Surface creation now specifies the exact format, not only the bit depth
of each surface which is used for rendering.
Additionally we now actually store the surfaces in that format, instead
of converting everything to 32bpp when drawing or e.g. handling palettes.
In order to be able to support 16bit canvases on 32bit screens and 32bit
canvases on 16bit screens we need to handle format conversion when drawing
RedPixmaps.
The way this works now for X11 is that we only have one PIXELS_SOURCE_TYPE
for pixmaps, which always has a pixman_image_t for the data, but additionally
it has an XImage (shared mem or not) if the screen the pixmap was created
for (i.e. an explicit one or the default screen) has the same format as
the pixmap.
When we draw a pixmap on a drawable we have two variants. If the pixmap
has a XImage and it matches the format of the target drawable then we
just X(Shm)PutImage it to the drawable. If the formats differ, then we
create a temporary XImage and convert into that before drawing it to
the screen.
Right now this is a bit inefficient, because we always allocate a new
temporary image when converting. We want to add some caching here, but
at least this lets things work again.
We need to know the format for other drawables too (like for instance
the native format of a window), so we're pushing this down.
This changes a bunch of references to be RedDrawable::, but not all.
The the old RedPixmap:: references still work, but will be phased out.
We now support 16bit format pixmaps as well as the old ones. Including
both 555 and 565 modes.
We drop the palette argument for pixmap construction as it was only
used for black/white anyway.
Canvas creation is simplified so that there is no separate set_mode
state. Canvases are already created in the right mode and never change.
The current glx code is looking for a rgb32 visual and always failing
if there is none. This means not even software rendering starts up
on e.g. 16bit visuals. This commit makes it pick software fallbacks
on 16bit visuals.
Long term we need to fix the gl implementation to do 16bpp too.
While the fix could have been more effective,
it seems like this patch stream better with the coding
logic that was there..., maybe later we will want to change
the locking into more effective way.
(There is just the primary surface to protect in reiality)
Signed-off-by: Izik Eidus <ieidus@redhat.com>
This is pretty straightforward, although there are two weird issues.
The current encoder has two bugs in the yuv conversion. First of all
it switches red and blue, due to something of an endianness issue. We
keep this behavior by switching red and blue. Maybe we want to
change this in the new protocol version since switching this may
cause jpeg compression to be worse.
Secondly, the old coder/decoder did rgb to/from yuv420 wrongly for
jpeg, not using the "full scale" version of Y that is used in jpeg,
but the other one where y goes from 16 to 235. (See jpeg/jfif
reference on http://en.wikipedia.org/wiki/YCbCr for details.)
The new decoder uses the full range in order to get better quality,
which means old encoders will show slightly darker images.
This completely removes all ffmpeg usage in the client
Now we can send commands from the server to the client
to destroy surfaces (right now just the primary surface)
Needed for offscreens support)
Another patch`s on the way.
Signed-off-by: Izik Eidus <ieidus@redhat.com>
Replace all "$(top_srcdir)/common" with "$(SPICE_COMMON_DIR)"
and all "$(top_srcdir)/client" with custom "$(CLIENTDIR)"
This would (after following patches) enables building the client from
either spice/ (top directory) or spice/client.
In Fedora 13, the linker doesn't pull in DT_NEEDED libraries anymore,
so we have to list the things that we depend on explicitly.
This affects several X extension libraries, and also the pthread
library.
Every place that does a regular malloc/calloc and aborts on failure
should use spice_malloc/spice_mallo0 instead, which is leaner and cleaner.
Allocations of dynamically sized arrays can use g_malloc_n or g_new etc
which correctly handle multiplication overflow if some of the arguments
are not trusted.
Instead of having two virtualizations of the canvas we push the
virtualization into the canvas code itself. This not only avoids
the duplication of this code, it also makes the exposed API for the
canvas much smaller (in terms of exported API).
It also lets us use the virtualization to implement basic support
for operations in canvas_base which is then overridden by each canvas
implementation.
