This adds a first draft of systemd-consoled. This is still missing a lot
of features and does some rather primitive rendering. However, it shows
the direction this code is going and serves as basis for further testing.
The systemd-consoled binary should be run as `systemd --user' unit. It
automatically picks up any session marked as Desktop=SYSTEMD-CONSOLE.
Therefore, you can use any login-manager you want (ranging from /bin/login
to gdm) to create sessions for systemd-consoled. However, the sessions
managers must be prepared to set the Desktop= variable properly.
The user-session is called `systemd-console', only the daemon providing
the terminal environment is called `systemd-consoled' (mind the 'd').
So far, only a single terminal session is provided on each opened
user-session. However, we support multiple user-sessions (even across
multiple seats) just fine. In the future, the workspace logic will get
extended so you can have multiple terminal sessions in a single
user-session for easier access.
Note that this is still experimental! Instructions on how to run it will
follow shortly.
As it turns out, I can actually send data to the pty faster than the
terminal can read. Therefore, make sure we read as much data as possible
but bail out early enough to not cause starvation.
Kernel TTY buffers are 4k, so reduce the overall buffer size, but read
more than once if possible (up to 8 times sounds reasonable).
Instead of increasing the screen-age on redraw, we now increase it only on
real updates. This is effectively the same, but avoids increased age
counters on backbuffer rendering. Therefore, we can now check age counters
against fronbuffers safely, while rendering frames in background.
Terminals use pseudo color-codes mixed with 8bit and 24bit colors. Provide
a color-converter so external renderers only have to deal with ARGB32
colors.
This requires a color-palette as input as there's no fixed mapping. We
provide a default, but maybe we wanna support external palettes in the
future.
This is the most simple way to render cursors: flip attr->inverse of the
cursor cell. This causes the background and foreground colors of the
cursor-cell to be inversed.
Now that we render cursors ourselves, make subterm not call into the
parent terminal to render cursors.
Matching keyboard shortcuts on internationalized keyboards is actually
non-trivial. Matching the actual key is easy, but the modifiers can be
used by both, the matching and the translation step. Therefore, XKB
exports "consumed-modifiers" that we use to figure out whether a modifier
was already used by the translation step.
The new IDEV_KBDMATCH() helper can be used to match on any keyboard
shortcut and it will do the right thing.
Implement the feed_keyboard() handling by mapping XKB keys according to
DEC-VT behavior.
Public information on terminal key-mappings is pretty scarce. We only
implement the most basic mapping for now. Further improvements welcome!
In case we cannot render a glyph, we want a fallback we can display
instead. If we rely on the font itself to provide the fallback character,
we have nothing to display if that character is not available. Therefore,
add a static fallback that we can use at any time.
If a pipe is enabled/disabled, we have to clear crtc->applied of the
linked CRTC. Otherwise, we will not run a deep modeset, but leave the crtc
in the pre-configured state.
We really want more sophisticated aging than just 64bit integers. So
always provide front *and* back buffers to renderers so they can compare
arbitrary aging information and decide whether to re-render.
Instead of looking for available back-buffers on each operation, set it to
NULL and wait for the next frame request. It will call back into the pipe
to request the back-buffer via ->target(), where we can do the same and
look for an available backbuffer.
This simplifies the code and avoids double lookups if we run short of
buffers.
Coverity complained about this code and is partially right. We are not
really protected against integer overflows. Sure, unlikely, but lets just
avoid any overflows and properly protect our parser loop.
Lets return the parsed length in term_utf8_decode() instead of a buffer
pointer. Store the pointer in the passed argument.
This makes it adhere to the systemd coding-style, were we always avoid
returning pointers, but store them in output arguments. In this case, the
storage is not allocated, so it doesn't fit 100% to this idiom, but still
looks much nicer.
We currently select front-buffers as new back-buffer if they happen to be
the last buffer in our framebuffer-array. Fix this by never selecting a
new front buffer as back buffer.
Spotted with coverity. If parsing both /etc/os-release and
/usr/lib/os-release fails then null would be passed on. The calls
to parse the two files are allowed to fail. A empty /etc may not
have had the /etc/os-release symlink restored yet and we just
try again in the loop. If for whatever reason that does not happen
then we now pass on 'n/a' instead of null.
Coverity seems to think that we can later end up with the "them"
fd having a negative value. Even after a succesful barrier_create.
Add some test to verify that the constructor went well. If coverity
still complains then it must mean that it thinks the the value is
overwritten later.
Example from Tom Gundersen is included using xi:include.
The copyright notice stands out a bit. Maybe it should be removed,
and the code placed in public domain.
The caller may have an existing DUID that it wants to use, and may
want to use some other DUID generation scheme than systemd's
default DUID-EN.
[tomegun: whitespace - we never use tabs]
This patch adds a transient user unit directory under
`$XDG_RUNTIME_DIR/systemd/user/` and stores transient user-instance
units (such as those created by `systemd-run --user`) under there
instead of putting them in $XDG_CONFIG_HOME/systemd/user/.
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=67331
