Now that this driver works on most systems (hopefully) without any configuration,
it makes sense to use it as a fallback default.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
We can't free driverNames[i], because it's part of abi,
and other drivers might give us statically-allocated strings.
Fixes: #1848
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
When starting an X server from a terminal running
inside another X server, the "host" X server sees
the key press event, but not the key release event,
and misinterprets this as a long keypress.
With this patch, we forge the missing key release event,
so the server doesn't think that the user is holding a key down.
This is different from Ctrl + Alt + F* vt switching, because
there we don't want the X server to see the key press event at all,
and we want to forge key release events for the Ctrl + Alt keys.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
Upcoming work needs a few fixes in the Intel driver.
(mainly not using internal EDID parser defines anymore)
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
It will take some time for a new libpciaccess to be released and
even then bumping the dependency for libpciaccess isn't attractive.
If an older libpciaccess is used just add a static inline define.
Signed-off-by: Tautvis <gtautvis@gmail.com>
In the case of multiple display devices that are not VGA devices
the 'boot_display' attribute read by libpciaccess can disambiguate.
Signed-off-by: Tautvis <gtautvis@gmail.com>
When rapidly switching vt's, it can happen that shadow tries to
draw to the screen before it's initialized.
In that case, we return NULL, and we should also return a zero size,
because shadow doesn't check for NULL.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
os.h still defining our own timingsafe_memcmp() prototype when this
symbol isn't set - this is causing trouble with drivers on FreeBSD.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Zapping is the fastest way to kill the X server, faster than vt switching,
or killing the window manager and configuring the X server to die with it.
It is very useful when debugging, or if the system runs very low on memory
This adds the `-nozap` argument, for restoring the old behavior.
The `-zap` argument is also kept for backwards compatibility.
Signed-off-by: stefan11111 <stefan11111@shitposting.expert>
Allow the parseLine() function to use tabs alongside spaces to separate
tokens. Without this patch, using tabs in the config file actually
confuses the parser and makes it think that the tab is part of the
option's name.
Signed-off-by: Aggelos Tselios <aggelostselios777@gmail.com>
The previous comment made no sense. Add a proper comment to document the
function instead.
Signed-off-by: Aggelos Tselios <aggelostselios777@gmail.com>
Our naming convention for headers is that private ones (not exported,
ie. not part of xf86 sdk) shall be suffixed by "_priv.h".
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
This functions is designed for some legacy keyboard drivers
(eg. xf86-input-keyboard) that need to get the fd to the console
device, so they don't need to directly access xf86Info field anymore.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
