The name fork_pty_child was misleading. It doesn't fork at all but
instead sets up the child's environment.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This ioctl is really not needed and performs needles kernel context
switches. We can simply read into our buffer without checking how many
data is available.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If the child returns EWOULDBLOCK on write we need to save the input in a
buffer to avoid loosing data. We need to work in non-blocking mode to
avoid UI hangs so we simply use the new ring-buffer object to store
data.
This also changes the callback behavior. The pty is no longer closed
implicitely so the owner must call pty_close now even if the close
callback is called. This avoids circular callbacks.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
There is no reason to keep two callbacks as the caller always registers
both. Hence, we can use a shared callback. Reading length 0 means closed
like reading from an fd.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
There is no reason to pass the eloop reference late at terminal/pty
open. If we pass it early on object creation we will always have a valid
reference and can remove several code paths.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
The output_cb handles output from the pty but it actually makes more
sense to call it input_cb as we are handling input to our console here.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
pty_write actually makes more sense as we are writing to the other end
and not reading input.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
When we register signals in the eloop we also block them in our signal
mask. The signal mask is inherited by the child. Therefore, if the child
does not reset its mask (e.g. bash and most normal processes), it will
not receive any of the signals that we handle. So for example C-c
(SIGINT) does nothing in the child process.
We now unblock all signals before we exec the child.
It's also worth noting that if we _ignore_ a signal -
sigaction(SIG_IGN) - this is also inherited and we must reset it to
default. However, we do not ignore signals so this is unneeded.
Here is some more discussion on signalfd and this problem:
https://lwn.net/Articles/415684/
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We recently moved log_warning to log_warn and I forgot to fix it when
merging the kbd-dumb backend. Fix this now.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
The paths can be off when building out of tree, so have make put them in
for us instead. This requires turning the straight genshader.c file to a
template.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This commit adds a very simple keyboard backend which does basic keycode
interpretation. It is used as a fallback when xkbcommon is unavailable
or is not required.
See the file header for what is supported.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Currently it's required by xkbcommon. However, we will need it for other
backends as well, even with xkbcommon support disabled.
xproto consisted only of static header files. We use the XK_ keysym
definitions. It is therefore only required during compilation and we
shouldn't needlessly duplicate it ourselves.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
log_warn is much shorter and we already use log_err instead of log_error
so this is more consistent now.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
GL_BGRA_EXT as internal format is not supported with full OpenGL so we
must use GL_RGBA as internal format.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
These dispatch debug messages are currently no longer needed and just
fill the log needlessly.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Unfortunately, there is no clean way I see to hook this up from the pty
object. We can (and will) have more than one pty object opened at a
time, but the semantics of signalfd make it impossible to deliver each
SIGCHLD to its rightful owner without complicating things.
[ From what I tested:
- If you have two signalfd's listening to the same signal, they will be
dispatched in some round-robin manner.
- Also, if more than one child exits before we read signalfd (possibly
beloging to different terminals), they will be compressed to one
event. ]
We therefore need to do the reaping from a central location, and need to
remember to copy this snippet over to main.c in the future.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Add a new terminal_open/terminal_close methods to correspond to the pty
ones, and notify when the terminal is closed.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This commit adds a new pty object.
The pty object takes care of all pseudo terminal handling, reading and
writing. It can be opened and closed, and notify through callbacks when
input arrives or the child process exits/dies. It can also receive input
and pass it along to the child process.
There is not yet any real VTE processing, so we display raw escape
codes and so on. However, this should provide immediate feedback for
any further vte development, as we start to act like a real terminal
emulator.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This commit ports the XKB handling to the new kbd interface, and makes
the input subsystem use it without any direct knowledge of XKB.
Some code is moved around but there are no functional changes.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This defines the API the keyboard backends will need to implement. It is
based on what's in input_xkb.h but cleaned up and simplified.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We now properly draw fonts with OpenGL. We now use FreeType2 instead of
pango to avoid big dependencies.
We also add a DejaVu font so we currently don't have to deal with font
selection.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
To avoid cairo dependencies we now take a compositor reference in the
font backend so fonts can be drawn with GL textures instead of cairo.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
OpenGL pulls in lot of X dependencies on linux. To avoid this we also
support OpenGLES2 now. This will also allow to run kmscon on
embedded/mobile platforms.
We still get X dependencies through EGL which we cannot avoid as EGL is
compiled with multiple backends on most systems. However, switching to
GLES2 reduces memory footprint by 30MB which is pretty good.
This also enables GLES2 as default in autogen.sh. However, default
behaviour of configure will still be GL.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We should first create a GL context before we try to find GL functions
to avoid errors on some platforms.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We should use the context of the compositor instead of directly calling
the GL functions to avoid direct dependencies.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We have no use for Control_L or Alt_R by themselves, the information is
in the "mods" field.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Instead of drawing with fixed-function GL pipeline we now use our own
shader for texture drawing. This also fixes test_console to no longer
depend on GL.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
We use a 4x4 matrix so we should use the correct uniform handler.
Otherwise we might get errors on some GL implementations.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
To switch to the new drawing subsystem we need a reference to a valid
compositor object so we can retrieve the GL context.
This also applies to the terminal object.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This converts the output test function to use the new kmscon_context for
drawing operations so we no longer depend directly on GL.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Besides the texture shader this also adds a second shader (def/default)
which is used to draw arbitrary colored vertices to the screen.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
