The CoglAngle type can still be used for focused optimisations since the type
and macros remain. Uses of CoglAngle within Cogl have been replaced with floats;
COGL_ANGLE_FROM_DEG is no longer used anywhere and the replacements for
cogl_angle_cos -> cosf (same for sin) have been fixed to convert float values
in degrees to radians. This fixes the cogl-primitives API.
It's necissary to replace COGL_FIXED_FROM_INT with a (float) cast otherwise
the replacement maths may end up with integer rounding errors. This was
causing text to not be displayed due to the texture coordinate calculation
always rounding to (0,0)
The previous patch broke some of the normalization done before the sine value
gets multiplied with CLUTTER_ALPHA_MAX. This e.g. broke test-actors when sine
values went through to -1, as the o-hands were scaled so large all you saw was
the red 'O'.
This commit doesn't actually include any direct changes to source; you
have to run ./fixed-to-float.sh. Note: the script will make a number of
commits itself to your git repository a various stages of the script.
You will need to reset these if you want to re-run the script.
* NB: Be carefull about how you reset your tree, if you are making changes
to the script and patches, so you don't loose your changes *
This aims to remove all use of fixed point within Clutter and Cogl. It aims to
not break the Clutter API, including maintaining the CLUTTER_FIXED macros,
(though they now handle floats not 16.16 fixed)
It maintains cogl-fixed.[ch] as a utility API that can be used by applications
(and potentially for focused internal optimisations), but all Cogl interfaces
now accept floats in place of CoglFixed.
Note: the choice to to use single precision floats, not doubles is very
intentional. GPUs are basically all single precision; only this year have high
end cards started adding double precision - aimed mostly at the GPGPU market.
This means if you pass doubles into any GL[ES] driver, you can expect those
numbers to be cast to a float. (Certainly this is true of Mesa wich casts
most things to floats internally) It can be a noteable performance issue to
cast from double->float frequently, and if we were to have an api defined in
terms of doubles, that would imply a *lot* of unneeded casting. One of the
noteable issues with fixed point was the amount of casting required, so I
don't want to overshoot the mark and require just as much casting still. Double
precision arithmatic is also slower, so it usually makes sense to minimize its
use if the extra precision isn't needed. In the same way that the fast/low
precision fixed API can be used sparingly for optimisations; if needs be in
certain situations we can promote to doubles internally for higher precision.
E.g.
quoting Brian Paul (talking about performance optimisations for GL programmers):
"Avoid double precision valued functions
Mesa does all internal floating point computations in single precision
floating point. API functions which take double precision floating point
values must convert them to single precision. This can be expensive in the
case of glVertex, glNormal, etc. "
Instead of returning CLUTTER_X11_FILTER_CONTINUE always from
clutter_x11_handle_event() return CLUTTER_X11_FILTER_REMOVE if
the event was on a stage and translated to a ClutterEvent.
Signed-off-by: Emmanuele Bassi <ebassi@linux.intel.com>
The cairo pkg-config is crafted such that #include <cairo.h> should be used.
If building with a private build of cairo installed to a custom prefix
<cairo/cairo.h> can cause an un-intended cairo header to be picked up during
compilation due to /usr/include being in the header search path.
When building out of tree the generated scripts for the unit tests
need to explicitly reference the original src dir to be able to find
test-launcher.sh, like this:
$(top_srcdir)/tests/conform/test-launcher.sh
Also test-launcher.sh now passes -m slow --verbose to gtester. Without
-m slow then the wrappers dont work for some of the timeline tests.
Our beloved math.h exports, from bits/mathcalls.h, a bare "y1" symbol.
Apparently, it's unthinkable for code including <math.h> to also declare
arguments or variable named "y0", "y1" and "yn".
Anyway, the quick fix already used elsewhere in Clutter's codebase is
to rename the colliding variables "y_0", "y_1" and "y_n" - and
obviously everything similar to them as well, using the same pattern.
The name of the parameter in the header and the one in the gtk-doc
annotation on top of a function must match.
Unfortunately, there is an index() function declared inside strings.h
which makes gcc complain for the "index" argument as soon as we
enable the extra compiler flags we use when distchecking.
Hence, we need to rename "index" to "index_" in the header and in
the source files.
* cairo-texture:
[cairo-texture] Remove the construct only restriction on surface size
[cairo-texture] Silently discard 0x0 surfaces
Re-indent ClutterPath header
Add a test case for the new cairo path functions
Add clutter_path_to_cairo_path and clutter_path_add_cairo_path
Warn instead of returning in the IN_PAINT check
Small documentation fixes
Print a warning when creating a cairo_t while painting
Do not set the IN_PAINT flag inside the Stage paint
Set the IN_PAINT private flag
[docs] Add ClutterCairoTexture to the API reference
Add ClutterCairoTexture
Require Cairo as a Clutter dependency
Conflicts:
Fix merge conflict in clutter/clutter-path.h
It is possible to change the surface size after construction with
clutter_cairo_texture_set_surface_size so it doesn't seem right to
restrict changing the properties.
clutter_cairo_texture_resize_surface_internal is called in a handler
for the notify signal. It is called there rather than directly in the
set_property handler so that changing both properties in a single
g_object_set will only cause one resize. The constructed override is
no longer needed.
resize_surface_internal will now bail out if the size of the surface
is already the right size.
Some drivers (e.g. Nvidia) get upset if you try to create multiple glx pixmaps
for the same server side pixmap object, even though you might have unique
client side names, we now avoid hitting this problem by destroying the current
glx pixmap early within clutter_glx_texture_pixmap_create_glx_pixmap.
A label is now displayed under the rectangle showing the current
gravity. The text for the gravity is taken from the GEnumClass. This
makes it easier to verify that the test is working correctly.
The calculation for cubic bezier curves had an extra multiplication by
3 which was causing the curve to go over 1.0 very quickly. This had
the affect of making test-animation appear to complete much before the
completed signal is emitted.
The patch makes it cast to double before subtracting the original
value from the target value. Otherwise if the target value is less
than the original value then the subtraction will overflow and the
factor will be multiplied by a very large number instead of the
desired interval.
The problem is demonstrable using the border-width property of
ClutterRectangle.
The current CairoTexture can be created with a surface size of 0
by 0 pixels, but a warning will be printed.
Worse, the surface can be resized to be 0 by 0 pixels without a
warning. The :surface-width and :surface-height properties accept
a minimum value of 0, and not check is performed on either the
constructor or set_surface_size() parameters to enforce the "greater
than zero" rule.
The correct and consistent behaviour is to allow a 0 by 0 pixels
surface size everywhere; inside surface_resize_internal(), the
current surface will be destroyed and if either :surface-width or
:surface-height are set to 0, the resizing terminates.
Attempting to create a Cairo context from a CairoTexture with
either :surface-width or :surface-height set to 0 will result in
a warning.
This allows:
- creating a CairoTexture with :surface-width or :surface-height
set to zero and delaying the surface resize at a later point;
- resizing the surface to 0 by 0 pixels to destroy the image
surface used internally;
- increase the consistency in the usage of CairoTexture.
Currently, the conformance test suite creates symbolic links pointing
to a wrapper script that just parses the name used to invoke it and
calls the gtester with the correct path.
Unfortunately, this presents two issues:
- it does not really work on file systems that do not
support symbolic links
- it leaves behind the symbolic links, which cannot
be automatically cleaning by 'make clean'
Both can be solved by creating a small script that invokes the wrapper
one with the test unit path.
The Makefile will use test-conform to extract the unit test paths
and generate a list that will be iterated over to create the
executable name (using the "test-name" convention also used by the
interactive tests, instead of "test_name"); the executable is then
just a simple shell script that invokes the wrapper script passing
the unit test path on the command line. The wrapper script will
use the first argument to work correctly, so it could be simply
executed like:
./test-wrapper.sh /path/to/unit_test
Which is another improvement over the current implementation, where
the wrapper script does not work when invoked directly.
The gdouble value represents an interval along the path from 0.0 to
1.0. This makes more sense than using an alpha value because paths are
not directly related to ClutterAlphas and the rest of the Clutter API
tends to expose gdouble arguments.
Since commit c7c5cf9b ClutterCloneTexture causes an extra paint of the
source actor during the paint run of the cloned texture if the source
is not yet visible. When the stage is first shown it is redrawn
immediatly before it is mapped which means get_paint_visibility will
fail on the source actor so the inner paint will be run. The paint
guards were a global variable so they didn't cope with a second actor
being painted.
The breakage didn't occur until commit d510a4b0 because
get_paint_visibility was also broken.
The ClutterParamSpecFixed constructor is declared in the header
as taking ClutterFixed values, but the implementation takes
ClutterUnit values instead. This obviously works because fixed
and units are exactly the same type.
This commit fixes the wrong parameters.
The nodes of the test path have been reordered because Cairo coalesces
multiple move operations into a single move so the comparison would
fail if the two move nodes are consecutive.
Fix the CairoTexture description, and some of the comments inside
the code, especially with regards to the alpha channel unpremultiplication
that we have to perform each time we upload the image surface to
GL.
If you create a Cairo context in the middle of a paint run and then
you destroy it, the CairoTexture will have to upload the contents of
the image surface to the GL pipeline. This usually leads to slow
downs and general performance degradation.
ClutterCairoTexture will warn to the console if Clutter has been
compiled with the debug messages and if create() or create_region()
are called while an actor is in the middle of a paint.
Since the CLUTTER_ACTOR_IN_PAINT private flag is set as part
of the paint process by clutter_actor_paint(), there is no
need to set it inside the ClutterStage paint function.
When calling clutter_actor_paint() we should be setting the
CLUTTER_ACTOR_IN_PAINT private flag. This allows signalling
to each Actor subclass that we are effectively in the middle
of a paint sequence. Actor subclasses can check for this
private flag and act based on its presence - for instance to
avoid recursion, or to detect performance degradation cases.
Move the ClutterCairo actor from a separate library to an in-tree
actor.
ClutterCairoTexture is a simple texture subclass that allows you
to retrieve a Cairo context for a private image surface. When the
Cairo context is destroyed it will cause the image surface
contents to be uploaded to a GL texture.
The image surface used is not hardware accelerated.
Cairo has been an indirect dependency for Clutter since 0.8, through
the PangoCairo API.
Now we explicitly depend on Cairo in order to merge the clutter-cairo
API into Clutter core.
When the apply_transform_to_point() and its relative variant
landed in Clutter 0.3, the initial approach was to modify the
passed vertex as an in-out parameter. This was later dropped
in favour of a more consistent out parameter.
Unfortunately, the implementation never changed: both methods
where modifying the passed vertex with the partial results of
the computations.
This commit copies the contents of the "point" ClutterVertex
argument inside a stack variable and, for good measure, constifies
the argument.
Thanks to Thomas Steinacher for catching this in Python.
* clutter/cogl/gl/cogl-defines.h.in:
* clutter/cogl/gl/cogl-context.h:
* clutter/cogl/common/cogl-mesh.c: Rename the glBufferDataSub
function to glBufferSubData. When calling glXGetProcAddress with
the former Mesa returns a stub dispatch function which will
segfault if you try to use it. With NVIDIA it returns NULL so
_cogl_features_init decides the card doesn't have VBO support.
