We already had a limitation on synching a max of 256 slots, regardless
what the device reports, this just extends the check to refuse any
device with >256 slots. The kernel already limits to 99 slots for uinput
devices so anything 256 and higher will be a malicious/buggy real
device.
We can't add a test to this because the kernel refuses uinput devices
with more than 99 slots.
Part-of: <https://gitlab.freedesktop.org/libevdev/libevdev/-/merge_requests/135>
If the start and end of a touch are dropped, the slot, according to the
kernel, may have a different state. We should inform the client of these
changes even if the slot is not currently active.
For most axes this doesn't matter too much as we expect them to change
during an active touch anyway so we don't expect the kernel's caching to
be a problem. However where the ABS_MT_TOOL_TYPE changed during a sync
we need to inform the client of the new tool type so that future
touchese won't be erroneously treated as e.g. palms.
For a full reproducer see the test case but it comes down to:
- touch down with MT_TOOL_PALM, make sure libevdev reads the state
- change that slot to MT_TOOL_FINGER, trigger a sync
- ensure that libevdev pushes out that tool type change even if the
slot is not currently active
Co-authored-by: Peter Hutterer <peter.hutterer@who-t.net>
Part-of: <https://gitlab.freedesktop.org/libevdev/libevdev/-/merge_requests/124>
If libevdev is used as subproject header lookup for libevdev.h fails
because our build directory isn't correctly set as one of the include
directories.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Not much of an effect but where libevdev is used as subproject those
files are now correctly accessible.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
A bit of a niche case but this helps with embedding libevdev as static
library when the tools don't matter.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Historically, I copied the docs on the server directly after pushing latest,
but (especially with meson) it's a lot easier to just build that specific tag
locally and rsync it to the correct target directory.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
On a test run it only prints the summary and one message per failed
test. While having this behaviour by default is nice it might be
interesting in some case to have more information print. Make the print
mode configurable from the environment variable CK_VERBOSITY, which can
have the values "silent", "minimal", "normal" or "verbose" so we can have
different outputs.
$ sudo CK_VERBOSITY=verbose ./test-libevdev
Running suite(s): libevdev init tests
libevdev_has_event tests
libevdev event tests
100%: Checks: 79, Failures: 0, Errors: 0
test-libevdev-init.c:23:P:test_new_device:test_new_device:0: Passed
test-libevdev-init.c:28:P:test_free_device:test_free_device:0: Passed
[ ... ]
Note that the default print mode doesn't change after this patch.
Signed-off-by: Enric Balletbo i Serra <eballetbo@redhat.com>
Otherwise, the distribution tarballs will be generated based on what's
in libevdev_la_SOURCES, which only includes the headers for the
operating system the maintainer happens to be running "make dist"
from. As a result of this, e.g. the 1.13.0 tarball only includes the
headers for Linux, making the tarball unbuildable for FreeBSD.
"meson dist" was already doing the right thing.
Signed-off-by: Alyssa Ross <hi@alyssa.is>
This is copied from libinput's CI but as one large change rather than
cherry-picking the process on how to get here. meson-build.sh is synched
with libinput's version - it is a more generic version anyway.
With this change we no longer require separate images for the qemu runs,
our default image is qemu-capable and can be run in qemu via
boot2container (b2c).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Only use the LIBEVDEV_SKIP_ROOT_TESTS env var in autotools where we need
it, in meson we can use meson to control which tests we (don't) want to
run.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Unlike autotools distcheck which ensures we didn't forget to add
anything to the makefiles, ninja dist just zips up the git repo.
It does run the tests though but without suite selection which is a
problem for us here.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Use the new needs-uinput suite specifier for the meson build job, and
use --no-tests for ninja dist in the autotools build job.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
In the no-check:meson job, the ninja arg was "dist" so the test would be
run as part of that anyway (and skipped, since we didn't have check).
In the no-doxygen-check:meson job, the ninja arg was zero so the test
would be skipped but since we don't have check we might as well just
run it as empty test suite.
And the same applies to the scan-build job, running the test shouldn't
hurt here.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We have meson with a proper test suite setup and this currently breaks
the CI due to some Arch issues with glibc debuginfo packages. Let's just
drop this so we don't run valgrind unconditionally.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Required by the debian sid containers, otherwise we fail because of a
missing /etc/apt/sources.list file.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Fix invalid bool usage which violates official meson specification and thus
breaks muon, an implementation of meson written in C.
Signed-off-by: illiliti <illiliti@protonmail.com>
While I was running the tests for libevdev on a
system, I had a test failure, and it told me to report it to
bugs.freedesktop.org.
This project uses GitLab now, so update the URL.
Signed-off-by: Douglas R. Reno <renodr@linuxfromscratch.org>
This is a noop, meson gets its compile dependencies from the compiler,
listing it here makes no difference.
Same is true for listing the other two linux/input/*.h headers as
sources.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The latter is deprecated, so let's bump the meson version requirement
and use the newer, shiny feature.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
With the 64 bit timestamps, the struct layout changes into a flatter
version, so let's use the input_event_(u)sec helpers to transparently
handle this.
Fixes#25
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
No big point building for Centos 7 anywmore, and Centos 8 is now Centos
Stream only which needs fixing in the CI templates first.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
A non-installed tool to make it easy to check if newly added codes are
indeed supported correctly.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The example code for creating uinput devices produces the following
warnings:
warning: unused variable ‘ev’ [-Wunused-variable]
| struct input_event ev[2];
| ^~
warning: unused variable ‘new_fd’ [-Wunused-variable]
| int fd, new_fd, uifd;
| ^~~~~~
warning: ‘err’ may be used uninitialized in this function [-Wmaybe-uninitialized]
| return err;
| ^~~
Signed-off-by: José Expósito <jose.exposito89@gmail.com>
Generate the snippet for whichever is the last version in the list for the
want_qemu tag.
And move the want_qemu tag up so it's more obvious in the config file.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The order of dict values is not deterministic in python leading to differing
header file generation which results in differing build output for the same
configuration. Sort to remove this inconsistency and make the output
reproducible.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Reviewed-by: Filipe Laíns <lains@archlinux.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Yes, the value we return is from the currently active slot, but there are a
few niche cases where the active slot changes from what the client may think
it is. So let's call it undefined like the other half of the documentation
already does.
Fixes#20
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This requires latest CI templates for the mkosi changes. Since the start_vm.sh
script is now gone, switch to using vmctl instead.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We still require Fedora for the various jobs with custom autotools/meson
configurations. But we might as well make it dependent on the config file
entries only than hardcoding it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
On some devices, a kernel input property has been set in error and we need the
ability to disable that property.
Signed-off-by: Scott Jann <sjann@knight-rider.org>
Requires some .gitignore pattern removal too and an autotools fix to actually
dist the man pages.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Remove the includsion of sys/signalfd.h again, it hasn't been needed
since cca90938 and was accidentally re-added, probably as a mismerge,
in a40e014e.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
And fix the script to sync the headers up so it syncs event codes for both bsd
and linux, but only syncs input.h for linux.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Add two FreeBSD specific tests:
test_uinput_check_devnode_bsd checks that libevdev_uinput_get_devnode()
returns something sensible. This is modelled on the Linux test
test_uinput_check_syspath_name, but uses devnode instead of syspath, since
reeBSD doesn't have sysfs.
test_uinput_check_syspath_bsd checks that libevdev_uinput_get_syspath()
always returns NULL.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
FreeBSD does not support force feedack events. Disable the test for
this event when running on FreeBSD.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
FreeBSD does not have CLOCK_MONOTONIC_RAW, instead use
CLOCK_MONOTONIC_FAST. This test checks that libevdev_set_clock_id()
fails when called with CLOCK_MONOTONIC_[RAW,FAST].
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
Implement a FreeBSD version of fetch_syspath_and_devnode().
FreeBSD does not have sysfs, so instead fetch the device node directly
as as this matches with what is returned by the UI_GET_SYSNAME ioctl().
Since there is no sysfs, libevdev_uinput.syspath will always be set to NULL.
If the ioctl fail, return -1 from fetch_syspath_and_devnode(), since
there is no other way to figure out the device node path.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
Move the definition of SYS_INPUT_DIR to where it is used, instead of at
the top of the file, to make it easier to find.
Undefine it at the end of usage to avoid accidental uses.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
Use baename(argv[0]) to get the program name (for usage), instead of
using program_invocation_short_name, which only exists on Linux, not
FreeBSD.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
Remove signalfd() use from the mouse-dpi-tool and touchpad-edge-detector
tools, in favor of using plain old signals.
FreeBSD does not have signalfd() without pulling in external libraries,
and with this change these tools can be compiled on FreeBSD.
Instead of providing two implementations, one using signalfd() and one
using signal(), just use the signal() implementation everywhere as it is
more portable.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
Add FreeBSD compatible input.h and uinput.h files.
This is done by moving the linux files to include/linux/linux, adding
the freebsd versions in include/linux/freebsd, and then changing
include/linux/[u]input.h to pull in the right one depending on which OS
we are compiling on.
Make sure that the build infrastructure in meson.build and
autoconf.ac/Makefile.am uses the correct files when building and as
dependency for targets, and ensure that make-event-names.py get the
correct files as arguments.
A similar change has been done in libinput in
61f3e3854458c556a01fb05d7abb22733fd2b7c1
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
Don't hardcode /bin/bash, use /usr/bin/env bash instead, since not all
platforms install bash as /bin/bash.
FreeBSD, as an example, installs bash in /usr/local/bin/bash by default.
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
This gives the developer enough time to file an MR after pushing a branch.
Having this run in the first stage means we get false positives because no MR
has been filed yet when the job is run.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
ci-templates now has a new tool ci-fairy that replaces our jinja generation
script with something (eventually) unified across project repositories. Let's
move the files to the expected locations .gitlab-ci/config.yml and
.gitlab-ci/ci.template.
ci-fairy also has a wrapper to delete images, let's start using that.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
On a device with more than 256 slots we would read (and copy) past our changes
stack-allocated changes array. Fix this by capping to MAX_SLOTS though this
also requires us to memset the target where it is larger than MAX_SLOTS.
There are no real devices with 256+ slots, so this is a theoretical issue
only.
Fixes#11
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Fixes:
../test/test-common.h: In function ‘print_event’:
../test/test-common.h:97:6: error: ‘const struct input_event’ has no member named ‘input_event_sec’
ev->input_event_sec,
^
../test/test-common.h:98:6: error: ‘const struct input_event’ has no member named ‘input_event_usec’
ev->input_event_usec,
^
../test/test-common.h:102:6: error: ‘const struct input_event’ has no member named ‘input_event_sec’
ev->input_event_sec,
^
../test/test-common.h:103:6: error: ‘const struct input_event’ has no member named ‘input_event_usec’
ev->input_event_usec,
^
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Fixes the following compile failure with old toolchains and meson
build (autotools build not affected):
../tools/mouse-dpi-tool.c: In function ‘handle_event’:
../tools/mouse-dpi-tool.c:115:13: error: ‘const struct input_event’ has no member named ‘input_event_sec’
m->us = ev->input_event_sec * 1000000 + ev->input_event_usec;
^
../tools/mouse-dpi-tool.c:115:45: error: ‘const struct input_event’ has no member named ‘input_event_usec’
m->us = ev->input_event_sec * 1000000 + ev->input_event_usec;
^
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Let's not duplicate the functionality, use the upstream templates because
they'll actually see improvements over time.
Define two jobs, one that is the ifnot-exists job and one that is the normal
container-build job. The second one only runs on schedules.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
gitlab doesn't support double-nested variables so we can assume that these
variables don't need further expansion
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is a test commit only to speed things up when testing the container-clean
phase. In the real instance, we don't want to remove our containers until the
test suite successfully completes on the new container.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Rather than raw curl requests to the API, use a python script using the gitlab
python package to access everything. This makes things a bit more readable.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
These templates get rid of the various distribution-specific naming and
instead default to the namespaced FDO_DISTRIBUTION_<foo> for whatever value we
need. So FEDORA_RPMS, DEBIAN_DEBS etc. becomes FDO_DISTRIBUTION_PACKAGES for
example.
By necessity this is one large commit. gitlab does not allow nested variable
expansion, so the previous approach of global variables didn't work.
Specifically, we'd end up with a template in this form:
variables:
FEDORA_TAG: 12345
.base_template:
variables:
DISTRO_IMAGE: $DISTRO_TAG
.fedora:
variables:
$DISTRO_TAG: $FEDORA_TAG
But the actual DISTRO_IMAGE variable would be the literal string $FEDORA_TAG,
not the value of that variable. So all of it needed to be reworked.
Specifically:
- the packages to install moved to the config yaml file
- the distribution tag is now in the config yaml file
- all distributions now share the same tag (because lazyness)
- there are .fedora:30, .debian:stable, etc. templates now with the variables
defined as needed, jobs will extends those templates as they need those
distributions
- qemu-prep jobs are now generated too (based on the config yaml file)
Overall, it ends up cleaner despite the mess in this patch.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Make the GitLab spelling consistent too and remove the git protocol link, it
looks awkward in the rendering and you can get all that from the gitlab repo
anyway
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Due to how pytest is used, we don't get the actual errors printed in the
commandline log if this job fails. This doesn't matter in merge requests where
the results are displayed nicely but where there's no merge request we can't
get this info out at all. So let's simply store the xml file for an easy
check.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is a leftover from ci-templates where the name of a distribution didn't
necessarily match the image (arch vs archlinux) and is no longer needed.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is a leftover from libinput where we have to allow failures because of
frequent issues with timing. We don't have these issues here so a failed job
in the VM indicates a bug.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Unclear what happend but I suspect a rebase went wrong so
e03cdd1d3f didn't actually update to the new
ci-templates.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The KVM tests use this for now, not the container builds where we run meson
directly.
The python script to convert meson test logs to junit results expects suite
names, so let's add all tests to suites so we don't need to carry local
modifications.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This gives us a warning where qemu failed to start and it gets rid of the
different "flavor" that only archlinux had different to the distribution name.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Only one change: the meson boolean to decide whether to build with meson is
now inside the build: block.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
To avoid dnf updates and outdated packages (and the resulting delay from a dnf
update) we just install the clang-analyzer package into the default Fedora
image. It won't mess with the build expectations too much.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This cannot ever be unset on any real device, but coverity is unhappy and
that's not making me happy.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Clang doesn't support variable length arrays inside a struct so we could
either make our life complicated or just assume no-one is using more than 256
slots and hard-code that. Let's go for the easy solution until someone
notices.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Where at least one touch ends during SYN_DROPPED, we send out two event
frames: one with all applicable touch sequences ending (tracking id -1) and
the second one with the whole device state *and* the applicable touch
sequences starting (tracking id != -1).
This requires us to also update the BTN_TOOL_ bits correctly so that they are
correct after the first frame. For that we count the number of previously
known touches and send a 0 event for the matching BTN_TOOL_ bit, together with
a 1 event for the currently known touches.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The previous event processing had subtle issues with touches stopping during
SYN_DROPPED. All of the device state was processed in the same frame, but if
any touch changed tracking ID during SYN_DROPPED, an inserted SYN_REPORT
resulted in a weird split of events:
- the first frame had all key/sw/abs updates including those slots that
changed tracking ID, but not the ones that were fully terminated.
- the second frame had only the slots states for newly started touches **and**
the slot state for touches terminated during SYN_DROPPED but not restarted.
In other words, where three fingers were on the touchpad and slot 0 was lifted
and put down again and slot 1 was lifted but *not* put down again, our frames
contained:
- frame 1: terminate slot 0, BTN_TOOL_TRIPLETAP 0, BTN_TOOL_DOUBLETAP 1
- frame 2: start slot 0, terminate slot 1
Where there was no touch changing tracking ID, only one frame was generated.
The BTN_TOOL updates were buggy, they may not match the number of fingers down
as seen on a frame-by-frame basis. This triggered libinput bug
https://gitlab.freedesktop.org/libinput/libinput/issues/422
This patch changes the above example to
- frame 1: terminate slot 0, terminate slot 1
- frame 2: start slot 0, BTN_TOOL_TRIPLETAP 0, BTN_TOOL_DOUBLETAP 1
Notably, the first frame no longer contains the BTN_TOOL bits. This patch is
one of two, the BTN_TOOL sync bits are part of a follow-up patch.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Go from:
if (a != b)
continue;
foo;
to:
if (a == b) {
foo;
}
Basically just an indentation change after the condition inversion, makes the
follow-up patch easier to review.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
In the near future, we will need to handle slot termination *before* any other
state synchronization. So let's start splitting things up.
This is functionally equivalent though dropping the need_tracking_id_changes
variable means we run through all slots now to check if we need to terminate
one of them. Given the normal number of slots on a device and that this should
only ever run very rarely anyway... meh.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Keep a better state of each touch before/after the SYN_DROPPED. Most of this
is currently unused, it's functionally the same as before but the new code
serves to increase readability and it can be passed around easier this way.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We don't really care that they're F31, that's an implementation detail. So
let's rename them so we can easily pick which job is which on the pipeline
overview.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Notable: the meson builds don't have a "nm is missing" target because meson
needs it for itself.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Centos doesn't run meson because it's too hard to install the package with dnf
and I can't be bothered going through pip.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
There is so much duplication between the various jobs that it's hard to keep
track of it manually. Let's employ a python script to generate those bits,
reducing the actual gitlab-ci.yml to the hand-written parts only.
The new script takes the .gitlab-ci/gitlab-ci.yml.in and simply appends the
generated parts to it. Most of it is straightforward, only centos needs some
custom parts because of missing doxygen.
The diff is a bit hard to review, thanks to the python script we now group
based on distribution, not based on name (i.e. all fedoras in one group
instead of all container-preps in one group).
And since we're generating anyway, some of the in-between stages were removed
(e.g. $DISTRO-build@template).
A new CI job is added to run a diff against the .gitlab-ci.yml that's checked
in and the one generated by this script. If they differ, we fail.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Move the centos builds to after the ubuntu builds and swap the two fedora
builds. Just we have the same order for things here as in the container
prep/clean phases and to make a future patch easier to review.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Replace it with a stack-allocated one. This saves us a bunch of confusing
allocations and size calculations.
And in the process use uint32_t/int32_t to ensure the struct is actually the
expected size.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Coverity tries to supply system headers and fails badly at it. A bunch of
_Float... sizes are pulled in by math.h but not provided anywhere. So as a
workaround, let's add an option to explicitly enable coverity support that
simply #defines those types to ones we do know about and let's go on hoping
it'll eventually work.
See the equivalent addition to libinput in commit 8178339b5baa717.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
v3.4 was released in 2012, every kernel since has that ioctl. So instead of
assuming you're running new libevdev on an 8 year old kernel, let's assume
that any error from the ioctl() is an actual error and handle it accordingly.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
For debugging it's more important to be able to quickly run a single test
rather than grouping them together, we don't have thousands of tests here
anyway. So let's add a macro to put every test func into its own TCase,
allowing for test selection via the environment variable CK_RUN_CASE.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is a GNU C extension, and is not available in ISO C.
Instead, just explicitly initialize other indices to -1.
Signed-off-by: Michael Forney <mforney@mforney.org>
Statement expressions are a GNU C extension and are not available
in ISO C.
On compilers that don't have them, define these macros as plain
conditional expressions, since they are only ever used with expressions
that have no side-effects.
The statement-expression version is still retained as an added
safety measure on GNU-compatible compilers.
Signed-off-by: Michael Forney <mforney@mforney.org>
With b65be7aa79 the centos 7 RPMs don't include doxygen anymore, so let's drop
the distcheck job. Centos 7 is old enough at this point that we don't need to
care too much.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
With b65be7aa79 the centos 7 RPMs don't include doxygen anymore, causing a CI
pipeline failure (but only where the centos image is being rebuilt).
Fedora is a better early-warning system about possible breakages than Centos -
which was originally chosen precisely because it doesn't update much and was
faster to run in the CI (this was before custom docker images).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We're only testing EV_SW event delivery here and SW_LID has the tendency to
suspend the host when we don't specifically inhibit it. So let's just swap for
the next one.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Mostly copy/paste from libinput but we do build on debian and centos as well.
The special builds are basically the same as before.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
That code did not compile because those functions were only renamed in
header and code back then, but not in the example.
Fixes: ab2f20bfd6 ("Revamp the API once again")
Signed-off-by: Alexander Dahl <ada@thorsis.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The id_* fields are 16 bits in linux/input.h and we mirror
the kernel API here. Even though we accept an int for this
fields in ABI the value is truncated at 16 bits.
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Previously, enabling or disabling ABS_MT_SLOT would not change the actual
slots, it was treated as a normal bitflag. This means we couldn't initialize a
libevdev context from scratch and have it behave like a correct MT context.
Fixes#4
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Revert back to the normal doxygen looks, even though they're pretty awful. But
at least they're consistent, the custom stylesheets had all sorts of weird
corner cases that didnt' render correctly and I don't have enough CSS
knowledge to fix this correctly.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This causes some weird rendering, let's split it into a list (which also
happens to be more readable).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Two new function pairs:
libevdev_event_code_from_code_name()
libevdev_event_type_from_code_name()
libevdev_event_code_from_code_name_n()
libevdev_event_type_from_code_name_n()
These functions look up event codes/types by the name of the event code only,
removing the need to figure out what event type an event code has. So if all
you have is "BTN_TOUCH", you can now look up the type and code for that,
without having to check the prefix yourself to guess at the type.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
With the previous approach, every libevdev_next_event() invocation triggered a
read() on the device fd. This is not efficient, the kernel provides whole
event frames at a time so we're guaranteed to have more events waiting unless
the current event is a SYN_REPORT.
Assuming a fast-enough client and e.g. a touchpad device with multiple axes
per frame, we'd thus trigger several unnecessary read() calls per event frame.
Drop this behavior, instead only trigger the read when our internal queue is
empty and we need more events.
Fallout:
- we don't have any warning about a too-slow sync, i.e. if a SYN_DROPPED
arrives while we're syncing, we don't get a warning in the log anymore.
the test for this was removed.
- the tests that required the specific behavior were rewritten accordingly
- a revoke on a kernel device doesn't return ENODEV until already-received
events have been processed
The above shouldn't be an issue for existing real-world clients.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
ABS_RESERVED was added to 4.20 for that reason, to keep that event code
reserved so we can't use it for anything else (and thus mess up the fake MT
detection).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Two variable renames for less ambiguity
Two changes from an long if condition to a "if foo in [...]"
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This flag was used to disable test runs during make distcheck. Now that we
have more checks and the environment variable, we can drop this flag.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Presumably this worked, but it certainly doesn't work anymore (neither on
RHEL7 nor on Fedora 28)
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Depending on the container, or other checks don't always work. Add an extra
environment variable instead.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We previously had this separate because it tested separate things. Now the
setup is generic enough that we should just re-use it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Even if we're root we may be running in a container without input device
nodes. In which case we should skip the test, not fail.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
It's not really autodetection, we just declare the test suites that need root
privs. But this way we can generically check for it from the main() that we
re-use across tests.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
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>
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>
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>
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>
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>
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>
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>
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.
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>
Side-effect of 240ba34ebd 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>
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>
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>
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>
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>
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>
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>
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>
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/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>
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>
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>
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>
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>
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>
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 debe9b030c, this patch avoids
the ioctl/close calls if the fd isn't set.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Some devices scale the frequency based on the input and will provide
recordings with different frequencies each time. Recommend to measure multiple
times since we can only know what the highest frequency is.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
clang looks at GCC pragmas, but doesn't understand -Woverride-init. Instead,
it uses -Winitializer-overrides.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Relative devices don't provide a physical resolution to the host. For things
like pointer acceleration, the physical amount of movement is better as
baseline than the movement in device units.
Alas, many devices don't come with any information at all, so the users have
to guess. Help that guesswork by providing a tool that does the calculations
for them.
This tool measures the device units covered, then prints the frequency and an
lookup table for various resolutions (in dpi) to match to the physical
movement of the device.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
These were removed in (7da329b) because for some reason they got copied into
the output. That was either a buggy doxygen or just some other problem. Add
them again, makes it much easier to read the header file.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
When building for multiple prefixes from the same repo it's nice to see a
reminder for what the current prefix is.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
I can't recommend it, but I can't stop people from doing it, so at least
document the ground rules.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We ignore anything that starts with an underscore, anything that starts with
libevdev and main (since we test the test-static-link binary) and a couple of
gcov-related functions. This should catch any functions we accidentally
export.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We rebuild the libevdev object files with the gcov flags for the tests, so we
need to make sure those flags are present on all binaries.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This can't fail in the kernel anyway, so cast it to shut up Coverity.
Error message:
"Calling function "ioctl(int, unsigned long, ...)" without checking return
value (as is done elsewhere 35 out of 36 times)."
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Using LIBEVDEV_UINPUT_OPEN_MANAGED can leak the fd if an error occurs after
opening it.
Found by Coverity.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The static library currently leaks log_msg and log_priority. Both are too
generic, so rename them, with a leading underscore to hint they're supposed to
be private.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
12717d79 "Add libevdev_event_type/code_from_name() resolvers" added the
lookup functions for types and codes, this commit adds the missing ones for
input properties.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
The python check was inserted between "checking whether to build with gcov"
and the answering "yes", causing a lonely "yes" or "no" to wander around the
config.log in vain, haunted by solitude and wondering about the binary
simplicity of its life.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
A simple ck_assert() just fails, a ck_assert_int_eq() also shows the two
values. Makes failures a bit easier to debug.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We use the negative errno internally, but the proper errno is always positive.
Fixes device creation failures on kernels that don't support UI_SET_PROPBIT.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Older kernels don't have UI_GET_SYSNAME, and upstream is adding a few more
ioctls to the uinput code. So ship the header we're using to avoid compilation
errors.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Better protection against buffer overflow, though by the time someone
is manipulating your sysfs, libevdev is unlikely to be the biggest worry.
Slight change in functionality: before we checked the timestamp of
/sys/devices/virtual/input/inputXYZ before looking at /inputXYZ/name, now we
just check the name file for the timestamp.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
The global log handler isn't a good choice for a low-level library. In the
caser of the X server, both evdev and synaptics are now using the libevdev but
are loaded from the same server process. Thus, there's only one log handler,
but evdev and synaptics don't talk to each other (a bit childish, I know).
Add a per-device log handler that overrides the global log handler, and fall
back to the global log handler if no device log handler is set. The log
macros take care of that automatically, especially as we can't do per-device
log handlers for the uinput code.
Note that we use the same struct for the global and device logging, so in each
instance one of the two function pointers is NULL. Suicide triggers are in
place in case we mess that up.
This also makes libevdev_new_from_fd() a bit less useful since we can't set
the log handler beforehand.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
These macros come from systemd, but CC_CHECK_CFLAGS_SILENT was removed in
systemd commit eb2e280f9c59b66965c9316eadc4c113a13ca744, breaking some of
them.
CC_FLAG_VISIBILITY doesn't need to require CC_CHECK_WERROR because
CC_CHECK_FLAG_APPEND always append -Werror anyway. Which kinda brings into
question why we have CC_CHECK_WERROR in the first place, but whavever.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The tests create devices on the host system, avoid running them in parallel to
avoid interference between the test devices.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
The kernel ring buffer drops all events on SYN_DROPPED, but then continues to
fill up again. So by the time we read the events, the kernel's client buffer is
essentially like this:
SYN_DROPPED, ev1, ev2, ev3, ...., evN
The kernel's device state represents the device after evN, and that is what
the ioctls return. For EV_KEY, EV_SND, EV_LED and EV_SW the kernel removes
potential duplicates from the client buffer [1], it doesn't do so for EV_ABS.
So we can't actually sync while there are events on the wire because the
events represent an earlier state. So simply discard all events in the kernel
buffer, synchronize, and then start processing again. We lose some granularity
but at least the events are correct.
[1] http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/input/evdev.c?id=483180281f0ac60d1138710eb21f4b9961901294
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Follow-up to
commit 41334b5b40
Author: Peter Hutterer <peter.hutterer@who-t.net>
Date: Thu Mar 6 11:54:00 2014 +1000
If the tracking ID changes during SYN_DROPPED, terminate the touch first
In normal mode, we may get double tracking ID events in the same slot, but
only if we either have a user-generated event sequence (uinput) or a malicious
device that tries to send data on a slot > dev->num_slots.
Since the client is unlikely to be able to handle these events, discard the
ABS_MT_TRACKING_ID completely. This is a bug somewhere in the stack, so
complain and hobble on along.
Note: the kernel doesn't allow that, but we cap to num_slots anyway, see
66fee1bec4.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
We can't allocate in sync_mt_state since it may be called in the signal
handler. So pre-allocate based on the device's number of slots, store that in
the libevdev struct and use it for the sync process.
This fixes a remaining bug with the handling of ABS_MT_TRACKING_ID. If a
device had > MAX_SLOTS and a slot above that limit would start or stop during
a SYN_DROPPED event, the slot would not be synced, and a subsequent touch in
that slot may double-terminate or double-open a touchpoint in the client.
For the effects of that see
commit 41334b5b40
Author: Peter Hutterer <peter.hutterer@who-t.net>
Date: Thu Mar 6 11:54:00 2014 +1000
If the tracking ID changes during SYN_DROPPED, terminate the touch first
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
For protocol A devices we won't get the information from the kernel anyway and
we expect all axes to be updated in the next event.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
The EVICOCGMTSLOTS ioctl returns all slot values for the requested code or an
error code, it doesn't return the number of bytes successfully transferred.
Thus all values in the input array are always defined (on success), we don't
need to memset it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
dev2 by definition doesn't initialize, we expect it to fail. Freeing it after
is a bad idea. Also initialize it to NULL so this is a bit more obvious now.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
A max of num_slots -1 caused the first MT_SLOT event to be skipped, leading to
wrong tracking IDs in the slots.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If a touch starts and terminates while in SYN_DROPPED, the tracking ID appears
to stay at -1, but the other axes may update. We need to pass these on to the
client since the kernel may buffer the next event with the same value
otherwise. Note this in the documentation so that client's don't create touch
points based on out-of-touchpoint updates.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This doesn't really do much here, but strictly speaking: if asprintf returns
-1, devnode is undefined. So reset it to NULL to avoid weird pointers. And
also free the rest of the names if we ever have more than one device - which
also shouldn't happen.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
test-libevdev-events.c: In function ‘test_double_syn_dropped_event’:
test-libevdev-events.c:187:2: warning: ignoring return value of ‘read’,
declared with attribute warn_unused_result [-Wunused-result]
This read was there to drain events even when there shouldn't be any on the
pipe anyway. So let's add an assert.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Triggered with -O3
../libevdev/libevdev.c: In function ‘libevdev_get_event_value’:
../libevdev/libevdev.c:1112:6: warning: ‘value’ may be used uninitialized in
this function [-Wmaybe-uninitialized]
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Triggered with -O
../libevdev/libevdev.c: In function ‘libevdev_has_event_code’:
../libevdev/libevdev-util.h:58:20: warning: ‘mask’ may be used uninitialized
in this function [-Wmaybe-uninitialized]
Optimisation clearly shuffles things around here: in the code, if no max is
found, we return -1 and bail out before we access mask.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
If a client doesn't sync expliciltly, make sure we sanitize the events when we
update the internal library state.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
"Related Pages" in doxygen are ordered in the order they appear in the source
file. The internal test suite is least likely to be of interest to the
reader, so move it to the bottom.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This is getting a bit complex, so add some high-level documentation that we at
least know what we're trying to do.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Clients may not care about the events generated during SYN_DROPPED, but the
current slot must be updated to avoid a client being out-of-date.
Same with tracking IDs, if they changed, the caller will likely have to update
some internal states.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Most clients can't deal with tracking ID changes unless a -1 is sent first. So
if we notice that the tracking ID has changed during the sync process, send a
set of ABS_MT_TRACKING_ID -1 events for each of those, then send the rest of
the events.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
If multiple slots have changed during the sync handling, the client must be
re-set to the current slot before continuing with normal events.
Signed-off-by: Benjamin Tissoires <btissoir@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Devices with ABS_MT_SLOT-1 are fake MT devices, they merely overlap the
axis range but don't actually provide slots. The EVIOCGABS ioctl won't work to
retrieve the current value - the kernel does not store values for those axes
and the return value is always 0.
Thus, simply ignore those axes for fake MT devices and instead rely on the
next event to update the caller with the correct state for each axis.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This allows libevdev-util.h to be used by tests, it no longer relies on
libevdev internal structs.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
A malicious device may announce N slots but then send a slot index >= N. The
slot state is almost always allocated (definitely the case in libevdev and
true for most callers), so providing a slot number higher than the announced
maximum is likely to lead to invalid dereferences. Don't allow that.
Likewise, don't allow negative slot numbers.
Note that the kernel filters these events anyway, the only way to trigger this
is to change the device fd to something outside the kernel's control.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
When syncing, we expect the slot to stay the same until the client has
processed the events. This already worked, just add a check to make sure.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
As seen on 3M devices, which seems to be the maximum seen so far. Some Stantum
devices report 255 touches but are only capable of 10, so the are not affected
by our limits.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Changes the algorithm: before we'd ioctl all axes for all slots, then generate
events for all slots one-by-one.
Now we ioctl the slot state for each axis, copy the new event value into
the device and mark a bitfield that we've updated the value. Then loop through
the slots and generate events where changed.
Side-effect: this makes it easy to check if anything in the slot has updated,
so we can skip empty slot events during sync.
Min memory requirement for the state storage was:
MAX_SLOTS * (ABS_MT_CNT + 1) * sizeof(int) = 1980
Min memory requirement now:
(ABS_MT_CNT + 1) * sizeof(int) + NLONGS((MAX_SLOTS * ABS_MT_CNT) bits) = 544
This is sigsafe code, so this was stack memory. Reducing the requirement
allows us to up MAX_SLOTS in the future if we need to.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Instead of relying on a static MAX_SLOTS array, allocated it based on the
number of slots we have on the device. The previous checks for MAX_SLOTS were
incomplete, causing out-of-bound reads.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If a device has more than MAX_SLOTS slots, we'd run out-of-bounds on the sync
array. This function is sig-safe, so we can't alloc here, merely limit the
access.
Reported-by: Jonas Ådahl <jadahl@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
For an invalid fd, or a failure to open the device, the pre-allocated uinput
device struct would leak.
We can drop the open_uinput() function now, since skipping to the error
handling means we'll return -errno anyway.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If a device is assigned a name, uniq and/or phys before calling
libevdev_set_fd(), those values would leak.
Change the default alloc to calloc, so name, uniq, and phys are initialized to
zero before we call libevdev_reset
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
It's a pointer to the copy libevdev has, but that copy may be freed by the
caller if someone calls e.g. libevdev_set_name()
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If a message is higher than the current priority, filter it. And add a few
tests that the priority is handled the way it should.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
We're a bit inconsistent here anyway because of the ample use of "non-zero"
which can't really be converted to non-0. But let's at least be consistent
with the use of 0 instead of zero.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Bump once to the stable interface we have now. This isn't really required as
we're compatible with everything but functions deprecated with 0.4 but we
might as well do it now to force a rebuild of everyone. Should've done that
for 0.6 when we actually dropped the function, but ...
Now that we have proper library versioning we shouldn't need any
backwards-incompatible soname bumps for a while.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Explicit symbol versioning allows us to provide multiple versions of
incompatible API changes. It is a very common practice in GNU world and
avoids the problems occuring if distributions try to ship multiple version
of a single DSO.
Background information available at:
http://people.redhat.com/drepper/dsohowto.pdf
To see the symbol-versions, use objdump:
objdump -T libevdev/.libs/libevdev.so.1.0.0
This can also be used to verify that all symbols are correctly
exported (typos in linker-scripts are silently ignored by ld).
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Updated for evdev 0.6's API.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If the first event after a completed device sync is a SYN_DROPPED, warn the
user that they're not fast enough handling this device.
The test for this is rather complicated since we can't write SYN_DROPPED
through uinput so we have to juggle the device fd and a pipe and switch
between the two at the right time (taking into account that libevdev will read
events from the fd whenever it can).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Add two log functions, one that aborts on a received message. We know when we
expect to receive an error, so anytime this happens unexpectedly should
terminate the test.
And for those tests do issue a log message, let them ignore it and don't
print anything.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Make sure we have a queue that is at least large enough to do a full sync
after a SYN_DROPPED, plus store a few extra events in case some came in after
the sync.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
We only take one option (path to linux/input.h) anyway so drop the argparse
requirement. This way libevdev builds on RHEL6 too which doesn't ship
argparse.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This is a leftover from when the file was in evemu and used to generate python
headers too.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
The Check test framework forks by default which is annoying when running gdb.
Try to detect whether we're inside gdb by ptracing ourselves. If that works,
we're not inside a debugger. If it doesn't, then assume we're inside a
debugger and set CK_FORK to "no".
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This test doesn't do anything but compile and link against libevdev. It's a
simple protection to avoid linker errors. If we ever have libs we depend on
and they don't get resolved properly, this test should warn us in time.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Some devices (PS3 sixaxis controller) merely have a bunch of axes, without the
semantic information that linux/input.h requires. For those, the ABS_MT range
may be merely another axis, not the special range that we need to treat it
with.
Use a simple heuristic: if ABS_MT_SLOT - 1 is enabled, don't treat ABS_MT as
multitouch axes. The ABS_MT_SLOT - 1 axis is not used for a real axis.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
All clients that want to handle SYN_DROPPED correctly need to pass an EV_SYN
through their own handlers before starting with the syn events. Rather than
letting them synthesize that, guarantee that the event is defined the first
time LIBEVDEV_READ_STATUS_SYNC is returned.
This does not change existing behavior, it merely documents it so we can rely
on it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We shouldn't have a separate API for that, the whole point of libevdev is to
abstract the quirkyness of the ioctls into a common interface. So let's
export the two EV_REP values through libevdev_get_event_value.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
We don't provide man pages (the ones created by doxygen are pretty
terrible) so provide a placeholder page to provide the minimum info and point
people in the right direction.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If we don't set AUX_DIR, I get errors during ./configure:
/bin/sh: /home/david/missing: No such file or directory
configure: WARNING: 'missing' script is too old or missing
This error is ignored by default, but may cause failures with newer
autoconf versions. Simply set the AUX_DIR to "build-aux" like many other
autotools projects do.
[This is an automake 1.14 bug, see the upstream report:
http://lists.gnu.org/archive/html/bug-automake/2013-11/msg00007.html]
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
-Wpedantic is a relatively new option, with -pedantic being the old version of
it that works on older gcc versions too.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
For the scenarios the module supports it is not required to
be specified.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The reason why this did not work is that we do not have a fixed list
of html doc files and it would be pain to maintain.
The 'find' hack gets a list of html files but only when the package
has already been built. Let's complement this hack with the
html/index.html file which act as a build trigger as it is the target name.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Everytime make is invoked, doxygen is invoked to generate the html doc.
Dependencies are added so it does rebuild when necessary.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Every time a make target is run in the doc directory when the html dir
is absent, you get this error. Even with 'make clean'.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
It reflects the values on the build machine.
Same pattern as libevdev.pc vs libevdev.pc.in.
For example, in libevdev.doxygen:
HTML_EXTRA_STYLESHEET = ./libevdev.css
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
A user of libevdev may be compiled with -Wpedantic. Our header files should
not produce any warnings, so add a simple test that merely includes both
public header files and compiles with -Wpedantic -Werror.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
libevdev.c:921:134: warning: ISO C does not allow extra ';' outside of a
function [-Wpedantic]
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
This seems like a reasonable minimum version to require.
Available in RHEL6 (2010) or Ubuntu (2010).
Version 2.7 contains the argparse module so the check could be removed.
The argparse module was introduced in v2.3.5 as an optional module.
http://www.python.org/doc/versions/
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Given the close ties with input drivers and xserver, might
as well use the same level of compiler.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Creating the m4 directory is not required under any circumstances
when the package is properly configured. We have git files under m4.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The configure script is invoked from the build directory rather than
from the source directory.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Merge potentially useful patterns taken from other projects.
Some application specific patterns were move to their respective directories.
The only noticeable change is that *.patch is ignore to prevent accidental
checkin of patches. The pattern "test-driver" could not be found and was
removed.
The test directory had not been updated since the move of all test cases
in a single binary.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We recompile the sources anyway in the tests and we have our own CFLAGS there.
Drop them from the main library code so we don't leave gcov files everywhere.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Use AC_LINK_IFELSE instead of AC_COMPILE_IFELSE to test for flags that
might succeed during compilation but not during linking. An example is gcc
compiled with libssp support but gnu-ld without it. In this case
-fstack-protector works fine during compilation but fails during linking
as several internal helpers are missing.
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This macro is used to verify that $srcdir is set correctly by testing for
existance of the passed file. Shouldn't hurt setting it to catch wrong
build setups.
Cc: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Three new helpers are added:
(1) libevdev_event_type_from_name() takes a string describing an EV_*
event type and returns the given event-type constant.
(2) libevdev_event_code_from_name() takes a string describing an event
code and returns the given event-code constant.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Additionally to type->string mapping tables we now also create
string->type lookup tables. The lookup tables are sorted by their name
so binary-search will work.
We create one lookup table for EV_* types and one for all event-codes.
More tables (like bus-names, input-properties, ..) can be added later.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Simple helper that use strncmp() to test whether a given string
starts with a given suffix. While it doesn't reduce code-size by much,
it is a lot more readable than the strncmp() usage.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We can rely on CLOCK_MONOTONIC and CLOCK_REALTIME to be different at any
time. However, this does not apply to the ms/us/ns parts of the current
time. Both may be in sync regarding the current micro-seconds state. So
remove the wrong clock us-comparison.
I was able to trigger this on my machine. Chances that both are in sync
are very low so I assume my RTC only provides low granularity and thus
both clocks are sync during boot for higher granularity.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
According to Gaetan, that's what we really require. And as a bonus point, this makes libevdev build
on RHEL6 out-of-the-box.
CC: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
There's a gap in the range between EV_SW and EV_LED. Trying to enable one
of those bits will segfault.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Set the bit during device reset and make sure that if we're checking
for the event type we always return true for EV_SYN.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
libevdev_set_fd may fail at a number of points. If it does, it errors out but does nothing
otherwise. Thus, a client may call set_fd again for the same struct but on a different fd and have
it succeed. Depending on when set_fd bailed out the first time, some fields may already be set.
Thus, reset the whole struct at set_fd time to make sure we're nulled out appropriately.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
This avoids a number of otherwise required ifdefs when building on older kernels
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Add a new flag for "initialized" and separate that from the fd logic. This way,
we can call libevdev_change_fd(dev, -1) to signal that the current fd should be
dropped.
Otherwise libevdev can't be told to release the fd and always keeps a reference
to it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
These are deprecated, but were missing the deprecated attribute.
And fix up the tests that were still using those deprecated calls.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Only add files to EXTRA_DIST, to avoid having a full copy of doc/html as
doc/html/html in the distributed tarball.
Signed-off-by: Stephen Kitt <skitt@debian.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
http://lists.freedesktop.org/archives/input-tools/2013-September/000332.html
See above a detailed reasoning, but Colin Walters put it this way:
"At a high level, I think components (git repositories) should feel free
to set up default warning flags and possibly use a targeted subset of
-Werror=foo. But please don't inject non-warning flags like this unless
there is a very good reason.
The right way to do -fstack-protector is to have something like
redhat-rpm-config or other global CFLAGS system controlling *all*
components."
Thus, reverting the -fstack-protector flag. This doesn't completely address
Colin's comments, we still use a few other flags. But this one is the one
that causes real headaches, so drop it.
This reverts commit f5e65ea3ce.
We removed this previously due to build failures in:
commit 14ac764ec8
Author: David Herrmann <dh.herrmann@gmail.com>
Date: Mon Sep 9 16:03:41 2013 +0200
build: remove -fstack-protector
Reintroduce it but this time disable it if it's not supported. We use the
CC-flags testing to prevent build-failures on gcc without libssp support.
Cc: Giovanni Campagna <gcampagn@redhat.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Use the CC_CHECK_FLAGS_APPEND() m4 macro to test for availability of
CFLAGS and LFLAGS. It automatically drops the unavailable flags.
This is also used by systemd, so it ought to work with ostree and other
non-standard build-environments.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
There is a legitimate reason for clients to force a sync on the device.
X.Org drivers lose the fd when the device is disabled and re-enabled. When
the device comes back, a simple libevdev_change_fd() doesn't update the status
on the device.
Button states, etc. may have changed, etc. So a driver may call FORCE_SYNC after
re-connecting to the fd to make sure the library and the driver get the current
state of the device.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Rename from LIBEVDEV_READ_foo to LIBEVDEV_READ_FLAG_foo to differentiate
better from LIBEVDEV_READ_STATUS_foo.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Improved readability in callers, changing magic numbers 0 and 1 to
rc = libevdev_next_event();
if (rc == LIBEVDEV_READ_STATUS_SUCCESS)
do_something();
else if (rc == LIBEVDEV_READ_STATUS_SYNC)
do_something_else()
No ABI changes, the enum values are the previously documented values,
this is just a readability improvement.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Another look at the current API showed some inconsistencies, rectified
in this commit:
libevdev_kernel_*: modify the underlying kernel device
libevdev_event_type_*: something with an event type
libevdev_event_code_*: something with an event code
libevdev_event_*: struct input_event-related functions (i.e. not device-related)
libevdev_property_*: something with a property
libevdev_*: anything applying to a device
Hopefully that's the last API change. Current symbols deprecated and aliased.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
If gcc is built without libssp support, it loudly fails linking due to
missing __stack_chk_*() helpers. Unfortunately, gcc isn't smart enough to
disable it automatically.
systemd recently got a CC_CHECK_FLAG_APPEND helper to work around such
issues:
http://cgit.freedesktop.org/systemd/systemd/tree/m4/attributes.m4
I didn't want to add it now, so let's just drop -fstack-protector. If we
want it, we can try adding it later again.
This partially reverts:
commit 43752ec17d
Author: David Herrmann <dh.herrmann@gmail.com>
Date: Sun Sep 1 17:45:04 2013 +0200
Add some gcc/ld optimizations and magic
All other gcc/ld options are kept.
Reported-by: Giovanni Campagna <gcampagn@redhat.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The logging we do use atm inside the library is largely
to spot application errors. Log that to stderr by default so
it doesn't get lost.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Change to the previous code in that we continue looking at devices
even after we've found one. However, this way we can warn
the user when we can't guarantee syspath correctness.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
There's no need to have separate logging function for each device created.
More likely, libevdev will be hooked up once into the logging system and
expected to deal with it.
Plus, this allows us to log from the uinput code where we don't
have the context anyway.
Requires a rename to libevdev_set_log_function to avoid ABI breaks, and
while we're breaking the ABI make the logging function more sophisticated
to log line, number, etc.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
There are several gcc/ld flags that optimize size and performance without
requiring explicit code changes. In no particular order, this adds:
- gcc -pipe to avoid temporary files and use pipes during compilation
- gcc -fno-common avoids putting uninitialized global variables not
marked as "extern" into a common section. This catches compilation
errors if we didn't mark global variables explicitly as "extern".
- gcc -fno-strict-aliasing allows us to use unions for some binary magic.
Otherwise, -O2 might assume that two different types never point at the
same memory. We currently don't rely on this but it's common practice
so avoid any non-obvious runtime errors later.
- gcc -ffunction-sections and -fdata-sections put each function and
variable into a separate section. This enables ld's --gc-sections to
drop any unused sections (sections which aren't referenced from an
exported section). This is very useful to avoid putting dead code into
DSOs. We can now link any helper function into libevdev and the linker
removes all of them if they're unused.
- gcc -fstack-protector adds small stack-corruption protectors in
functions which have big buffers on the stack (>8bytes). If the
stack-protectors are corrupted, the process is aborted. This is highly
useful to debug stack-corruption issues which often are nearly
impossible to catch without this.
- ld --as-needed drops all linked libraries that are not actually
required by libevdev. So we can link to whatever we want and the linker
will drop everything which is not actually used.
- ld -z now, resolve symbols during linking, not during runtime.
- ld -z relro, add relocation-read-only section. This allows to put
read-only global variables and alike into a read-only section. This is
useful for variables that need a relocation and thus cannot be
explicitly put into a read-only section. This option tells the linker
to mark them read-only after relocations are done. (that's why -z now
makes sense in combination with this)
All of these options are common in other open-source projects, including
systemd and weston. Don't ask me why they are not marked as default..
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If we have it, stop searching for it. Otherwise a second device with the
same name would overwrite the first, causing a leak.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
All other functions that check the fd for validity return EBADF, which also makes it
easier to debug if the actual device goes away.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
From errno(3):
ENOMEM Not enough space (POSIX.1)
ENOSPC No space left on device (POSIX.1)
when we run out memory the reason is a failed malloc, for which ENOMEM
seems more appropriate.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
The ioctls return the number of bytes copied into the destination, so just
copy them into the device state instead of individually flipping bits.
For easier review: rc is the return value of the EVIOCG* ioctl, which is
the number of bytes copied.
state variables must be initialized to 0 now, in case the kernel's FOO_MAX
is smaller than libevdev's FOO_MAX. If not initialized to 0, the bytes
between the two max values is undefined and we may end up generating bogus
events.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This lets libevdev provide a relatively generic interface for the
creation of uinput devices so we don't need to duplicate this across
multiple projects.
Most of this is lifted from the current test implementation, with a
couple of minor changes.
EV_REP needs special handling:
Kernel allows to set the EV_REP bit, it doesn't set REP_* bits (which we
wrap anyway) but it will also set the default values (500, 33).
Device node is guessed based on the sysfs path:
The sysfs path contains a eventN file, that corresponds to our
/dev/input/eventN number. Use it so clients can quickly get the device
node, without a libudev dependency.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
GCC_CFLAGS was set by configure, but never actually used. So we didn't build with hidden symbols,
rather just weeded them out later through libtool.
Do this properly now, mark the symbols as visibility default.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
These are just to shut the compiler up, in all three cases we only access
the respective values if they're defined.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This will check for invalid types for us, and we can do a better check
for signed/unsigned comparison.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Having the declarations in the form of [0...EV_MAX] = NULL together
with the actual definitions causes warnings for every true definition if
-Woverride-init is enabled.
We can drop them for most as they're zero, but still need them for
ev_max (which defaults to -1), not zero. So use the GCC pragma to
disable the warnings for this file.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Python 3 doesn't have a print statement, only a print function.
Fixed with:
2to3 make-event-names.py | git apply
Print as function requires Python 2.6 which is reasonable enough given
that even RHEL6 ships that.
Even though it's not needed for 2.6, use
from __future__ import print_function
to avoid accidentally introducing a print statement in the future.
With this line, print "blah" is now a syntax error in python 2.
This line was added manually, after the 2to3 conversion.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Previous find command wouldn't actually list the generated documentation
files, so they'd be missing in the tarball.
...and add a dist-hook to make sure this won't happen during release.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
This way any ABS_MT_ event value that comes in will also be stored in abs_info.
That always corresponds to "current slot", so if a user calls
libevdev_set_event_value() or libevdev_get_event_value() they're actually
modifying the current slot value.
When the current slot changes, sync the state back into the absinfo values.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Other clients of an evdev device need to have the events they receive
be separated, in moment in time, from other events by an EV_SYN/
SYN_REPORT. This is the responsibility of the client who writes events
into the stream.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
The "Entering/Leaving directroy /xyz" messages only add noise to
make-output. Even without we still get short hints which directory is
currently processed. Compare the two outputs below:
With --no-print-directory added:
make --no-print-directory all-recursive
Making all in doc
doxygen libevdev.doxygen
Making all in libevdev
python2 ./make-event-names.py --output=c > event-names.h
make all-am
CC libevdev.lo
CCLD libevdev.la
Making all in tools
CC libevdev-events.o
CCLD libevdev-events
Making all in test
CC test-main.o
CC test-event-names.o
CC test-libevdev-init.o
CC test-libevdev-has-event.o
CC test-int-queue.o
CC test-libevdev-events.o
CC libevdev.o
CC test-common-uinput.o
CC test-common.o
CCLD test-libevdev
Without it:
make all-recursive
make[1]: Entering directory `/home/david/dev/libevdev'
Making all in doc
make[2]: Entering directory `/home/david/dev/libevdev/doc'
doxygen libevdev.doxygen
make[2]: Leaving directory `/home/david/dev/libevdev/doc'
Making all in libevdev
make[2]: Entering directory `/home/david/dev/libevdev/libevdev'
python2 ./make-event-names.py --output=c > event-names.h
make all-am
make[3]: Entering directory `/home/david/dev/libevdev/libevdev'
CC libevdev.lo
CCLD libevdev.la
make[3]: Leaving directory `/home/david/dev/libevdev/libevdev'
make[2]: Leaving directory `/home/david/dev/libevdev/libevdev'
Making all in tools
make[2]: Entering directory `/home/david/dev/libevdev/tools'
CC libevdev-events.o
CCLD libevdev-events
make[2]: Leaving directory `/home/david/dev/libevdev/tools'
Making all in test
make[2]: Entering directory `/home/david/dev/libevdev/test'
CC test-main.o
CC test-event-names.o
CC test-libevdev-init.o
CC test-libevdev-has-event.o
CC test-int-queue.o
CC test-libevdev-events.o
CC libevdev.o
CC test-common-uinput.o
CC test-common.o
CCLD test-libevdev
make[2]: Leaving directory `/home/david/dev/libevdev/test'
make[2]: Entering directory `/home/david/dev/libevdev'
make[2]: Leaving directory `/home/david/dev/libevdev'
make[1]: Leaving directory `/home/david/dev/libevdev'
We don't really win any useful information from these messages.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We put m4 files into a separate directory. If we ever put our own files in
there, we need to include it. Furthermore, this suppresses annoying
autotools warnings.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This enables libevdev_get_event_value(dev, EV_LED, LED_NUML); to check
if a LED is on or off.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Simply enabling the bits is not enough, we need to provide axis
information too if we want to enable this properly.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
libevdev has no external dependencies and both check and libudev are
only required for running the unit-tests. Make them optional.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
No real effect on the test, this check is just so that a follow-up patch to
fix the tests for the new uinput backend is a bit easier to review.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Changing a single value on an abs axis is slightly more common than
having to enable that axis outright. Provide a set of accessors for
doing so.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Deprecated:
* libevdev_get_abs_min, libevdev_get_abs_max
* libevdev_get_input_prop_name
Will be removed in one or two versions.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Just enabling EV_REP sets them to zero, but when enabling them directly,
a value is required.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
Except for the API documentation, that'll stay on github for now until
I figure out where to put it (and have the space to put it in).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The helper function uses abs.value for the axis code (to save a parameter).
Reset that to 0 before we write to the device
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
These aren't really that useful since they're just one or two lines in
code, but it saves callers from accidentally misplacing braces, etc.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
A caller may start syncing but switch back to normal half-way through the
sync. In that case, we need to drop all sync events and continue with
regular events only.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Not all clients need nonblocking read, so add a flag to read
in blocking mode. In that mode, events are only read from the fd when
the queue is empty.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
libevdev needs to maintain the correct state of the device, even if the caller
decides to drop all delta events after a SYN_DROPPED.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We can't enable axes on devices at runtime in the kernel, not even
for uinput devices. So this API can't work anyway, remove it before
someone thinks it does work.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This test fails because we can't actually enable bits at runtime, guess I
should've thought of that before. Either way, commit this test (and revert it
again) so the test is in the repository in case we do get this ability later.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
When a sync is complete and the first event is a SYN_DROPPED again,
we have a problem. Log that so that the process can print a warning.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
uinput doesn't route a SYN_DROPPED, so we neeed to handle fds in
a rather complicated manner. For tests, avoid this and instead force
the library to sync if a specific flag is given.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Make check will simply overwrite results of a previous test. Hook up gcov so that the test has to be
explicitly passed to avoid confusion here.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Doxygen is bad for man pages, but having an online API documentation
is better and easier to maintain.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Return 0 for an unallocated queue, and return the actual number of events (rather than one too
little).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This avoids messing around with libtool for debugging. I may revert this
once the library is actually slightly stable, but for now this makes things
easier.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
These were hidden due to the include order from the .c files, but really,
libevdev-int should include them.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If a client doesn't care about syncing that's one thing, but
libevdev functions still need to return the right _current_ (i.e. synced)
value.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Enum values fo 3 and 4 to avoid callers using true/false values for grabbing
and ungrabbing. If the evdev mute ioctl is merged, we can extend this
without having to worry about breaking callers.
[1] http://lwn.net/Articles/476830/
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
libevdev_new() can only fail on allocation failures, but with an fd the
failure options are more. Split into a new function that can return an error
code.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Ideally, we could sync by pre-pending all the sync events and then pretend
nothing happened but our queue is too small and likely too full for any
extra events. So drop all events, then add the sync events to the queue.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Provide a couple of handler functions to avoid erroneous queue updates. This
will make it easier if we need to swap queue implementation later.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Callbacks looked good on paper, but synaptics ran into an issue already that
it just couldn't easily pass around the state needed in the actual event
processing function.
Replace with a new interface that only returns the next event (still reading
more off the fd while doing so).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Two main goals of this library:
- 'transparently' handle SYN_DROPPED events
- avoid errors in ioctl handling by providing a simpler interface.
Keeps a cached copy of the device for quick querying.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>