test: automate test suite handling

Move all tests to a special section, then loop through that section
to fetch all test suite. The result is that new tests only need to add the
source files without having to update everything else as well.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

test: rename some tests for less ambiguity

"key" usually refers to one of KEY_ of EV_KEY in general, but here we're
testing event codes.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Change all URLs to gitlab.fdo

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Update struct input_event

The struct input_event is not y2038 safe.
Update the struct according to the kernel patch:
https://lkml.org/lkml/2018/1/6/324

Signed-off-by: Deepa Dinamani <deepa.kernel@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: sync event codes with kernel 4.16

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.5.9

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

test: skip tests when we're not root

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Blacklist SW_MAX so it doesn't shadow SW_PEN_INSERTED

They have the same value, so the _MAX code would shadow the real code, causing
issues in any client that needs to get all event names from libevdev.
Specifically, the loop of:
  for each code in 0 to max-for-type:
      print(name)
would not show up the code (but the _MAX) code instead. This causes issues
with clients that rely on name resolution that works. And the _MAX values are
special values anyway.

Blacklist it in the script here, causing it to resolve from name to code, but
not from code to name (like other duplicated codes).

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Blacklist REP_MAX so it doesn't shadow REP_PERIOD

They have the same value, so the _MAX code would shadow the real code, causing
issues in any client that needs to get all event names from libevdev.
Specifically, the loop of:
  for each code in 0 to max-for-type:
      print(name)
would not show up the code (but the _MAX) code instead. This causes issues
with clients that rely on name resolution that works. And the _MAX values are
special values anyway.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Drop the python map printing

leftover from when this was part of evemu

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

libevdev 1.5.8

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Sync with kernel 4.15 (BTN_STYLUS3)

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

test: add --no-install flag to be able to debug directly

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

When changing the fd, reset our grab state to ungrabbed

Previously, calling grabbing a device after changing the fd was a no-op
because libevdev's grab state didn't match the fd:

libevdev_grab(LIBEVDEV_GRAB);
  .. fd is grabbed
  .. internal state is 'grabbed'
libevdev_change_fd();
  .. new fd is ungrabbed
  .. internal state is 'grabbed'
libevdev_grab(LIBEVDEV_GRAB);
  .. argument matches internal state and we exit without grabbing the device

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: sync with kernel 4.13

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: add a script to sync the kernel header files

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: sync with kernel v4.12

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

uinput: make a note that the syspath we return is the input node

And not the syspath for the /dev/input/eventX node

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

test: remove test for ULONG_MAX queue allocation

Fixes the warning:
../libevdev/libevdev-int.h:231:15: warning: argument 1 value
‘18446744073709551615’ exceeds maximum object size 9223372036854775807
[-Walloc-size-larger-than=]

That's now part of gcc's -Wall, so let's rely on that for code. Arguably, the
queue code is simple enough that we don't need a test for ENOMEM anyway.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

test: drop unused parameters from test's main()

Not needed, so let's get rid of the compiler warning.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

test: disable coredumps during test suite runs

It's a test suite, it shouldn't fill up the file system or the journal with
coredumps.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

libevdev 1.5.7

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Fix an indentation issues

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

doc: add links to git repositories on the main page

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

autogen: add default patch prefix

Signed-off-by: Mihail Konev <k.mvc@ya.ru>

autogen.sh: use exec for configure

No point in waiting around until it finishes

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

autogen.sh: escape $srcdir before configure

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: print an error if we don't have any matching events in the dpi tool

Beats crashing by dereferencing a null-pointer (when we access
m->frequencies[idx])

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

tools: move udev printf into the print_summary() helper

No functional changes

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

configure.ac: enable subdir-objects

The main thing holding us back here was our gcov hacks. We used to rebuild the
libevdev sources locally inside test/ with the gcov flags so that we could
leave the main libevdev sources untouched. This doesn't work well with
subdir-objects - we have to link to libevdev.la instead.

To enable gcov, we now have to apply the gcov flags to the main library
object. But this also means that when running, the notes files will be
somewhere within the libevdev/ directory, not the test/ directory. Working
around this in automake gets nasty quickly, so just add a script that knows
how to search for things.

No functional changes unless --enable-gcov is given at configure time - then
don't install the library.

The gcov reports are now in test/gcov-reports/

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

test: fix a Makefile.am comment

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.5.6

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Fix typos in Documentation

Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: update to v4.9 linux/input.h headers

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.5.5

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Disable test runs on make distcheck

The tests all need root, but running distcheck as root is not ideal. Disable
the test runs (but not the build) to make it easier to verify distcheck works
as intended.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Fix minor error in doxygen example code

The incorrect variable was checked as a return code in an example, and
that can be confusing for people trying to learn or use the example code
as a starting point.

tools: print the mean frequency together with the max frequency

And if they're 30% out, print a warning. On the ThinkPad X1 Wireless Touch
Mouse (when connected via bluetooth) we get a bunch of events at the start of
the movement, all less than 1ms apart. Best guess is that the device goes to
low-power, then notices the movement and buffers the event until the BT
connection is back up. Then it sends all events at once. Usually they're less
than 1ms apart, but at one recording showed a 37ms delay before we go back to
the normal 70ms (~15Hz) the mouse has otherwise.

This is unpredictable enough that we can't just work around it so instead
print a warning to the user so they can go investigate.

https://bugs.freedesktop.org/show_bug.cgi?id=97812

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

tools: rename frequency to max_frequency in the dpi tool

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: use uint64_t, not doubles for the µs parameters

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: fix kernel-announced width/height

Side-effect of 240ba34ebd483 was that "touchpad size as listed by the kernel"
was now dependent on the values we got. This one is a static one based on the
axis info.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

doc: drop the HTML_TIMESTAMP

Allows for reproducible builds. Debian carries a patch for this, and the
timestamp doesn't really add much since the doc is either in sync with master
or represents the release tag. Might as well drop it.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.5.4

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: fix the touchpad resolution calculation

Previous ones used the absinfo from the kernel but since we never updated that
from within the tool, the output was always the same.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.5.3

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Don't bother sanitizing disabled event codes

Filter them immediately instead of passing them on and relying on the actual
event handling code to filter them.

Reproducer: if EV_ABS is disabled on an Apple MagicMouse we still get events
passed into sanitize_event(). But the code handling EV_ABS events doesn't
update the state, so we end up complaining about double tracking IDs, even
though that is not actually correct.

https://bugzilla.redhat.com/show_bug.cgi?id=1361325

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

tools: require a minimum size for touchpads

This mostly aims to catch users trying to specify the size in inches.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: change touchpad-edge-detector to require physical size

Almost no-one does the calculations for me to update the udev rules (and some
rules were submitted with the <x resolution> placeholders left in).
Require the user to specify the physical size so we just copy/paste the actual
udev rule.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

libevdev 1.5.2

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: update to v4.6 linux/input.h header

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.5.1

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev: Properly distribute uinput.h

Signed-off-by: Armin K <krejzi@email.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.5

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: allow numeric axis values in tweak-device

Not all axes a device may have are assigned a semantic name.

https://bugs.freedesktop.org/show_bug.cgi?id=95029

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: fix tweak-device option parsing

https://bugs.freedesktop.org/show_bug.cgi?id=92880

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

doc: reduce doxygen file to the non-default settings

Makes it easier to check what we changed from the defaults.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Fix typo in doc string

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

uinput: support the new UI_DEV_SETUP/UI_ABS_SETUP ioctls

Available in kernel 4.5 and later (uinput version 5), these ioctls allow us to
set the absinfo correctly, i.e. including the resolution.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

uinput: split uinput device creation into a helper function

No functional changes. This is prep work for supporting the new
UIDEV_DEV_SETUP ioctl.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

uinput: drop an unused argument

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

include: update to v4.5 linux/uinput.h

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Add parsing linux/input-event-codes.h, update to kernel v4.5

The kernel has split most of the event codes out to a new header but the FF_
bits are missing from that header. Until this is fixed upstream, change the
event code parsing so it can take two files (using cat and stdin) so we can
update the kernel headers again.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

test: ignore a failure to UI_DEV_DESTROY the device

Coverity complains about it.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

test: drop unnecessary return value

We abort if anything goes wrong anyway, so we never returned anything but
success.

Found by coverity.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

libevdev 1.4.6

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

uinput: fix race condition in uinput syspath check

In theory, the device could change between stat() call and open(), resulting
in us opening the new device. Change to open() first, then fstat() on the fd.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>

tools: shut up coverity about a potential close(-1)

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: fix coverty "may be used uninitialized" warnings

tools/libevdev-tweak-device.c:390: uninit_use_in_call: Using uninitialized
value "changes" when calling "parse_options_abs".

tools/libevdev-tweak-device.c:376: warning: 'led' may be used uninitialized in
this function

tools/libevdev-tweak-device.c:375: warning: 'axis' may be used uninitialized
in this function

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Document that the fd should be drained before libevdev_set_fd

This is the caller's responsibility, for two reasons:
* we don't know if O_NONBLOCK is set, so draining the fd isn't a simple matter
  of read() until EAGAIN. A select() + read() could work around this of
  course.
* for stateless information, keys and relative data, it is not a problem when
  there are events waiting on the fd already, they are processed correctly,
  albeit with a delay.

So punt this decision to the caller, they openend the fd, they know if they
care about delayed events, they can drain the fd before handing it to us.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Acked-by: David Herrmann <dh.herrmann@gmail.com>

libevdev 1.4.5

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: fix touchpad-edge-detector udev rule output

off-by-one error on the ABS_ axes printed. ABS_X/Y are 0 and 1.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Add some more files to gitignore

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.4.4

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Fix invalid absinfo range values reported by certain mtk soc

This change will only affect certain touch screens, for which the driver
integration code does not provide meaningful values for the allowed range
of ABS_MT_TRACKING_IDs. The reported range [0, 0] will be overwritten with
[-1, 0xFFFF]

Signed-off-by: Andreas Pokorny <andreas.pokorny@canonical.com>
[Changed from INT_MAX to 0xFFFF to match the kernel, add device name to log
message]

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: print out evdev override udev rule from touchpad edge detector

We're starting to collect overrides for custom devices, making this easier for
users and saves us time. Once we measured everything, print out a
guesstimated udev rule and instructions on how to calculate the resolution.
Extra output now is:

Touchpad size as listed by the kernel: 132x111mm
Calculate resolution as:
x axis: 6076/<width in mm>
y axis: 5021/<height in mm>

Suggested udev rule:
# <Laptop model description goes here>
evdev:input:b0005v05ACp030E*
 EVDEV_ABS_01=-2694:2862:<x resolution>
 EVDEV_ABS_02=-20:121:<y resolution>
 EVDEV_ABS_35=-2694:2862:<x resolution>
 EVDEV_ABS_36=-20:121:<y resolution>

The ABS_MT_ axes are only printed if the device have them, if the device isn't
bluetooth/usb we print the dmi modalias instead.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools - tweak-device: add a --resolution command

So far, 100% of the usages for tweak-tool was to set the x/y resolution of a
device. Make --resolution a shortcut for this.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

tools - tweak-device: revamp to reduce use of globals

Make the code base a bit more modular so it's easier to add new commands.
Main change here is: options are parsed twice now, first time for the mode
(abs/led) and the device path, then again for the mode-specific options.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Note that libevdev_free() does not close the fd

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools - tweak-device: fix wrong error message if the LED doesn't exist

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools - tweak-device: close the fd on exit

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.4.3

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: update to 4.1 header

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: complain about devices that don't have abs x/y axes in the edge detector

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: widen frequency resolution to µs in the DPI tool

The Microsoft Arc Touch Mouse claims 8000fps which is higher than we can
measure in the current milliseconds resolution.
http://www.cnet.com/products/microsoft-arc-touch-mouse-black-series/specs/

https://bugs.freedesktop.org/show_bug.cgi?id=90540

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.4.2

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: fix help string for manual DPI calculation

https://bugs.freedesktop.org/show_bug.cgi?id=90143

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: update to 4.0 header

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.4.1

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Accept LIBEVDEV_READ_FLAG_BLOCKING as valid flag

https://bugs.freedesktop.org/show_bug.cgi?id=89798

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

ABS_MAX counts as MT axis for the event queue

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

cosmetic: fix a comment

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Disable test-static-link in case of shared only build.

Fixes the following buildroot compile failure:

libtool: link: [..]/host/usr/bin/arm-buildroot-linux-uclibcgnueabi-gcc -std=gnu99 -I.. -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Os -static -o test-static-link test_static_link-test-link.o  ../libevdev/.libs/libevdev.so -Wl,-rpath -Wl,[...]/build/libevdev-1.4/libevdev/.libs
[...]/arm-buildroot-linux-uclibcgnueabi/bin/ld: attempted static link of dynamic object `../libevdev/.libs/libevdev.so'
collect2: error: ld returned 1 exit status
Makefile:719: recipe for target 'test-static-link' failed

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev 1.4

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

cosmetic: fix a couple of duplicate/missing empty lines

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

test: use -std=c89 instead of c90 for the build test

They're identical, but RHEL6 doesn't support -std=c90

https://gcc.gnu.org/onlinedocs/gcc/Standards.html
https://gcc.gnu.org/onlinedocs/gcc/C-Dialect-Options.html#C-Dialect-Options

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

tools: add a tool to change kernel devices

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

test: always build the build-tests

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

configure: rename BUILD_TESTS to ENABLE_RUNTIME_TESTS

Avoids confusion when we build tests in the future even though BUILD_TESTS is
disabled.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Use memcpy/memmove instead of loop operations

Signed-off-by: Thilo Schulz <thilo@tjps.eu>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

libevdev_uinput_destroy: don't close non-open FD

The returned errno from libevdev_input_create_from_device was returned
incorrectly because libevdev_uinput_destroy() would try to close the
unset value of ->fd, overwriting errno.

That was fixed in debe9b030c8069cdf78307888ef3b65830b25122, this patch avoids
the ioctl/close calls if the fd isn't set.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>