platform/upstream/kmscon.git
12 years agouterm: vt: return error in real_deactivate() if VT is dead
David Herrmann [Mon, 1 Oct 2012 11:24:41 +0000 (13:24 +0200)]
uterm: vt: return error in real_deactivate() if VT is dead

We used to return 0 if we don't know the VT id in real_deactivate().
However, the VT id is nowadays always known so we must return an error
instead.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: vt: make real_activate() behave like real_deactivate()
David Herrmann [Mon, 1 Oct 2012 11:23:38 +0000 (13:23 +0200)]
uterm: vt: make real_activate() behave like real_deactivate()

real_activate() has alsmost the same semantics as real_deactivate() so we
should also return -EINPROGRESS when we scheduled the VT switch. This
isn't used by kmscon currently, but may be used by other uterm users.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: fix out-of-bounds access when drawing sb
David Herrmann [Mon, 1 Oct 2012 10:46:09 +0000 (12:46 +0200)]
tsm: screen: fix out-of-bounds access when drawing sb

When drawing scroll-back buffer, we cannot guarantee that the lines are as
long as the current screen. Therefore, check that we aren't accessing out
of bounds and draw empty glyphs instead.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: add --term, --login, --palette and --sb-size
David Herrmann [Mon, 1 Oct 2012 10:41:06 +0000 (12:41 +0200)]
wlt: add --term, --login, --palette and --sb-size

These terminal options are simply copied from kmscon and behave the same
way.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: add scrollback-buffer grabs
David Herrmann [Mon, 1 Oct 2012 10:24:03 +0000 (12:24 +0200)]
wlt: add scrollback-buffer grabs

This copies the grabs from kmscon over to wlterm. It is now possible to
scroll up and down in the scrollback-buffer if you manually increase the
buffer size.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: add shl_misc to all its users
David Herrmann [Mon, 1 Oct 2012 10:23:15 +0000 (12:23 +0200)]
build: add shl_misc to all its users

Even though it's just a header, we want to be safe so add it to all users.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: input: plain: remove unused array
David Herrmann [Mon, 1 Oct 2012 10:22:34 +0000 (12:22 +0200)]
uterm: input: plain: remove unused array

We do not parse num-lock anymore, so remove the unused array. GCC didn't
warn about it, which is weird, but it's definitely not needed, anymore.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoshl: misc: add SHL_HAS_BITS() helper
David Herrmann [Mon, 1 Oct 2012 10:21:10 +0000 (12:21 +0200)]
shl: misc: add SHL_HAS_BITS() helper

This helper checks whether a bitmask has all the given bits set.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoMove grab-parsing into conf.c
David Herrmann [Mon, 1 Oct 2012 09:23:21 +0000 (11:23 +0200)]
Move grab-parsing into conf.c

We really need to clean this up and allow parsing of grabs in conf.c
again. xkbcommon is now mandatory so we can fix all the input layers to
use it.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoMake xkbcommon mandatory
David Herrmann [Mon, 1 Oct 2012 08:10:50 +0000 (10:10 +0200)]
Make xkbcommon mandatory

We really need xkbcommon. There is so much stuff (including parsing of
keyboard shortcuts in conf.c) that depends on it. Therefore, we make it
mandatory now which allows us to use xkbcommon functions all over the
place.

Note that xkbcommon itself has no runtime dependencies so it is a small
self-contained library. The only reason I didn't do this ealier is that
xkbcommon has not seen a public release, yet. However, that should be done
in the near future.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: implement proper minimal sizes
David Herrmann [Sun, 30 Sep 2012 22:16:57 +0000 (00:16 +0200)]
wlt: toolkit: implement proper minimal sizes

A widget may have different constraints regarding minimal sizes and the
size it occupies. Therefore, we need to pass a hint with the minimal size
together with the prepare-resize round.

The catch-all user can then decide to resize the catch-all field until it
is big enough to hold all minimal-sizes.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: terminal: remove dead non-snap code
David Herrmann [Sun, 30 Sep 2012 21:51:34 +0000 (23:51 +0200)]
wlt: terminal: remove dead non-snap code

This code isn't needed at all. It doesn't make sense to disable snapping
so remove this dead code.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: terminal: do not clear terminal background
David Herrmann [Sun, 30 Sep 2012 21:50:12 +0000 (23:50 +0200)]
wlt: terminal: do not clear terminal background

We snap to terminal borders now so there is no need to clear the
background as we paint it during redraw, anyway.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: fix scheduling resize on widget-creation
David Herrmann [Sun, 30 Sep 2012 21:49:24 +0000 (23:49 +0200)]
wlt: toolkit: fix scheduling resize on widget-creation

Instead of scheduling for redraw, we have to schedule for resize as the
new widget might want to change the size constraints.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: fix forcing redraw during unneeded resize-round
David Herrmann [Sun, 30 Sep 2012 21:39:33 +0000 (23:39 +0200)]
wlt: toolkit: fix forcing redraw during unneeded resize-round

If we resize the buffers, we skip redrawing as resizing implies this.
However, if the resize-handler skips resizing because the size didn't
change, it also skips redrawing.
Hence, we add a new flag to force redrawing even if the resize-operation
is skipped. This fixes a bug where we didn't redraw during stalling resize
operations by the user.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: fix moving cursor when scrolling during resize
David Herrmann [Sun, 30 Sep 2012 21:38:30 +0000 (23:38 +0200)]
tsm: screen: fix moving cursor when scrolling during resize

If we scroll the screen during resize to keep the screen-content constant,
we should also move the cursor position. Otherwise, the application has an
inconsistent state.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: terminal: snap to console-size on resize
David Herrmann [Sun, 30 Sep 2012 21:01:24 +0000 (23:01 +0200)]
wlt: terminal: snap to console-size on resize

Instead of resizing linearly, we now always resize only to the next
possible console size.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: allow widgets to control new size
David Herrmann [Sun, 30 Sep 2012 20:50:38 +0000 (22:50 +0200)]
wlt: toolkit: allow widgets to control new size

This changes the prepare-resize logic to allow widgets full control of the
new size. The initial size starts at 0 and each widget can increase it.
The last widget should be a catch-all that uses all size that is available
and computes a good new size based on the requirements that the previous
widgets provided.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: fix buffer-overflow when resizing and scrolling
David Herrmann [Sun, 30 Sep 2012 19:59:45 +0000 (21:59 +0200)]
tsm: screen: fix buffer-overflow when resizing and scrolling

We call screen_scroll_up() under special conditions when resizing. We
require it to allocate the lines with the new width even though we didn't
set it yet. Therefore, we need to set the width before calling it and then
adjust the height.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agokmscon: allow paths with --vt=<xy>
David Herrmann [Sun, 30 Sep 2012 17:00:45 +0000 (19:00 +0200)]
kmscon: allow paths with --vt=<xy>

The getty variants out there (including agetty) require an relative path
to the /dev directory as argument. This is really odd but we want to be
backwards-compatible to them so we allow this, too.

--vt now accepts:
  * A positive number which is internally converted into /dev/ttyXY
  * A string that does not start with '/' or '.' which is interpreted
    relative to /dev as /dev/%s
  * Everything else is interpreted as path

This option still selects only the TTY on seat0. On all other seats we do
not use controlling TTYs.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoshl: add shl_strtou() helper
David Herrmann [Sun, 30 Sep 2012 16:57:40 +0000 (18:57 +0200)]
shl: add shl_strtou() helper

The standard C/POSIX helpers are really ugly to use. This small helper
returns 0 on success, otherwise an error.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoFix printf() specified bugs all over the code
David Herrmann [Sun, 30 Sep 2012 16:01:35 +0000 (18:01 +0200)]
Fix printf() specified bugs all over the code

When enabling the printf() logic for the logging subsystem, several
warnings were produced about missing or wrong specifiers. This fixes all
those occurrences.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: do not encode invalid UTF8
David Herrmann [Sun, 30 Sep 2012 15:59:36 +0000 (17:59 +0200)]
tsm: unicode: do not encode invalid UTF8

We must under all conditions avoid encoding invalid UTF8. Otherwise, we
would rely on other applications to do error-recovery.
Unfortunately, this is no syntactical change but a semnatical fix as the
Unicode standard defines several codepoints which are invalid or which
must never be used in UTF8.
See the Unicode standard if you're interested in these codepoint ranges.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agolog/llog: add gcc format checks
David Herrmann [Sun, 30 Sep 2012 13:41:08 +0000 (15:41 +0200)]
log/llog: add gcc format checks

Add gcc attributes so the printf formats are correctly checked. This
causes a lot of warnings which will be fixed in the following commits.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoRelease kmscon-5
David Herrmann [Sun, 30 Sep 2012 13:15:41 +0000 (15:15 +0200)]
Release kmscon-5

This is a stable release for kmscon. It includes mainly bugfixes and
cleanups. The TSM library was created out of the internal VTE layers and
mostly code-cleanups took place.
Happy testing!

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotest_vt: fix compilation with new allocation API
David Herrmann [Sun, 30 Sep 2012 13:15:06 +0000 (15:15 +0200)]
test_vt: fix compilation with new allocation API

We changed the uterm API so we need to fix this test again.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoUpdate NEWS entry
David Herrmann [Sun, 30 Sep 2012 13:05:13 +0000 (15:05 +0200)]
Update NEWS entry

Add all important changes since kmscon-4.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: autogen.sh cleanup
Matthew Monaco [Sat, 29 Sep 2012 08:23:08 +0000 (02:23 -0600)]
build: autogen.sh cleanup

Use "set -e" to force exit on failure and convert $* to "$@".

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agokmscon: add --vt=<num> option to select VT for seat0
David Herrmann [Sat, 29 Sep 2012 09:49:54 +0000 (11:49 +0200)]
kmscon: add --vt=<num> option to select VT for seat0

VTs are only available on seat0 so this option does not affect other
seats. On seat0 it selects the VT that we run on. We do _not_ fall back to
another seat if it fails but disable this seat instead.

Reported-by: Matthew Monaco
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconf: add INT datatype
David Herrmann [Sat, 29 Sep 2012 09:47:42 +0000 (11:47 +0200)]
conf: add INT datatype

This is very similar to the UINT datatype but allows signed values. This
will be required for the coming --vt option as default value.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: add maximize toggle functionality
Scott Moreau [Sat, 29 Sep 2012 07:10:01 +0000 (01:10 -0600)]
wlt: toolkit: add maximize toggle functionality

Use wl_shell_surface_set_maximized() to notify the compositor that we
support being maximized now. The server actually decides whether it
resizes our window.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: fix compilation if _Static_assert() is not available
David Herrmann [Sat, 29 Sep 2012 08:13:45 +0000 (10:13 +0200)]
build: fix compilation if _Static_assert() is not available

I just noticed that older compiles do not support _Static_assert() so we
now check in configure.ac for it. We do not use it heavily, anyway.
Dynamic asserts are avoided entirely.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: don't warn when removing bad fds from epoll-set
David Herrmann [Thu, 27 Sep 2012 17:15:28 +0000 (19:15 +0200)]
eloop: don't warn when removing bad fds from epoll-set

People might want to close file-descriptors before destroying the ev_fd
object if they actually have no direct control of it. For instance if used
through a library.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: terminal: forward HUP event to the caller
David Herrmann [Thu, 27 Sep 2012 17:08:41 +0000 (19:08 +0200)]
wlt: terminal: forward HUP event to the caller

This forwards the PTY-HUP event to the caller so we can close the terminal
window when the client application exits.

Reported-by: Alexander Preisinger
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: vte: be less verbose about unhandled escapes
David Herrmann [Thu, 27 Sep 2012 16:57:50 +0000 (18:57 +0200)]
tsm: vte: be less verbose about unhandled escapes

This converts a warning into a debug message to avoid spending too much
time in the logger when parsing unknown content. This speeds up "cat
/dev/urandom" by like 1000x.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: make timerfd non-blocking
David Herrmann [Thu, 27 Sep 2012 12:29:33 +0000 (14:29 +0200)]
eloop: make timerfd non-blocking

We actually allow draining a timer so we cannot be sure that a timer has
valid data that we can read. We already handle the non-blocking case
everywhere, we just missed setting this flag.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: repeat key events
David Herrmann [Thu, 27 Sep 2012 10:56:18 +0000 (12:56 +0200)]
wlt: toolkit: repeat key events

This adds a timer that automatically repeats key-events while pressed. In
the XKB-1 protocol embedded in wayland, key-repeates are not sent for
performance reason so we have to emulate them.

It is currently unclear where to take key-repeat rates from. This will be
solved in the future, though.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: allow NULL for timer updates
David Herrmann [Thu, 27 Sep 2012 10:54:37 +0000 (12:54 +0200)]
eloop: allow NULL for timer updates

If NULL is passed for timer udpates, we simply assume that the timer
should be disarmed and use a zeroed itimerspec.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: add some --font-* options
David Herrmann [Thu, 27 Sep 2012 10:36:10 +0000 (12:36 +0200)]
wlt: add some --font-* options

Copy the options from kmscon which are font-* engine, size, dpi and name.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: fix cleaning buffers on resize
David Herrmann [Thu, 27 Sep 2012 10:26:30 +0000 (12:26 +0200)]
tsm: screen: fix cleaning buffers on resize

There is some nasty bug where we do not correctly clean buffers when
resizing. So we now clear the whole offscreen region on resize to go sure
everything is clean.

Note that cells can pretty much move everywhere during their
offscreen-life so this is a quite sophisticated task to keep track which
cells are clean and which not. Therefore, simply clean all of them when
taking them on-screen.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: fix accessing symbol-array out of bounds
David Herrmann [Thu, 27 Sep 2012 10:13:18 +0000 (12:13 +0200)]
tsm: unicode: fix accessing symbol-array out of bounds

The array type used to be from glib which did that check automatically. We
now have to check explicitely that we do not access it out-of-bounds.

This fixes a nasty resizing-bug of TSM.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: fix not recreating the default-table all the time
David Herrmann [Thu, 27 Sep 2012 10:11:52 +0000 (12:11 +0200)]
tsm: unicode: fix not recreating the default-table all the time

We actually created a new table all the time which led to huge memory
leaks. We now actually use the default table if it is available.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: fix use after free of dp_fd
David Herrmann [Thu, 27 Sep 2012 09:34:15 +0000 (11:34 +0200)]
wlt: toolkit: fix use after free of dp_fd

We need to keep dp_fd alive for all calls to wl_display_*(). Otherwise,
the callback will use it and access invalid memory.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: fix sending two surface.attach() reqs in one frame
David Herrmann [Thu, 27 Sep 2012 08:45:20 +0000 (10:45 +0200)]
wlt: toolkit: fix sending two surface.attach() reqs in one frame

The need_frame and idle_frame logic was slightly wrong. We ended up not
correctly requesting the frame-cb after a _real_ frame-cb. Hence, the
idle-cb did simply send the next attach() request.

This fixes this small race and adds a comment how all this is supposed to
work.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: main: increase default window size to 600x400
David Herrmann [Wed, 26 Sep 2012 22:45:32 +0000 (00:45 +0200)]
wlt: main: increase default window size to 600x400

Current default size is pretty small and there are still some resizing
bugs so increase it to have a working terminal.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: terminal: handle keyboard events
David Herrmann [Wed, 26 Sep 2012 22:45:05 +0000 (00:45 +0200)]
wlt: terminal: handle keyboard events

Pass keyboard events straight through to the TSM handler.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: add keyboard events
David Herrmann [Wed, 26 Sep 2012 22:43:23 +0000 (00:43 +0200)]
wlt: toolkit: add keyboard events

This implements the wl_keyboard client-side of the wayland protocol and
takes care of pushing keyboard events to the correct window.

A new callback is added to all widgets which is called when keyboard input
is sent to the specific window.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: terminal: draw terminal to buffers during redraw-cb
David Herrmann [Wed, 26 Sep 2012 21:48:35 +0000 (23:48 +0200)]
wlt: terminal: draw terminal to buffers during redraw-cb

Use the font subsystem to blit the console glyphs to the wlterm window
during redraw-callbacks.

This is mostly copied from src/terminal.c and the blitting functions from
uterm-fbdev.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: load font modules during startup
David Herrmann [Wed, 26 Sep 2012 21:47:12 +0000 (23:47 +0200)]
wlt: load font modules during startup

We need the text-font modules to render fonts in wlterm. Therefore, load
all the font modules during startup before initializing the application.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlterm: add hard-dependency to xkbcommon
David Herrmann [Wed, 26 Sep 2012 21:45:02 +0000 (23:45 +0200)]
wlterm: add hard-dependency to xkbcommon

wlterm needs xkbcommon as the wayland protocol depends on XKB states.
Hence, we add a hard-dependency for wlterm.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agopty: remove useless debug message
David Herrmann [Wed, 26 Sep 2012 20:45:38 +0000 (22:45 +0200)]
pty: remove useless debug message

Remote pty debug message about resizing.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotext: add helpers to (un)load all modules at once
David Herrmann [Wed, 26 Sep 2012 20:00:01 +0000 (22:00 +0200)]
text: add helpers to (un)load all modules at once

Instead of doing all this in main() we now use the two helpers. This makes
the code much more readable and avoids too many stuff in main().

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: build text_font subsystem as separate static library
David Herrmann [Wed, 26 Sep 2012 19:54:37 +0000 (21:54 +0200)]
build: build text_font subsystem as separate static library

We want to use the font-rendering layer in wlterm so we need to split this
out. Gladly, the layer has only a build-time dependency on uterm and not
other hard-coded stuff. That is, we have no cleanup to do.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: terminal: add TSM and PTY objects
David Herrmann [Wed, 26 Sep 2012 19:39:15 +0000 (21:39 +0200)]
wlt: terminal: add TSM and PTY objects

Create TSM screens and VTE objects plus a PTY and connect everything so we
have a working terminal. Keyboard input still needs to be hooked up and
the drawing functions aren't implemented, yet.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: toolkit: add wlt_window_get_eloop() helper
David Herrmann [Wed, 26 Sep 2012 19:38:00 +0000 (21:38 +0200)]
wlt: toolkit: add wlt_window_get_eloop() helper

This helper allows widgets to retrieve the eloop that is used by the
window. Otherwise, we couldn't dispatch events there.
No refcounting is needed as all widgets are destroyed before the window is
destroyed. Other users need to perform ref-counting by themself.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agopty: set argv via helper function
David Herrmann [Wed, 26 Sep 2012 19:24:34 +0000 (21:24 +0200)]
pty: set argv via helper function

This makes the pty layer independent of kmscon global state. This allows
us to use it in other applications bundled with kmscon. There is still
some work to do to make it fully independent so we can integrate it into
TSM. But that's not really needed, yet.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agopty: set "TERM" via helper function instead of kmscon_conf
David Herrmann [Wed, 26 Sep 2012 17:25:35 +0000 (19:25 +0200)]
pty: set "TERM" via helper function instead of kmscon_conf

Instead of accessing global state, we now set the TERM value via a helper
function. This is needed to make the pty layer independent of kmscon.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agowlt: theme: fix calculating window size regarding buttons
David Herrmann [Wed, 26 Sep 2012 17:12:16 +0000 (19:12 +0200)]
wlt: theme: fix calculating window size regarding buttons

We need to avoid reducing window size beyond our minimal width used to
draw buttons.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: update autotools logic and add "wlterm" application
David Herrmann [Wed, 26 Sep 2012 16:56:41 +0000 (18:56 +0200)]
build: update autotools logic and add "wlterm" application

Sorry for the big commit, but I was working on the wlterm application and
then thought I can rework the whole configure-logic again. This mainly
renames all build-defines to BUILD_DEFINE_* and BUILD_HAVE_* and allows
specifying which applications to build via --enable-kmscon/--enable-wlterm
and similar.

wlterm is a new application which is a native wayland client with no
external dependencies. It serves several purposes:
  * It uses TSM (not yet implemented, but will come soon) to create a
    console independent from kmscon. This shows how TSM can easily be used
    to create independent terminal emulators.
  * It is a native wayland application (probably the first independent
    wayland app so far?) and is used to test how well the wayland API
    works. As wayland is still under heavy development, we need more
    application-writers who report back whether the wayland-API makes
    sense to them and whether it works correctly.
  * A proper terminal-emulator for wayland! There is currently no proper
    emulator so we really need something that we can work with.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconf: remove uterm depedency
David Herrmann [Mon, 24 Sep 2012 19:57:23 +0000 (21:57 +0200)]
conf: remove uterm depedency

We move the GRAB type into kmscon-main.c to avoid any uterm dependency in
the conf subsystem. It is still open how we can better handle the
key-parser without a valid uterm_input object, but when there will be a
xkbcommon release, we can hopefully add a hard depedency to it.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agomain: pick up non-FBDEV_DRM devices without --fbdev
David Herrmann [Fri, 21 Sep 2012 12:31:27 +0000 (14:31 +0200)]
main: pick up non-FBDEV_DRM devices without --fbdev

Now that we can differentiate between fbdev-DRM and normal fbdev devices,
we can just pick up the normal fbdev devices by default.

--fbdev now makes kmscon use the FBDEV_DRM devices instead of pure DRM.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: monitor: mark DRM-fbdev devices as such
David Herrmann [Fri, 21 Sep 2012 12:25:10 +0000 (14:25 +0200)]
uterm: monitor: mark DRM-fbdev devices as such

Most DRM drivers also provide a legacy fbdev device so fbcon can pick it
up (and more importantly, we get kernel panics on it). However, as an
application developer, I don't want to use two devices which drive the
same physical hardware.

This marks all such DRM fbdev devices as FBDEV_DRM so kmscon doesn't pick
them up automatically.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video: ignore NULL buffers in vector requests
David Herrmann [Fri, 21 Sep 2012 11:20:21 +0000 (13:20 +0200)]
uterm: video: ignore NULL buffers in vector requests

When multiple requests are pushed into the video backends at once, we
should ignore NULL buffers. Otherwise, users might have to re-order
buffers just so we don't segfault. Instead, we now ignore them and the
application can set requests to NULL to signal us that it is unused.

This fixes some bugs with multi-monitor setups and kmscon.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agouterm: video: fbdev: disable double buffering
David Herrmann [Fri, 21 Sep 2012 11:17:43 +0000 (13:17 +0200)]
uterm: video: fbdev: disable double buffering

Kernel fbdev drivers are really broken in respect to virtual framebuffer
sizes. They all report success when allocating and even mmap()'ing them
but memory access will result in SIGBUS or even SIGSEGV.

Therefore, we disable it by default now. We might consider re-enabling it,
but there is really no need to work too much on fbdev backends.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotest_key: add small helper for client-side key-input
David Herrmann [Thu, 20 Sep 2012 11:16:32 +0000 (13:16 +0200)]
test_key: add small helper for client-side key-input

We sometimes need to test what keys a client receives when specific keys
are pressed. This small helper simply runs in a terminal and receives raw
keyboard input and prints it to stdout with debugging information.

This can definitely be improved with the help of the TSM state-machine to
print more useful information and directly parse the input. However, this
is better than nothing.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agopty: set VERASE character to backspace during setup
David Herrmann [Wed, 19 Sep 2012 19:13:48 +0000 (21:13 +0200)]
pty: set VERASE character to backspace during setup

Some crazy guy decided to set VERASE to DEL by default in the kernel. So
when starting an application which does not initialize the terminal on
startup, they may not notice that we actually send BACKSPACE as
erase-character (like /bin/login). Therefore, initialize VERASE to 010 so
everyone is happy.

Thanks to Etam for reporting this!

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: vte: send original VT220 escape codes for Shift+F1 to F12
David Herrmann [Wed, 19 Sep 2012 18:27:48 +0000 (20:27 +0200)]
tsm: vte: send original VT220 escape codes for Shift+F1 to F12

Nearly all terminals behave differently regarding Shift+FX keys. We now
mimic the behavior of the classic VT220 instead of the new-style function
keys. Nearly all applications expect the old codes.

This might be changed in the future, though. It does make much more sense
to send the new codes as they provide more information to the application.
And remappings should be done via XKB instead of inside of kmscon.
However, as long as xkbcommon does not have a config-file, we will be
stuck with this little hack.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agodocs: add systemd kmscon unit
David Herrmann [Tue, 18 Sep 2012 18:41:55 +0000 (20:41 +0200)]
docs: add systemd kmscon unit

This is a very basic unit that can start kmscon with a login shell. We may
have to improve it to replace agetty/etc fully, but it is a good start.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agopty: remove public eloop dependency
David Herrmann [Tue, 18 Sep 2012 17:22:09 +0000 (19:22 +0200)]
pty: remove public eloop dependency

We now create an eloop object internally to avoid requiring public eloop
headers. Functionality is still the same but now hidden in the library.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: vte: convert to use llog
David Herrmann [Tue, 18 Sep 2012 17:21:28 +0000 (19:21 +0200)]
tsm: vte: convert to use llog

Instead of using kmscon log_* functions, we now use the independent llog
subsystem.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: make libtsm build separately
David Herrmann [Tue, 18 Sep 2012 15:58:35 +0000 (17:58 +0200)]
build: make libtsm build separately

This builds libtsm as separate library so we can use it in other projects,
too.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: vte: introduce own modifiers
David Herrmann [Tue, 18 Sep 2012 15:36:29 +0000 (17:36 +0200)]
tsm: vte: introduce own modifiers

We do not want to depend on uterm just for the modifiers so introduce new
modifier-names which are in-sync with the UTERM names.

Inside of kmscon we still use the UTERM names everywhere, but inside of
TSM we now rely on the new names.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: vte: remove unused keyboard-action enum
David Herrmann [Tue, 18 Sep 2012 15:34:55 +0000 (17:34 +0200)]
tsm: vte: remove unused keyboard-action enum

This enum is no longer used to remove it.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: add *_set_palette() helper
David Herrmann [Tue, 18 Sep 2012 15:25:34 +0000 (17:25 +0200)]
tsm: add *_set_palette() helper

Instead of accessing kmscon-state we now add a helper to select the
palette and make the terminal-subsystem use it on initialization.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: vte: rename kmscon_vte object to tsm_vte
David Herrmann [Tue, 18 Sep 2012 14:59:39 +0000 (16:59 +0200)]
tsm: vte: rename kmscon_vte object to tsm_vte

Perform final rename of the remaining vte objects.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: rename files
David Herrmann [Tue, 18 Sep 2012 14:54:25 +0000 (16:54 +0200)]
tsm: rename files

Rename all TSM files to have a tsm_* prefix.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: vte: move vte layer to tsm_vte_*
David Herrmann [Tue, 18 Sep 2012 14:37:17 +0000 (16:37 +0200)]
tsm: vte: move vte layer to tsm_vte_*

This pushes the whole VTE layer into TSM and performs the huge rename
which turned out to be quite easy to accomplish.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal: drain timer after wakeup
David Herrmann [Tue, 18 Sep 2012 14:24:48 +0000 (16:24 +0200)]
terminal: drain timer after wakeup

Use the new timer_drain() helper to reset the timer after waking up.
Otherwise, we get spurious warnings that the CPU is too slow.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoeloop: add ev_timer_drain() helper
David Herrmann [Tue, 18 Sep 2012 14:24:06 +0000 (16:24 +0200)]
eloop: add ev_timer_drain() helper

This helper reads the current expiration-count from a timer. This can be
used when waking up from an idle-period or similar to reset the timer.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: switch to llog
David Herrmann [Tue, 18 Sep 2012 14:04:03 +0000 (16:04 +0200)]
tsm: screen: switch to llog

We already have an llog context so use it instead of relying on kmscon-log
subsystem. TSM-screen is now fully independent. Next step is TSM-vte.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: introduce TSM_SCREEN_OPT_RENDER_TIMING
David Herrmann [Tue, 18 Sep 2012 13:58:54 +0000 (15:58 +0200)]
tsm: screen: introduce TSM_SCREEN_OPT_RENDER_TIMING

Instead of accessing kmscon state from TSM, we now introduce options to
control the behavior of TSM-screens. Apart from logging, TSM is now
independent of any kmscon state/code.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: add screen-options helpers
David Herrmann [Tue, 18 Sep 2012 13:54:31 +0000 (15:54 +0200)]
tsm: screen: add screen-options helpers

Allow setting global options for screen objects. These are different from
flags as the latter affects the state-machine but options affect global
operation modes.

Options should be set by the application while flags are set by the VTE
handler. The latter might change often, while options should be set on
startup and then remain mainly constant.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: fix build without uterm-keysyms.h
David Herrmann [Tue, 18 Sep 2012 13:53:07 +0000 (15:53 +0200)]
build: fix build without uterm-keysyms.h

uterm-keysyms.h is no longer available but the makefile still references
it. Remove all occurrences of it to fix build again.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: add tsm_log_t llog-compatible log object
David Herrmann [Tue, 18 Sep 2012 13:01:45 +0000 (15:01 +0200)]
tsm: add tsm_log_t llog-compatible log object

We need to avoid logging to stderr directly in TSM so introduce the
tsm_log_t object similar to eloop.h.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: screen: move filenames to tsm_screen.[ch]
David Herrmann [Tue, 18 Sep 2012 12:53:46 +0000 (14:53 +0200)]
tsm: screen: move filenames to tsm_screen.[ch]

All TSM files use the "tsm_*" prefix and the object is now named "screen"
so rename the files to resemble this.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: fix header comments
David Herrmann [Tue, 18 Sep 2012 12:50:02 +0000 (14:50 +0200)]
tsm: fix header comments

Remove "kmscon" from any TSM headers and replace it by "tsm".

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoconsole: rename prefix to tsm_screen_*
David Herrmann [Tue, 18 Sep 2012 12:12:38 +0000 (14:12 +0200)]
console: rename prefix to tsm_screen_*

This is part of the TSM library creation. We also rename "console" to
"screen" as this layer actually manages the screen.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: remove tsm_symbol_get_u8()
David Herrmann [Tue, 18 Sep 2012 11:49:11 +0000 (13:49 +0200)]
tsm: unicode: remove tsm_symbol_get_u8()

This is no longer used. You should first retrieve the UCS4 string and then
use the UCS4 to U8 conversion helpers instead.

All users have already been converted so we can remove this helper safely.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoterminal: don't include tsm_unicode.h
David Herrmann [Tue, 18 Sep 2012 11:47:19 +0000 (13:47 +0200)]
terminal: don't include tsm_unicode.h

This is not needed. And we want to avoid any dependencies to this layer
wherever possible.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotext: pass symbol-data directly into text layer
David Herrmann [Tue, 18 Sep 2012 11:42:22 +0000 (13:42 +0200)]
text: pass symbol-data directly into text layer

Instead of converting symbols into UCS4 string in each backend, we now
pass the whole data from the console layer into the renderers.

This makes all renderers indepedent of any recently introduced
symbol-tables and they can be implemented inside of TSM without exporting
them. However, we still need to pass the IDs to the text layer. The text
layer must not use them for anything but identification. Moreover, it must
never assume that they are valid tsm_symbol_t values.
We do this so the backends can still have fast hashtable lookups rather
than allocating big keys containing the UCS4 string+length and using these
for lookups.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: add helper for ucs4->u8 string conversions
David Herrmann [Tue, 18 Sep 2012 11:41:17 +0000 (13:41 +0200)]
tsm: unicode: add helper for ucs4->u8 string conversions

This small helper allocates a string big enough to hold the whole u8
string. This should be used for short and temporary strings only! It
allocates way to much memory for bigger or long-living strings.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agobuild: make xkbcommon built-time dependency
David Herrmann [Tue, 18 Sep 2012 10:25:12 +0000 (12:25 +0200)]
build: make xkbcommon built-time dependency

We really need xkbcommon-keysyms.h for building kmscon/uterm/TSM/etc.
However, the recent fix was ugly and didn't really help. Instead we copy
the keysyms file into external/xkbcommon/ so we can just include the real
xkbcommon files from any source but have a fallback in external/.

Hence, you can still build kmscon without xkbcommon with this fallback,
but this will be removed the first day when xkbcommon sees a public
release.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: fix error-path in tsm_symbol_append()
David Herrmann [Tue, 18 Sep 2012 10:17:41 +0000 (12:17 +0200)]
tsm: unicode: fix error-path in tsm_symbol_append()

We never checked the memory helpers for errors because they used to be
from glib. However, with our own helpers we need to check for errors to be
sure.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agoshl: hashtable: add shl_hashtable_remove() helper
David Herrmann [Tue, 18 Sep 2012 10:14:53 +0000 (12:14 +0200)]
shl: hashtable: add shl_hashtable_remove() helper

This adds a helper that removes entries from a hashtable. This hasn't been
needed, yet, so we never provided it. However, the new unicode-helpers
will need it for proper error recovery.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: add symbol-table helpers
David Herrmann [Tue, 18 Sep 2012 10:03:17 +0000 (12:03 +0200)]
tsm: unicode: add symbol-table helpers

Add three helpers to create and manage symbol-tables. Also fix internal
default-table to use them.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: merge tsm_symbol_get() into table__get()
David Herrmann [Tue, 18 Sep 2012 09:48:02 +0000 (11:48 +0200)]
tsm: unicode: merge tsm_symbol_get() into table__get()

We do not do locking in tsm anymore so we can remove the table__get()
helper.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: export tsm_ucs4_to_utf8()
David Herrmann [Tue, 18 Sep 2012 09:37:59 +0000 (11:37 +0200)]
tsm: unicode: export tsm_ucs4_to_utf8()

This helper function may be useful to other external code and allows us to
always return UCS4 strings. Other code can then use this helper to convert
it into UTF8.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: add symbol-table contexts
David Herrmann [Tue, 18 Sep 2012 09:30:38 +0000 (11:30 +0200)]
tsm: unicode: add symbol-table contexts

We should avoid any global state in shared libraries. As the TSM code is
becoming a shared library, we definitely need contexts for symbol tables.
However, we don't want to fix up all code now so we use a default table
NULL instead.

This can be fixed later but is ok for now.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: unicode: remove log_* calls
David Herrmann [Tue, 18 Sep 2012 09:04:50 +0000 (11:04 +0200)]
tsm: unicode: remove log_* calls

The logging-layer is not a dependency of TSM so we cannot use it. It is
also not needed anymore, as the unicode-layer is working pretty well.

Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
12 years agotsm: move unicode.[ch] to tsm_unicode.[ch]
David Herrmann [Tue, 18 Sep 2012 08:54:06 +0000 (10:54 +0200)]
tsm: move unicode.[ch] to tsm_unicode.[ch]

All TSM related files will get the tsm_* prefix so move unicode headers
and sources.

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