Instead of requiring to create uterm-screens, it is now possible to blit
and fill displays directly similar to fake-blends.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If we deactivate a session, we must not activate the next session we find
but we need to look for an enabled session instead.
This also changes the session_deactivate() helper to allow deactivating
disabled sessions. This shouldn't be relyed on, but it is a safe fallback
if some session turns out to be disabled but still active.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We use this naming-scheme for all the session-helpers so we should also
use it for the callback notifications.
Also rename all the callback-helpers to *_call_* so we have no
name-clashes.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If we want to allow dummy sessions, we need some way to disable these so
they are never activated on a seat. This patch disables every session by
default and you need to explicitely enable them to be available for
activation.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This helper allows to check whether a given session is currently the
active session on its registered seat.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This helper allows API users to check whether a session object is still
registered to a seat.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We currently do not handle modifier-changes during key-repeats. This is
odd as pressing shift should change a repeating key.
To avoid duplicating a lot of code, this patch puts most of the
key-handling into helper functions and cleans it up. We now handle all
kinds of key-repeat specialties and everything should work fine.
Reported-by: Ran Benity <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
A key-repeat is about keycodes, not keysyms. We repeat the keycode and the
events that are generated by it, so we must also stop repeating if the
given keycode is released.
This fixes a bug where a key would still be repeated if we press:
<a-down> - <Shift-down> - <a-up> - <Shift-up>
Reported-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
The XKB functions actually return <0 if the modifier is invalid. This
cannot happen for the hard-coded modifiers that we use, but it is still
safer to check correctly whether the returned value is >0.
Reported-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Even though I really like the high repeat-rates, it might not be the best
default option. Hence, change it to some sane default.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
It isn't clear from the xterm documentation which buffer to clear when
resetting private mode 1047. However, testing showed that the
alternate-screen is cleared, not the main-screen.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
The alternate screen is mainly used to avoid linking lines to the
scrollback buffer. This implements this feature so only main-screen lines
end up in the scrollback buffer.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We correctly free all new space and this is really needed due to rotations
in the active screen and scrollback buffer.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We didn't correctly initialize new lines on resize. In fact, we didn't
initialize new cells at all. This was all covered by a nasty-fix in the
screen-resize handler.
This fixes the line-resize handler to initialize all new cells correctly.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We have to free both, main and alternate buffers on shutdown. We only used
to free the currently active buffer.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
On DECSET 47, 1047-1049 we switch between normal screen mode and alternate
screen mode. xterm and friends support a titeInhibit resource that
disables this feature, so we add a flag to set this, too. However, there
is currently no way to set this on the command-line. If someone needs
that, we can easily add that.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
The VTE layer can now set the TSM_SCREEN_ALTERNATE flag to switch to the
alternate screen-buffer and back.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This changes the line-allocation helpers of TSM-screen to allocate an
alternate buffer along with the main-buffer. We still keep the "lines"
pointer to the main-buffer all the time, but this will allow to implement
alternate-xterm-screens later.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
On seats other than seat0 we do not have any session-management, because
VTs are not available. Furthermore, if we want to get rid of CONFIG_VT
entirely, we also need to provide session-management for seat0.
This commit introduces sessions. Every seat (seats are now managed in
kmscon_seat.c) can have registered sessions. One of the sessions is active
and gets control over all displays. Session switching is entirely handled
inside of kmscon so there is always an active session (except if no
session is registered at all).
This also reworks the seat-management. kmscon_main.c now only manages the
seat allocation/deallocation and video-objects. The seat itself is handled
inside of kmscon_seat.c and does not know of uterm_video objects. Instead,
it is assigned a list of displays that it can use. Everything is still
hotplugging capable so user-experience should be the same as before.
The kmscon_terminal layer is reworked to be session based. So every
terminal is now a single session. By default, a single terminal-session is
created for each seat. This may be changed, though.
There is currently no input-control to change between session with
keyboard hotkeys. However, this will be added when we have more than one
session.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
In some circumstances it might not be desirable to run ./configure
together with ./autogen.sh. You can now set NOCONFIGURE=1 to avoid this.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We didn't send these events as the fake-display for fbdev video-objects
was created during video-creation and thus no listener could have been
registered so far.
However, this caused every video listener to iterate over all displays on
wake-up. This is really ugly and we want to avoid that. Therefore, we now
send the UTERM_NEW event in an idle-handler and the GONE event on
shutdown.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Configuration handling is totally independent from kmscon_main.c so we
move it into a separate file to avoid cluttering up kmscon_main.c.
This also does some basic initialization in kmscon_conf.c which is solely
related to configuration handling and logging.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If the caller didn't pass a callback to be called on HUP, we now
automatically reopen the terminal as default behavior.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Everything in this header is solely related to configuration parsing so we
should name it properly. We will also move the configuration-handling from
kmscon_main.c to kmscon_conf.c in some of the next commits to clean up
kmscon-main.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
The variable xkbcommon_enabled was no longer nedded
after removing xkbcommon from the build configuration notice.
Signed-off-by: Detlef Riekenberg <wine.dev@web.de>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We need to pass the +8 shifted codes instead of the real hardware codes as
XKB needs these for X11 legacy reasons.
Reported-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We removed a lot of helpers from uterm-input because XKB is now a
mandatory dependency. So use it directly from test_input to avoid all the
uterm helpers.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We will be adding more applications to this repository, so to keep better
overview, we prefix kmscon sources with kmscon_*.
The only sources that have no prefix are either shared between
applications (i.e., statically linked) or they were not cleaned up, yet.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
To allow users to specify key-repeat rates/delays, we now implement
software key-repeat. This is mostly copied from wlt_toolkit.c which
already does this.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If we switch to a VT that has KBMODE set to K_OFF, we do not set it back
to K_OFF when leaving, instead, we set it to K_UNICODE. This allows
recovering when kmscon died by simply restarting kmscon.
There is really no need to let a VT stay in K_OFF! This causes the user to
be stuck at this VT and use the sysrq (if enabled) keys to recover.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
shl_dup() duplicates a memory region similar to strdup() but without
reading the data length via strlen().
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If a user has multiple active XKB layouts but only one of them has ASCII
keysyms on the base level, then ctrl+<XY> might actually never work,
because these keys aren't available in the current layout. This patch
tries to find a layout of the user that actually _has_ ascii keysyms on
the base level and passes this information along with the normal keysym
information.
The TSM layer can now use this ascii keysym instead of the normal unicode
keysym to handle ctrl+<XY> shortcuts. This is the same way xterm et. al.
handle this, so it seems to be a good idea to do this in TSM, too.
Reported (and mainly written) by Ran Benita.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
xkbcommon supports reporting multiple keysyms per key-event. There is no
keymap which uses this, yet. However, this feature is not meant to enhance
performance by reducing the number of calls into xkb, but instead multiple
keysyms are to be handled as one big keysym.
But there are no examples, yet, so we cannot actually perform any
conversions on them. But we can add the infrastructure for it.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This changes the uterm-input infrastructure to use XKB directly instead of
using a modularized infrastructure. There is no need to use something else
anymore.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We made xkbcommon mandatory some time ago and there is no reason to keep
this plain backend around anymore. It isn't tested at all and provides no
real advantage over xkb.
Even for debugging it is easier to use XKB.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If multiple handlers are called on the same input-event, we must notify
handlers whether the event was already handled by a previous callback. We
push this decision to the handlers by allowing them to modify the
"handled" flag for an input event.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We should use get_stride() if this check fails as this is the new
function. We use this check only for backwards-compatibility reasons but
should not require it to pass for new builds.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
AM_PROG_AR is a new feature since autoconf 1.12.
autogen.sh failed on Ubuntu 11.10 without this patch with:
configure.ac:30: error: possibly undefined macro: AM_PROG_AR
If this token and others are legitimate, please use m4_pattern_allow.
See the Autoconf documentation.
autoreconf: /usr/bin/autoconf failed with exit status: 1
AM_PROG_AR is required only for autoconf-1.12 so we can just skip it if
using an autoconf version that does not include it.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Current warning without the patch:
src/genunifont.c: In function 'print_data_row':
src/genunifont.c:85:3: warning: format not a string literal and no format arguments [-Wformat-security]
src/genunifont.c:88:3: warning: format not a string literal and no format arguments [-Wformat-security]
We use fputs() to avoid any format-string parsing and instead directly
write the string into the file.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
When adding new elements, we should add them at the tail, not at the
front. And when removing elements, we should remove them from the tail,
too.
We also convert the whole stuff to the shl_dlist API so we do not have to
manage the linked-list ourself.
Note that this silently breaks the idea of having multiple listeners with
the same function+data in the hook. This is because removing the listener
may now change order of two identical entries, as we don't know which of
them to remove. That means, when adding two identical entries, you cannot
rely on them to retain their position in regard to each other.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
These macros do the same as the already available macros but in reversed
order. This is useful when list-order is important.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We now set KBMODE to K_OFF so we are totally independent of kernel input.
Instead, we handle VT switches with uterm-input now. This also allows us
to have full control of which keyboard input is parsed by us and which is
parsed by the kernel.
We still need to set a flag for uterm-input events that they were handled
to avoid having the TSM layer handle these events again.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This modifies the generic VT layer to register input-cbs for all VT types
and dispatch the event to the correct handler.
This will allow us to handle VT-switches for real VTs ourself instead of
relying on VT input.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This whole concept was broken from the beginning. With hotkey based
activation we have a much better debugging tool. This is still very
fragile, but better than nothing. And we are doing pretty well in error
recovery during hijacked VT switches so that's not as problematic as one
might think.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This adds a lot more verbose error messages to the whole real-VT handling
so we can debug this fragile system way better than before.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Recent systemd was updated to parse XDG_SEAT in PAM so we can assign
logins to the correct seat.
This patch allows pty users to specify what seat they run on so the PTY
can correctly set the seat variable. If no seat is specified, then
XDG_SEAT is not set so we still allow non-seated logins.
Note that if kmscon_pty is run with XDG_SEAT set, this will also be set
for the client PTY so unset it if you don't want the environment to be
copied to the client (like any environment variable).
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Compiling with -Wl,--as-needed can cause the check for
gbm_bo_get_stride() to wrongly fail. Sanitize the environment further,
assuring that unneeded compiler flags are not involved in the conftest.
Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
You can now use --xkb-repeat-rate/delay to configure the Xkb key-repeat
settings instead of using the default 25/250.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
