= KMSCON Release News =

CHANGES WITH 4:
	* The "conf" subsystem was modularized so we can use it in other
	  binaries as well. It is also much easier to add new parameters now.

	* Remove internal dependencies. Several subsystems will be split into
	  separate libraries (like eloop, uterm, vte+console) so we should avoid
	  any internal dependencies between these. This shouldn't affect
	  performance, though.

	* New --seats argument which allows to specify what seats kmscon is
	  running on.

	* The old vt.[ch] subsystem is now merged into uterm_vt.c and a fake VT
	  backend was added. This allows to activate/deactivate kmscon via
	  keyboard shortcuts. This is for debugging only, though.

	* The VTE layer was extended a lot to support more advanced control
	  sequences and fix several bugs.

	* --dumb now forces kmscon to use non-accelerated rendering

	* The scrollback buffer now works. You can specify keyboard shortcuts
	  via --grab-* options. Default is: <Shift>Up/Down/Prior/Next

	* GL_EXT_unpack_subimage is not required anymore.

	* _Lots_ of speed improvements in the rendering path.

	* The unifont font backend was added. This is similar to 8x16 but has a
	  lot more characters.

	* ... and a lot of bugfixes

CHANGES WITH 3: (development release)
	* VTE support has been improved greatly. vt220 should be supported fully
	  except for multi-line characters.

	* glib dependency removed. Hashtables and unicode support is now
	  implemented in kmscon.

	* Configfiles are now supported. /etc/kmscon.conf and ~/.kmscon.conf are
	  read by kmscon on startup. They can include the same options as the
	  command line options of kmscon. However, command-line options have
	  precedence.

	* Multiple video objects are now supported per seat. That is, if you
	  have more than one graphics card per seat, all of them will be used by
	  kmscon. Multiple displays were supported from the beginning.

	* Modularized input backend. The XKB backend is fully optional and can
	  be activated/deactivated during runtime. The Dumb backend has been
	  renamed to "plain".

	* The eloop library is fully documented and should be API stable from
	  now on.

	* The font renderer subsystem has been rewritten and replaced with a new
	  subsystem which can load different font-renderers during runtime.
	  Several backends including static 8x16 backend has been added. It is
	  possible to compile kmscon without freetype2/pango dependencies now.

	* 2D blitting to framebuffers has been improved greatly. More devices
	  are supported now and blending is available, too. The 3D devices have
	  been updated to support these blitting operations via OpenGL. This may
	  be slow on older devices, though.

	* New text renderer subsystem has been added. It can load multiple
	  different text-renderers during runtime. The default renderer (bblit)
	  uses 2D blitting operations to render text. However, the gltex
	  renderer is also available (optionally) which renders with OpenGL
	  textures. This can improve performance greatly on modern devices.

	* Gcc optimizations are now enabled by default to improve rendering
	  performance.

CHANGES WITH 2: (development release)
	* Build scripts were improved and simplified. They also allow much more
	  control about the build process now. The source was split into several
	  different libraries.

	* Added llog.h helper for logging in libraries.

	* Eloop was moved into separate library and supports more more
	  use-cases now. It can be used as a standalone project.

	* Terminal emulator (vte.c) was improved a lot and can run vim now.

	* Console layer was rewritten and is now much faster and much more
	  standards compliant.

	* Uterm monitor was added. The monitor uses udev and systemd-logind to
	  watch the system for new seats and devices. This allows us to run on
	  multiple seats simultaneously and being hotplug-capable.

	* Uterm vt layer was added. This simply wraps the older VT layer so we
	  can switch to different VT systems on-the-fly.

	* New uterm video backends: The fbdev backend is now fully functional
	  kmscon can run on any fbdev device now. Furthermore, the dumb backend
	  was added which is simply an unaccelerated DRM backend that should
	  work even without EGL, GL, gbm, etc.

	* New 2D rendering API. In case OpenGL is not available, we now allow
	  blitting images into the framebuffer. This is used for fbdev and dumb
	  video devices.

	* Font renderer now depends on pango. This is is needed only until I get
	  the time to rewrite the freetype backend. But for now we depend on
	  pango so have full font support.

	* lots of small fixes

CHANGES WITH 1: (development release)
	* First kmscon release. For testing purposes only. Includes a rendering
	  pipeline and a very rundimentary terminal emulator.