>
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1056>

quirks: touchpad quirk for ICL Si1516/Si1512

Exactly the same quirk as for Graviton N15i-k2 and Graviton N15i.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1054>

quirks: touchpad quirk for Graviton N15i

Exactly the same quirk as for Graviton N15i-K2.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1054>

CI: skip the valgrind tests for marge-bot

marge-bot rebases and edits the commit message but there's not way for
it to introduce a memleak that wasn't spotted in the user pipeline
first. Since those tests are very flaky, let's skip them when running
the marge-bot pipeline.

Closes #1042

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1057>

Fix a few potential NULL dereferences

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1052>

util: add a safe version of strlen

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1052>

quirks: adjust pressure range for Google Krane, Wormdingler and Coachz

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1055>

tools: add --set-calibration to the debug-events man page

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1051>

test: fix macro grouping in the litest.h header

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1050>

test: use litest_assert_event_type instead of direct type check

This provides better debugging logs and is slightly less code in
the checks too.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1050>

test: use litest_wait_for_event instead of a NONE event check

Technically we're not really waiting here since we expect the
event to already be there but for these tests the distinction doesn't
matter.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1050>

test: fix use of wrong enum for setting clickfinger maps

Hidden by ck_asset_int_eq type-casing everything to intmax_t

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1050>

test: use litest_assert_empty_queue instead of a manual none check

This provides better debugging printfs and is the now-recommended way to
check this.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1050>

test: add litest_dispatch() for better test debugging

Wraps libinput_dispatch() with a location which will make things a bit
easier to track. Output (in --verbose) is something like:

  gestures_swipe_3fg_unaccel_fn():1346 - dispatching

Which makes it easier to associate the various calls to libinput
dispatch with the other output from libinput.

This patch switches all uses of libinput_dispatch() in test cases over
but not the litest functions that may call dispatch too. Remains to be
seen if that is necessary.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1048>

test: abort on libinput bugs again

This was always intended but a bug prevented the actual abort.
strstr returns NULL when we cannot find the substring so we always
triggered the first noop condition on bugs.

Fixes: bd7b91065b13 ("evdev: warn if our event processing lags by 10ms or more")
Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1048>

test: improve event debugging a bit

Add a litest_checkpoint macro and convert a few of the litest_assert
macros to make use of that - this gives us a printf of the call site in
case it fails.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1048>

test: highlight the gesture state machine log messages in yellow

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1048>

gestures: use a colon-suffixed debug prefix like the other state machines

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1048>

test: make litest_wait_for_event_of_type auto-terminating

Wrap this in a macro so we cannot forget to pass -1

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1048>

test: make litest_drain_events_of_type auto-terminating

Wrap this in a macro so we cannot forget to pass -1

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1048>

triage-policies: add reminder for info in commit messages

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1043>

triage-policies: add a blurb for the help-needed label

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1043>

tools/debug-events: move the event type str to a helper function

This isn't really necessary for the code as-is it makes debugging things
easier.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1044>

gestures: cancel the hold before initializing the pinch

If we're in state HOLD we have an ongoing hold gesture - make sure we
cancel that one first before we initialize the pinch.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1046>

triage-policies: point to the new hid-recorder written in rust

It provides a more useful and easier to read report output, so let's go
with that.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1045>

CI: retry valgrind jobs if they fail

Our valgrind jobs are very timing-sensitive so it's very common that
they fail with an error when the reason is just valgrind being slower
and we miss a deadline somewhere.

Retry them if they fail, hopefully that gives us more reliable
pipelines.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1040>

tools/debug-events: add ability to compress motion events

If --compress-motion-events is given (and stdout is a tty) reduce
the output printed to one line per repeated motion/axis/scroll sequence
(with a count). Example output:
 event6   POINTER_MOTION          108 +1.912s  1.00/ -1.00 ( +1.00/ -1.00))
 event6   POINTER_BUTTON              +2.008s BTN_LEFT (272) pressed, seat count: 1
 event6   POINTER_BUTTON              +2.074s BTN_LEFT (272) released, seat count: 0
 event6   POINTER_MOTION           39 +5.249s  0.00/  0.99 ( +0.00/ +1.00)
 event6   POINTER_BUTTON              +5.385s BTN_LEFT (272) pressed, seat count: 1
 event6   POINTER_MOTION           66 +6.031s -1.00/  0.00 ( -1.00/ +0.00)
 event6   POINTER_BUTTON              +6.401s BTN_LEFT (272) released, seat count: 0

The event count (108, 39 and 66) is only printed for more than one event
in sequence so the output is otherwise identical (but 4 spaces wider
now)

If stdout is not a tty the event count is printed but no compression
happens since we rely on a ansi escape sequence for that. Could be fixed
by changing the current print statements to print a \n before the
current event instead of at the end of the current line.

This makes debugging events easier as button events and similar are no
longer obscured by pages of motion events in between.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1041>

tools/debug-events: ignore old-style pointer axis events completely

Since they're interleaved with the more modern finger/wheel/continuous
events they mess up our repeate count, preventing compression for those
events.

Part-of: <https://gitlab.freedesktop.org/libinput/libinput/-/merge_requests/1041>