The headers <stdio.h> and <unistd.h> were included twice.
Signed-off-by: Johan Korsnes <johan.korsnes@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
There main reason for this change is that versionsort is not widely
supported at the moment (for example, it is not present in FreeBSD and
the manual page for GNU C library scandir documents versionsort as GNU
extension).
It's good enough for our purpose and it is more portable. Also, because
argv[0] is not globally accessible, let's define a new global variable
for that called progname.
This makes evtest more portable. We could try to use ifdefs to include
sys/syslimits.h on FreeBSD to get PATH_MAX but it feels to be an
overkill for evtest. Let's keep it simple as just replace PATH_MAX with
4096.
NB: it's fine to leave linux/input.h there, because this header is
actually available on FreeBSD via the devel/evdev-proto port.
It's not necessary to install signal handlers for SIGINT and SIGTERM
when not in grab mode, as there is nothing for evtest to do when
exiting, such as releasing the grab.
This causes problems with umockdev's event capture which relies on being
able to catch the SIGINT signal itself to know to flush its capture
buffers before exiting.
See https://github.com/martinpitt/umockdev/issues/96
Fixes build with musl libc that does not include limits.h indirectly via
other headers.
evtest.c: In function ‘scan_devices’:
evtest.c:886:14: error: ‘PATH_MAX’ undeclared (first use in this function); did you mean INT8_MAX’?
char fname[PATH_MAX];
^~~~~~~~
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
evtest.c:887:9: warning: ‘%s’ directive output may be truncated writing up to 255 bytes into a region of size 53 [-Wformat-truncation=]
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The functions are often declared as "warn unused result", which causes
compiler time warnings. Invalid user input may also lead to not entirely
correct utility behavior.
Signed-off-by: Dmitry Torokhov <dtor@chromium.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Add more key definitions that were added to newer kernels, but were
missing in evtest.
Signed-off-by: Dmitry Torokhov <dtor@chromium.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The newer "Accelerometer" input device property was missing in evtest,
let's add it.
Signed-off-by: Dmitry Torokhov <dtor@chromium.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Kernel recently got a new key definition for desktop assistant key,
let's plumb it through evtest as well.
Signed-off-by: Dmitry Torokhov <dtor@chromium.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Intentionally not printing keys here - led/snd/sw are long-term states, keys
are short term states so there's a chance that by the time the user looks at
the output, the key has already changed anyway.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Tested-by: Bastien Nocera <hadess@hadess.net>
We had the wrong pointer in the names[] array for EV_SYN, so for example
codename() would print "EV_SYN" instead of "SYN_REPORT", or "EV_ABS"
instead of "SYN_DROPPED", etc.
Output from 1.31:
Event: time 1425470463.069381, -------------- SYN_REPORT ------------
Output from 1.32:
Event: time 1425470948.149794, -------------- EV_SYN ------------
Output after this patch:
Event: time 1425471209.159696, -------------- SYN_REPORT ------------
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
For easy experimentation with different buffer sizes, it is more
convenient to specify the array size to read() with sizeof(ev) instead
of having the array element count in two places.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If we fail to read the input events fast enough, recent kernels give the
SYN_DROPPED event. Highlight the event to make it obvious that we missed
some events.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If the --grab flag is given in capture mode, evtest keeps an EVIOCGRAB
on the device. While this grab is active, other processes will not
receive events from the kernel devices. The grab is released again
when evtest quits.
TEST=Check that the cursor is frozen when --grab option is specified.
$ evtest --grab
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The current code simply counts the number of directory entries and takes
that as the maximum event node given. That isn't true if we have a hole in the
event range after removing a device, making the last device in the list
unaccessible.
Fix this by actually looking at the event node numbers.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
qiurui
Thomas Kuehne
Gergo Koteles
Sicelo A. Mhlongo
Harry Cutts
Harry Cutts
Harry Cutts
Johan Korsnes
Mateusz Piotrowski
Peter Hutterer
Bastien Nocera
Petr Vorel
Khem Raj
Leo
Baruch Siach
Dmitry Torokhov
Tommi Rantala
Joseph Hwang