Documentation fix, refer to correct function

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

Fix a typo in the documentation

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

build: set aux-dir

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>

Drop some leftover deprecated constants.

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

test-compile-pedantic: replace -Wpedantic with -pedantic

-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>

doxygen: specifying builddir is not required

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>

doxygen: fix 'dist' when the html docs have not yet been built

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>

doxygen: fix rebuild when not necessary

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>

doxygen: fix error msg: 'find: `html': No such file or directory'

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>

doxygen: remove redundant "dist_noinst_DATA = libevdev.doxygen.in" statement

All .in files in AC_CONFIG_FILES are distributed.

Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

doxygen: implement silent rules in the makefile

Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

doxygen: invocation of doxygen fails due to hard coded path

The makefile does not use $(DOXYGEN) variable.

Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

doxygen: libevdev.doxygen must not be distributed

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>

test: add a build test for -Wpedantic

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>

Drop semicolons after getter/setter macros

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>

Drop trailing comma from last element in the enums

If a caller is compiled with -pedantic, this causes a few complaints

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

Drop deprecated functions

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

libevdev 0.5

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

python: request version 2.6 or higher

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>

configure.ac: remove redundant AC_SUBST(PYTHON)

This is already done by AM_PATH_PYTHON.

Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

configure.ac: AC_PROG_CC_C99: use C99 compiler

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>

configure.ac: add copyright statement

Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

configure.ac: AC_INIT: add bug report url

Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

autogen.sh: remove unneeded m4 configuration workaround

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>

autogen.sh: fix broken out-of-source tree building

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>

Review/update .gitignore files

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>

Drop gcov flags from main libevdev sources

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>

build: check for build/link flags harder

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>

Update COPYING file.

The file must reflect actual Copyright statements in the source code.

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

build: set AC_CONFIG_SRCDIR

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>

Provide separate lookup functions for zero-terminated strings

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

Add name-resolver unit tests

A bunch of tests for the new name resolver.

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>

Add libevdev_event_type/code_from_name() resolvers

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>

Create event type/code lookup tables

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>

Add startswith() helper

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>

Fix wrong libevdev clock test

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>

Require autoconf 2.62

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>

Check max to see if an event type is valid

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>

Make sure EV_SYN is always set

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>

Disallow disabling EV_SYN event codes

The documentation already says that, make it happen.

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

Reset the struct on set_fd

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>

Revert "Check for linux/input.h during configure"

We now ship our own linux/input.h, so no need to have the header around locally.

This reverts commit fcf80ba3713b9d92480dc8be9ec4d88d95586ec0.

Document backwards-compatibility behaviour

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

Work around missing EVIOCGMTSLOTS ioctl

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

Make uinput-code backwards-compatible to missing UI_SET_PROPBIT ioctl

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

When running against a kernel without properties, continue as usual

Missing out on properties is not fatal.

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

Pass linux/input.h as argument to make-event-names.py

Rather than a hardcoded path inside the source.

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

Add a copy of linux/input.h

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>

configure: test for python and python-argparse

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

Don't allow a negative fd in libevdev_set_fd

Everything else responds with -EBADF, let's do so here as well.

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

Allow -1 as valid fd in libevdev_change_fd

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>

Check for linux/input.h during configure

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

Uppercase #define LIBEVDEV_INT_H

over-eager search/replace

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

Note that libevdev_change_fd requires a LIBEVDEV_FORCE_SYNC

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

Mark three deprecated functions as such

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>

Use AC_MSG_RESULT to print the few configuration options

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

Disable gcov by default

Having libevdev build with profiling enabled by default
leaves profiling files around. Profiling is something that only
libevdev developers should need, so let them enable it
case-by-case.

https://bugzilla.redhat.com/show_bug.cgi?id=1012180

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

Avoid shipping generated documentation twice

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>

Add extern C guards to libevdev-uinput.h

Reported-by: Stanisław Halik <sthalik@misaki.pl>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Capitalize uinput header guards

victim of some overly vicious search/replace

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

libevdev 0.4

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

Print an error on an invalid log priority

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

Merge branch 'clockid'

tools: make to function calls static

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

Fix a typo in the uinput documentation

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

Move libevdev_led_value documentation to the kernel group

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

tools: Honor GCC_CFLAGS

So we get compiler warnings from tools/ too.

Signed-off-by: Colin Walters <walters@verbum.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Drop duplicate compiler flag

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

Revert "Reintroduce -fstack-protector"

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 f5e65ea3ce2541fe8ccfafe9b0dd04325da75b34.

Reintroduce -fstack-protector

We removed this previously due to build failures in:

  commit 14ac764ec86452ca607403f314b0f8355d80290c
  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>

Test for CFLAGS/LFLAGS availability

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>

wrap EVIOCSCLOCKID into an API call

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

test: provide wrapper for fetching the devnode from a uinput test device

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

test: add a test for checking log data being passed around correctly

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

Fix a comment to reflect the new API

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

Remove warning comment from FORCE_SYNC

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>

Name-space the read flags better

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>

Enumerate libevdev_next_event() return codes

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>

Require check 0.9.9

0.9.9 is the first released version with fixed ck_assert_* macros that don't do
multiple expansion.

http://sourceforge.net/p/check/code/596/

Reported-by: Martin Minarik <minarik11@student.fiit.stuba.sk>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Fix comment typo

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

doc: don't show enum values in doxygen

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

Add missing group assignment to a bunch of functions and enumerations

Without these, they won't show up in the documentation.

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

Revamp the API once again

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>

build: remove -fstack-protector

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 43752ec17d09b132621a86f5cbc546ca6ab9e977
  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>

configure: add missing comma

Without that, it becomes impossible to build without gcov.

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

Support GNOME Build API

By adding an autogen.sh file with NOCONFIGURE support.

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

Merge branch 'logging-fix'

test: test for logging function

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

Log to stderr by default

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>

Fix comment typo

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

Match the kernel define for SYN_MAX

Will be defined as 0xf in 3.12, see
http://git.kernel.org/cgit/linux/kernel/git/dtor/input.git/commit/?h=next&id=52764fed5049655926bcecaefd52f0a415ceb105

And add the required ifdef guards for kernels before that.

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

Check for GNU ld and use the flags depending on the outcome

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

Warn if there are multiple devices with the same syspath

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>

Log a few errors, specifically application bugs

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

Drop per-device logging function, use per-library one instead

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>

Remove -g from default GCC_CFLAGS

-g should be set by debugging-options or in the default CFLAGS="", we
shouldn't force it in GCC_CFLAGS.

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

Add some gcc/ld optimizations and magic

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>

uinput: don't try to set the syspath twice

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>

Return -EBADF for functions that need the fd initialized

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

Return EBADF when trying to read from an uninitalized device

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>