platform/upstream/kmscon.git
12 years agovte: reset margins on soft-reset
David Herrmann [Sat, 14 Jul 2012 11:43:02 +0000 (13:43 +0200)]
vte: reset margins on soft-reset

Instruct console code to reset the console margins to default values (that
is, destroy the margins) on console soft-reset.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: implement DECSC DECRC
David Herrmann [Sat, 14 Jul 2012 11:39:43 +0000 (13:39 +0200)]
vte: implement DECSC DECRC

DECSC is used to save the current console state. DECRC can restore the
console to a previously saved state.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconsole: allow retrieving cursor state
David Herrmann [Sat, 14 Jul 2012 11:36:46 +0000 (13:36 +0200)]
console: allow retrieving cursor state

This adds two functions to retrieve x/y positions of the cursor. This can
be used to save and restore cursor positions.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoui: reopen terminal on HUP
David Herrmann [Sat, 14 Jul 2012 10:15:20 +0000 (12:15 +0200)]
ui: reopen terminal on HUP

We now automatically reopen the pty when the child process died. This will
most-often simply restart the shell or login program.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal: close pty before opening it
David Herrmann [Sat, 14 Jul 2012 10:14:25 +0000 (12:14 +0200)]
terminal: close pty before opening it

There is no explicit call to reopen the underlying pty so we simply
force-close it before (re)-opening it. This will allow the UI to reopen a
pty when the child died.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoxkb: fix null dereference
Ran Benita [Thu, 5 Jul 2012 10:16:40 +0000 (13:16 +0300)]
xkb: fix null dereference

Just an obvious silly mistake.

Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotests: fix tests to compile again
David Herrmann [Sat, 7 Jul 2012 11:45:29 +0000 (13:45 +0200)]
tests: fix tests to compile again

Whoops, I totally forgot about them. Fix the headers to be compatible with
eloop again.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoRelease kmscon-2
David Herrmann [Sun, 1 Jul 2012 18:29:12 +0000 (20:29 +0200)]
Release kmscon-2

This is the second release of kmscon. It's again a development release but
should now be good enough so it can run on any linux system.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoUpdate README
David Herrmann [Sun, 1 Jul 2012 18:26:30 +0000 (20:26 +0200)]
Update README

Fix the dependency information and add several TODO notes.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agounicode: replace glib types with native types
David Herrmann [Sun, 1 Jul 2012 17:35:16 +0000 (19:35 +0200)]
unicode: replace glib types with native types

There is really no need to use these glib types here. They are relicts
from the time when we used the glib hashtable so we can safely replace
them now.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agomisc: change hashtable compare function to return bool
David Herrmann [Sun, 1 Jul 2012 17:29:33 +0000 (19:29 +0200)]
misc: change hashtable compare function to return bool

It really doesn't make sense to expect an integer here as memcmp() and
friends use an integer, too, but in reversed logic. So use bool to avoid
confusion.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agomisc: remove glib dependency
David Herrmann [Sun, 1 Jul 2012 17:13:16 +0000 (19:13 +0200)]
misc: remove glib dependency

This adds a new drop-in replacement for the glib hashtable implementation.
The hash table can be found at github: github.com/rustyrussel/ccan

It's licensed under the terms of the LGPL and perfect for our
implementation. Please see COPYING for license information.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: monitor: make systemd optional
David Herrmann [Sun, 1 Jul 2012 14:43:09 +0000 (16:43 +0200)]
uterm: monitor: make systemd optional

If systemd is not used for multi-seat support, we simply fake-add seat
"seat0" and avoid looking for the "seat" tag. However, if devices still
use ID_SEAT, we still parse it but all other devices are taken for seat0
by default.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoAdd NEWS file
David Herrmann [Sun, 1 Jul 2012 14:31:54 +0000 (16:31 +0200)]
Add NEWS file

Whenever we do a release, this file will contain all the major changes as
a short list.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: remove TODO from DIST_EXTRA
David Herrmann [Sun, 1 Jul 2012 14:18:23 +0000 (16:18 +0200)]
build: remove TODO from DIST_EXTRA

The TODO list was moved to github long ago so remove this dead reference.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: fix signal creation
David Herrmann [Sun, 1 Jul 2012 14:17:04 +0000 (16:17 +0200)]
eloop: fix signal creation

The previous fix incorrectly registered new signals always to the last
found signal which is definitely incorrect. Therefore, we now correctly
traverse the list and register new signals as new signals.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: vt: correctly initialize ret to 0
David Herrmann [Sun, 1 Jul 2012 14:11:10 +0000 (16:11 +0200)]
uterm: vt: correctly initialize ret to 0

We need to return "true", not "false" if a mode is unknown. Otherwise, we
will block the VT subsystem.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: correctly clear GBM/EGL flags
David Herrmann [Sun, 1 Jul 2012 14:09:24 +0000 (16:09 +0200)]
build: correctly clear GBM/EGL flags

When building without DRM we should correctly clear the GBM and EGL flags.
Otherwise, we will be linking them either way.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: silence gcc warning
David Herrmann [Sun, 1 Jul 2012 13:49:36 +0000 (15:49 +0200)]
eloop: silence gcc warning

Gcc warns about an uninitialized variable (which technically is correct).
However, there is really no way this can really happen. But to make gcc
happy (and to enhance code readability) we now depend on !sig whether the
list-search was successful.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agollog: make debug statements produce real code
David Herrmann [Sun, 1 Jul 2012 13:47:11 +0000 (15:47 +0200)]
llog: make debug statements produce real code

Instead of defining them to nothing we now produce real code which
evaluates to "void", though. This allows us to use the macros whereever an
rvalue can be used.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: vt: silence gcc warning
David Herrmann [Sun, 1 Jul 2012 13:45:54 +0000 (15:45 +0200)]
uterm: vt: silence gcc warning

We initialize the variable to "true" to silence gcc warnings that we might
get an uninitialized variable (which we really cannot get here).

Anyway, in the future, if the number of types may increase, the correct
way to deal with unknown types is returning "true" so do this by default
now.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte/console: implement CSI 'M' (DL: delete line)
David Herrmann [Sun, 1 Jul 2012 13:35:28 +0000 (15:35 +0200)]
vte/console: implement CSI 'M' (DL: delete line)

The 'M' CSI mode is used to delete lines. This implements this mode
similar to the "INSERT LINES" CSI.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte/console: implement CSI mode L (insert line: IL)
David Herrmann [Sun, 1 Jul 2012 13:27:10 +0000 (15:27 +0200)]
vte/console: implement CSI mode L (insert line: IL)

The IL CSI ('L') is used to insert a given number of lines at the current
cursor position. All lines below are moved down.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: make opengl optional
David Herrmann [Sun, 1 Jul 2012 12:59:43 +0000 (14:59 +0200)]
build: make opengl optional

This removes the mode where we compiled uterm with OpenGL instead of
OpenGLES2 (it was broken anyway) and makes it fully optional now.

gl_shader.h is no longer compiled if OpenGL is not available and provides
dummy helpers instead.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agogl: add wrappers for glClear and glClearColor
David Herrmann [Sun, 1 Jul 2012 12:39:30 +0000 (14:39 +0200)]
gl: add wrappers for glClear and glClearColor

This helps us moving opengl dependencies into a single source file so we
can easily disable it.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agomain: use UTERM_VIDEO_DUMB as fallback when DRM is not available
David Herrmann [Sun, 1 Jul 2012 12:31:50 +0000 (14:31 +0200)]
main: use UTERM_VIDEO_DUMB as fallback when DRM is not available

If the uterm video object fails to initialize the DRM device, then we try
again with the dumb device. Only if both devices fail, we drop the device
and continue.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: add dumb-drm backend
David Herrmann [Sun, 1 Jul 2012 12:28:26 +0000 (14:28 +0200)]
uterm: add dumb-drm backend

Nearly all drm drivers provide a GEM/TTM buffer backend called "dumb
buffers". These buffers can simply be mmap'ped like fbdev device so 2D
acceleration is available. This patch adds a backend to uterm-video that
uses these buffers. It is quite similar to the drm backend but removes all
the egl/gl/gles2 dependencies.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: rewrite configure.ac
David Herrmann [Sat, 30 Jun 2012 17:52:20 +0000 (19:52 +0200)]
build: rewrite configure.ac

Remove all obsolete code and make it more readable. Nearly all features
can now be configured. However, the code doesn't work correctly if udev or
systemd is removed. That will be fixed soon, though.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: monitor: refactor systemd code
David Herrmann [Sat, 30 Jun 2012 10:53:23 +0000 (12:53 +0200)]
uterm: monitor: refactor systemd code

Move all systemd dependencies into local helper functions so we can easily
make them optional.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: move systemd dependencies to libuterm
David Herrmann [Fri, 29 Jun 2012 11:00:00 +0000 (13:00 +0200)]
build: move systemd dependencies to libuterm

We use the sd-login library in uterm exclusively so move the dependecy to
libuterm.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: move vt.[hc] to libuterm
David Herrmann [Fri, 29 Jun 2012 10:59:25 +0000 (12:59 +0200)]
build: move vt.[hc] to libuterm

The kmscon VT layer is used in libuterm exclusively so build it as part of
it.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: remove libudev.h includes if not needed
David Herrmann [Fri, 29 Jun 2012 10:49:37 +0000 (12:49 +0200)]
uterm: remove libudev.h includes if not needed

All udev-dependencies are now moved into uterm-monitor. This allows us to
easily make udev optional for hotplug-less systems or emergency setups.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: move uterm into shared library
David Herrmann [Fri, 29 Jun 2012 10:47:54 +0000 (12:47 +0200)]
build: move uterm into shared library

The uterm code was created to allow other applications to easily build new
user-space terminals for linux machines. Therefore, put it into a separate
library and install the header files correctly.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: move eloop into shared library
David Herrmann [Fri, 29 Jun 2012 10:37:02 +0000 (12:37 +0200)]
build: move eloop into shared library

We use the eloop code in several other projects so move it into a shared
library to allow code-sharing.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoxkb: first get keysyms, then update state
Ran Benita [Wed, 27 Jun 2012 10:23:38 +0000 (13:23 +0300)]
xkb: first get keysyms, then update state

This is the correct order, as specified in xkb_state_update_key header
comment.

Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoxkb: only repeat keys when they're supposed to repeat
Ran Benita [Wed, 27 Jun 2012 10:11:23 +0000 (13:11 +0300)]
xkb: only repeat keys when they're supposed to repeat

Support for this was added upstream, we can use it now. This mostly
means that modifiers won't repeat anymore.

Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoxkb: use unicode conversion from xkbcommon
Ran Benita [Wed, 27 Jun 2012 10:08:41 +0000 (13:08 +0300)]
xkb: use unicode conversion from xkbcommon

Support for this was added upstream, so we should avoid keeping the
entire conversion table twice in (readonly) memory. The old files are
still used by the dumb backend.

Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: move external sources into libkmscon-static
David Herrmann [Sun, 24 Jun 2012 19:15:08 +0000 (21:15 +0200)]
build: move external sources into libkmscon-static

Move all external sources into this library so we can use them everyone
but also be sure that we can link them statically. We will need this
guarantee later when adding LGPL stuff to libkmscon-static.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agodoc: fix header of gtk-doc makefile
David Herrmann [Sun, 24 Jun 2012 19:11:57 +0000 (21:11 +0200)]
doc: fix header of gtk-doc makefile

It still used the old name. So convert it to use the more general name as
we are building all documentation into one global document.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: add comments to Makefile.am
David Herrmann [Sun, 24 Jun 2012 19:11:19 +0000 (21:11 +0200)]
build: add comments to Makefile.am

Add more comments to Makefile.am to make it more readable.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: put static_* files into separate library
David Herrmann [Sun, 24 Jun 2012 18:43:34 +0000 (20:43 +0200)]
build: put static_* files into separate library

We now build all small helpers into a separate small library that can be
linked statically into all our others. That is, we don't need to keep
backwards compatibility or anything inside this library.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agostatic: move statically linked subsystems to static_*
David Herrmann [Sun, 24 Jun 2012 18:35:22 +0000 (20:35 +0200)]
static: move statically linked subsystems to static_*

Small stuff that is not worth putting into a separate library is no moved
into the "static" library which is statically linked into all our
subsystems.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: move to llog
David Herrmann [Sun, 24 Jun 2012 16:53:46 +0000 (18:53 +0200)]
eloop: move to llog

Move every use of log_* to llog_* and make all objects aware of llog
objects.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agollop: add default helpers
David Herrmann [Sun, 24 Jun 2012 16:16:12 +0000 (18:16 +0200)]
llop: add default helpers

We often need to return the same message over and over again with the same
return value. Use default log-messages now instead of ignoring the log.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agollog: add direct log helpers
David Herrmann [Sun, 24 Jun 2012 16:15:14 +0000 (18:15 +0200)]
llog: add direct log helpers

Sometimes, the parent object might not be available so we need to pass in
the log-object directly. Use "d" as prefix for those functions and add
variants for the most oftenly used functions.

If you need more direct variants, add them when you introduce the first
user.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: require llog function in eloop-creation
David Herrmann [Sun, 24 Jun 2012 15:57:38 +0000 (17:57 +0200)]
eloop: require llog function in eloop-creation

When creating a new eloop object, we now require an lloop function so we
can perform conditional logging.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: set LLOG_ENABLE_DEBUG
David Herrmann [Sun, 24 Jun 2012 15:55:59 +0000 (17:55 +0200)]
build: set LLOG_ENABLE_DEBUG

Set LLOG_ENABLE_DEBUG if debug mode is used.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agolog: add link to llog
David Herrmann [Sun, 24 Jun 2012 15:55:26 +0000 (17:55 +0200)]
log: add link to llog

Add helper function that can be used as llog backend.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agollog: add library log helpers
David Herrmann [Sun, 24 Jun 2012 15:28:31 +0000 (17:28 +0200)]
llog: add library log helpers

Libraries must never log unconditionally. Otherwise, applications are
unable to disable logging and can get confused. This header contains
helpers to implement conditional logging on context objects inside of
libraries.

It is compatible to the log interface in kmscon apart from the runtime
configuration.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agolog: change severities to be identical to kernel sevs
David Herrmann [Sun, 24 Jun 2012 15:21:56 +0000 (17:21 +0200)]
log: change severities to be identical to kernel sevs

The kernel defines 8 severities. Lets be compatible to these numbers so
our API will not have to be changes in the future.

This also causes LOG_ALERT to be added. It is not used, yet, but may be in
the future.

We also change the parameter type of severities to "unsigned int". Enum
variables can change types if new enums are added. We don't want that so
use a fixed type.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: set AUTO_WRAP mode on reset for console object
David Herrmann [Sun, 24 Jun 2012 10:20:18 +0000 (12:20 +0200)]
vte: set AUTO_WRAP mode on reset for console object

We reset all modes but forget to set them on the console object, too. The
AUTO_WRAP mode is the only mode that matters here, so we can ignore the
other modes.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_drm: set EGL_DISPLAY correctly
David Herrmann [Sun, 24 Jun 2012 09:09:21 +0000 (11:09 +0200)]
uterm: video_drm: set EGL_DISPLAY correctly

We force EGL_DISPLAY to drm now. Otherwise, we might cause segfaults if
the EGL library tries to use the X11 egl-backend.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agomain: use --fbdev options to switch between DRM/fbdev
David Herrmann [Sun, 24 Jun 2012 09:01:25 +0000 (11:01 +0200)]
main: use --fbdev options to switch between DRM/fbdev

The new --fbdev option allows runtime switching between the two backends.
That is, we only use fbdev devices when --fbdev is given. Otherwise, DRM
is used.

Technically, it would also be possible to use both. However, almost every
DRM device does also register an fbdev device for backwards compatibility.
Therefore, we must avoid using both, the DRM and fbdev device of the same
display at the same time. As this would also mean dealing with failures in
one backend and then switching to the other, we avoid this for complexity
reasons. Who needs fbdev and drm simultaneously, anyway?

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal: redraw on wake-up
David Herrmann [Sun, 24 Jun 2012 08:53:04 +0000 (10:53 +0200)]
terminal: redraw on wake-up

The drm video backend does automatically blit the framebuffer on wake-up,
however, the fbdev backend does not. Hence, we automatically redraw on
wake-up in the terminal backend now.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal: avoid adding displays multiple times
David Herrmann [Sun, 24 Jun 2012 08:52:05 +0000 (10:52 +0200)]
terminal: avoid adding displays multiple times

We now enumerate displays on wake-up so we might end up adding a display
twice. This causes the display to be redrawn twice which is nasty so we
avoid it.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoui: do not modify callback parameters
David Herrmann [Sun, 24 Jun 2012 08:50:16 +0000 (10:50 +0200)]
ui: do not modify callback parameters

Other registered callbacks might get confused if we change the parameters.
Hence, we use a static copy for further actions.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agofont: support 2D blitting
David Herrmann [Sun, 24 Jun 2012 08:32:42 +0000 (10:32 +0200)]
font: support 2D blitting

If OpenGL is not supported, we now allow 2D blitting into the video
framebuffer as alternative to shaders.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovideo: return -EOPNOTSUPP in video_use without OpenGL
David Herrmann [Sun, 24 Jun 2012 08:31:34 +0000 (10:31 +0200)]
video: return -EOPNOTSUPP in video_use without OpenGL

If OpenGL is not supported, we should _not_ return 0. Otherwise, there is
not convenient way to detect this scenario.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoui: enumerate displays on wakeup
David Herrmann [Sun, 24 Jun 2012 08:30:28 +0000 (10:30 +0200)]
ui: enumerate displays on wakeup

We might have missed a display when our handler was not registered, yet.
So enumerate these displays on wakeup so we can be sure that we have all
displays registered with the terminal subsystem.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agomain: do not depend on OpenGL
David Herrmann [Sun, 24 Jun 2012 08:28:40 +0000 (10:28 +0200)]
main: do not depend on OpenGL

We currently try activating the OpenGL context when adding a video device
to a seat. However, not all video backends provide OpenGL contexts so
remove this restriction. The terminal subsystem does all this by itself
now.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal: create shader only when OpenGL is available
David Herrmann [Sun, 24 Jun 2012 08:27:46 +0000 (10:27 +0200)]
terminal: create shader only when OpenGL is available

If the video backend does not provide an OpenGL context, there is no need
to create the gl shaders. We then use the 2D blitting functions instead.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal: add debug message when removing display
David Herrmann [Sun, 24 Jun 2012 08:27:07 +0000 (10:27 +0200)]
terminal: add debug message when removing display

This can help a lot when debugging displays that are registered multiple
times with a terminal object.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_fbdev: fix debug statements
David Herrmann [Sat, 23 Jun 2012 23:09:40 +0000 (01:09 +0200)]
uterm: video_fbdev: fix debug statements

Remove blitting debug statements as it is working fine now.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: add internal header for blitting support
David Herrmann [Sat, 23 Jun 2012 23:07:01 +0000 (01:07 +0200)]
uterm: add internal header for blitting support

Whoops, forgot to add it in the fbdev commit. Pushing it now so it works
again.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_drm: set .blit to NULL
David Herrmann [Sat, 23 Jun 2012 23:06:16 +0000 (01:06 +0200)]
uterm: video_drm: set .blit to NULL

We do not support blitting via drm (yet) so set it to NULL explicitely.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_fbdev: make double-buffering optional
David Herrmann [Sat, 23 Jun 2012 23:04:50 +0000 (01:04 +0200)]
uterm: video_fbdev: make double-buffering optional

This makes double-buffering optional as not all drivers provide this. In
fact, only very few drivers implement this. Furthermore, this also adds a
fake display-mode that is used instead of NULL. Otherwise, the screen
object might not work as expected.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video: add blitting support
David Herrmann [Sat, 23 Jun 2012 23:03:56 +0000 (01:03 +0200)]
uterm: video: add blitting support

All backends that do not provide OpenGL contexts can not implement buffer
blitting so we can at least draw rectangular areas to the framebuffer.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: add WAKEUP and SLEEP signals
David Herrmann [Sat, 23 Jun 2012 23:01:47 +0000 (01:01 +0200)]
uterm: add WAKEUP and SLEEP signals

Notify all listeners when going to sleep or waking up. This allows saving
energy in the listeners by not redrawing the screen while being asleep.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_fbdev: save stride on open
David Herrmann [Sat, 23 Jun 2012 20:54:52 +0000 (22:54 +0200)]
uterm: video_fbdev: save stride on open

Copy stride into the fbdev structure so we can use it throughout the code.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_fbdev: test whether bpp is power of 8
David Herrmann [Sat, 23 Jun 2012 20:52:24 +0000 (22:52 +0200)]
uterm: video_fbdev: test whether bpp is power of 8

I have never seen something different but lets be sure here and test for
modulo-8.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_fbdev: remove double check
David Herrmann [Sat, 23 Jun 2012 20:46:42 +0000 (22:46 +0200)]
uterm: video_fbdev: remove double check

There is no need to check this twice so move it beyond the last vinfo
modification.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_fbdev: discard non-xrgb32 devices
David Herrmann [Sat, 23 Jun 2012 20:39:48 +0000 (22:39 +0200)]
uterm: video_fbdev: discard non-xrgb32 devices

We discard all non-xrgb32 devices for simplicity reasons for now. We
definitely need to support other devices in future but that makes the
backend unnecessarily complex at the beginning so avoid it.

This also moves the resolution-change to the beginning and duplicates the
tests so we can be totally sure that everything is right after setting the
correct values.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video: return -EOPNOTSUPP when opengl is not supported
David Herrmann [Sat, 23 Jun 2012 20:04:06 +0000 (22:04 +0200)]
uterm: video: return -EOPNOTSUPP when opengl is not supported

When the backend does not implement display_use() then we return an error
instead of success to notify the application that it should use blitting
instead.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video_fbdev: remove display_use()
David Herrmann [Sat, 23 Jun 2012 19:56:43 +0000 (21:56 +0200)]
uterm: video_fbdev: remove display_use()

fbdev does not support OpenGL contexts so there is no need to implement
this function. Instead, we will provide blitting support later.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: add gtk-doc
David Herrmann [Sat, 23 Jun 2012 18:09:19 +0000 (20:09 +0200)]
build: add gtk-doc

This adds gtk-doc documentation builds to kmscon. All documentation is
build in a central place due to gtk-doc restrictions.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: turn comment into non-gtkdoc comment
David Herrmann [Sat, 23 Jun 2012 18:06:44 +0000 (20:06 +0200)]
eloop: turn comment into non-gtkdoc comment

This comment does not use gtk-doc syntax so remove the extra *.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: add fbdev video backend
David Herrmann [Sat, 23 Jun 2012 16:24:44 +0000 (18:24 +0200)]
uterm: add fbdev video backend

This adds a fully working fbdev backend to the uterm library. This allows
us to create our rendering pipeline on any linux machine.

The fbdev backend is not yet hooked up into kmscon. There are still some
remaining issues as we do not have OpenGL on fbdev if EGL is not compiled
with fbdev backend (which is usually not).

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconsole: rewrite console layer
David Herrmann [Sun, 17 Jun 2012 13:53:33 +0000 (15:53 +0200)]
console: rewrite console layer

This is a rewrite of the whole console layer so we can remove the
kmscon_buffer object to speed up the console.

This removes the split between the scroll region and the two margins so we
can resize margins in O(1). This also correctly merges the remaining
console modes so vte does not have to track them.

Btw. vim runs smoothly on kmscon with this console layer now (finally!).

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agofblog: add kernel patches for fblog driver
David Herrmann [Sat, 16 Jun 2012 22:29:04 +0000 (00:29 +0200)]
fblog: add kernel patches for fblog driver

The fblog drivers provides a replacement for fbcon and prints all kernel
log messages to all available framebuffer devices.

As long as the driver is not upstream, I will keep it here for reference.
See linux-serial@vger.kernel.org for discussion.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: allow setting margins
David Herrmann [Wed, 30 May 2012 18:49:39 +0000 (20:49 +0200)]
vte: allow setting margins

Support changing margin size of the terminal.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: fix HUP/HVP positioning bug
David Herrmann [Wed, 30 May 2012 18:48:22 +0000 (20:48 +0200)]
vte: fix HUP/HVP positioning bug

We mixed up X/Y coordinates. HUP and HVP send them reversed as often used
in terminals. Anyway, now it works correctly.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: more verbose SM/RM debug messages
David Herrmann [Wed, 30 May 2012 18:47:40 +0000 (20:47 +0200)]
vte: more verbose SM/RM debug messages

When we receive an unknown SM/RM message, be more verbose about what
exactly was received on the log output.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: add DA support
David Herrmann [Wed, 30 May 2012 18:37:33 +0000 (20:37 +0200)]
vte: add DA support

Send Device attributes as VT220.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconf: set TERM by default to vt220
David Herrmann [Wed, 30 May 2012 18:29:29 +0000 (20:29 +0200)]
conf: set TERM by default to vt220

vt220 seems to be the best alternative to "xterm" and should work with
every application.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: support cursor positioning CSI
David Herrmann [Wed, 30 May 2012 17:42:11 +0000 (19:42 +0200)]
vte: support cursor positioning CSI

Most of the CSIs for cursor movement are already implemented but HVP and
CUP are missing. This adds both CSI handlers to the VTE layer.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: update TODO items
David Herrmann [Wed, 30 May 2012 17:03:34 +0000 (19:03 +0200)]
vte: update TODO items

Both flags have been implemented recently so remove the TODO flag in the
comments.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconsole/vte: draw cursor
David Herrmann [Wed, 30 May 2012 17:02:17 +0000 (19:02 +0200)]
console/vte: draw cursor

If HIDE_CURSOR mode is not set, we draw the cursor at the current
position.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: set color to white on cell reset
David Herrmann [Wed, 30 May 2012 16:58:52 +0000 (18:58 +0200)]
vte: set color to white on cell reset

For the same reason as we reset it to white on cell creation we need to
reset it to white on cell reset. Otherwise we have odd artifacts in
reverse video mode.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: correctly reset console flags on soft reset
David Herrmann [Wed, 30 May 2012 16:53:43 +0000 (18:53 +0200)]
vte: correctly reset console flags on soft reset

We didn't reset the console flags on soft reset, however, this is required
to correctly synchronize the flags between vte and console.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconsole: reset font color to white by default
David Herrmann [Wed, 30 May 2012 16:52:20 +0000 (18:52 +0200)]
console: reset font color to white by default

Reset font color to white for new cells. This allows us to draw reversed
cells that are not occupied by any value. Otherwise, these cells would be
still black if reversed.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agofont: draw backgrounds
David Herrmann [Wed, 30 May 2012 16:48:59 +0000 (18:48 +0200)]
font: draw backgrounds

The pango backend now correctly draws colored or uncolored backgrounds.
This is needed to support drawing cursors.

We currently do not optimize this for performance. In fact, this reduces
performance a lot and we could avoid drawing backgrounds if we cleared the
background to the same color before. However, this also means we currently
could move drawing the background into the console backend again. The
transparency feature is awful, anyway and I don't see any reason
supporting it.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal/vte/console: support inverse-screen-mode
David Herrmann [Wed, 30 May 2012 16:19:49 +0000 (18:19 +0200)]
terminal/vte/console: support inverse-screen-mode

In inverse screen mode we switch background and foreground colors. As our
console layer supports transparent backgrounds, we have to do this in the
terminal layer.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconsole: allow retrieving console flags
David Herrmann [Wed, 30 May 2012 16:19:03 +0000 (18:19 +0200)]
console: allow retrieving console flags

This adds a new function that allows external subsystems to retrieve the
current flags of the console.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: implement origin mode
David Herrmann [Wed, 30 May 2012 16:00:06 +0000 (18:00 +0200)]
vte: implement origin mode

Convert the old relative-addressing mode of the console layer into origin
mode flag.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconsole: allow setting/resetting flags
David Herrmann [Wed, 30 May 2012 15:54:31 +0000 (17:54 +0200)]
console: allow setting/resetting flags

Instead of writing a function for each mode we now accept flags for the
console object. For now the flags are unused but other flags will be
added.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: implement local echo mode
David Herrmann [Wed, 30 May 2012 15:44:12 +0000 (17:44 +0200)]
vte: implement local echo mode

If SEND_RECEIVE_MODE is not set, we have to echo all data we send out.
That is, all escape sequences and all characters are directly interpreted
by us. To avoid recursion, we use a simply protection. However, some
escape sequences don't make sense in echo mode but we don't care here, as
this is a problem of the application/client, not us.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: add ESC+c hard-reset TODO item
David Herrmann [Wed, 30 May 2012 15:43:36 +0000 (17:43 +0200)]
vte: add ESC+c hard-reset TODO item

ESC+c is a hard-reset and now marked as TODO item.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: ignore unset csi parameters in SM/RM
David Herrmann [Wed, 30 May 2012 15:24:49 +0000 (17:24 +0200)]
vte: ignore unset csi parameters in SM/RM

When reading Setmode/Resetmode parameters, we should ignore all unset
parameters instead of printing a warning.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: fix up comments
David Herrmann [Tue, 29 May 2012 15:16:18 +0000 (17:16 +0200)]
vte: fix up comments

Fix up trivial comments. Remove already realized TODO items and add some
additional TODOs.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agovte: add auto-wrap support
David Herrmann [Tue, 29 May 2012 15:07:53 +0000 (17:07 +0200)]
vte: add auto-wrap support

VT510 manual says auto-wrap is disabled by default but most applications
(including bash) expect it to be on, therefore we enable it by default.

The console layer already supported it but the vte layer wasn't hooked up.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>