size buffer if it's big enough, and otherwise allocate memory.
* Added BLI_dynstr_appendf() to construct strings easily in printf
style, this should make it easier to construct menu strings for
example.
* merge the recent scons changes from trunk, so this means that the same rules now apply:
- scons-local available
- platform default configs updated
- check your user-config.py if you have made a full copy of default config
Note: BGE and blenderplayer still not working, but that is normal ;) (so turn them off!)
* Added more error prints for wrong definitions, for cases that
would laters cause problems compiling or crash at runtime, and
also made messages more clear.
* Added some skeleton code for main/ID/mesh/vertex types for testing.
* Added support for automatic arrays as collections using SDNA.
* Changed how pointers to data work. Now they are always wrapped
in a PointerRNA struct, which contains the data pointer and type,
and also the data pointer and type of the ID datablock that this
belongs to, since for example a vertex on it's own may not have
enough information for some operations, it also needs the mesh.
* Added some code for defining dependencies with RNA, and looking up
data with paths like: scenes[0].objects["Cube"].data.verts[7].co.
Note sure either will end up being used, this is experimental.
http://wiki.blender.org/index.php/BlenderDev/Blender2.5/RNA
This is the first code for the Data API, also known as RNA system in the
2.5 Branch. It does not include a user interface, and only wraps some
Scene properties for testing. It is integrated with Scons and Makefiles,
and compiles a 'makesrna' program that generates an RNA.c file.
http://wiki.blender.org/index.php/BlenderDev/Blender2.5/DataAPIhttp://wiki.blender.org/index.php/BlenderDev/Blender2.5/RNA
The changes are quite local, basically adding a makesrna module which
works similar to the makesdna module. The one external change is in
moving genfile.c from blenloader to the makesdna module, so that it can
be reused by the RNA code. This also meant changing the DNA makefiles.
It seems to be doing dependencies correct still in my tests, but if
there is an issue with the DNA not being rebuilt this commit might be
the one causing it. Also it seems for scons the makesdna and makesrna
modules are compiling without warnings. Not a new issue but this should
be fixed.
The RNA code supports all types as defined in the Data API design, so
in that sense it is fairly complete and I hope that aspect will not
have to change much. Some obviously missing parts are context related
code, notify() functions for updates and user defined / ID properties.
- operator that rips current area into new window
- implemented on the window level.
- sets C->area in current context if necessary
== fix ==
- small bugfix: missing return in WM_event_add_keymap_handler
==========
* Changed wmOperatorType, removing init/exit callbacks and adding cancel
callback, removed default storage in favor of properties. Defined return
values for exec/invoke/modal/cancel.
* Don't allocate operator on the stack, and removed operator copy for
handlers. Now it frees based on return values from callbacks, and just
keeps a wmOperator on the heap. Also it now registers after the operator
is fully finished, to get the correct final properties.
* Changed OP_get_* functions to return 1 if the property is found and 0
otherwise, gives more readable code in my opinion. Added OP_verify_*
functions to quickly check if the property is available and set if it's
not, that's common for exec/invoke.
* Removed WM_operatortypelist_append in favor of WM_operatortype_append
which takes a function pointer instead of a list, avoids macro's and
duplicating code.
* Fix a crash where the handler would still be used while it was freed by
the operator.
* Spacetypes now have operatortypes() and keymap() callbacks to abstract
them a bit more.
* Renamed C->curarea to C->area for consistency. Removed View3D/View2D/
SpaceIpo from bContext, seems bad to keep these.
* Set context variables like window/screen/area/region to NULL again when
leaving that context, instead of leaving the pointers there.
* Added if(G.f & G_DEBUG) for many of the prints, makes output a bit
cleaner and easier to debug.
* Fixed priority of the editors/interface module in scons, would otherwise
give link errors.
* Added start of generic view2d api.
* Added space_time with some basic drawing and a single operator to change
the frame.
- draw arrow in area that will removed and draw area that will be kept lighter (change from trunk: without arrow)
- not using notifications yet, but uses flag in ScrArea to determine draw type for area. (might be worth discussing)
- experimental: swapping of areas extended a bit, which allows user to choose area a bit more freely
This is a patch by Michael Fox (mfoxdogg) that make more
easy join two are, just RMB+Alt in the "origin" area,
drag the mouse to the "target" area and release the mouse.
Note that this is not a "final mouse-bind" for the join
operator, just an easy and simple option for test in
this early stage of 2.5.
Other note: maybe it's a good idea add azones here ?
After check this a little more I make some changes to the
API and now work on the following form:
WM_gesture_init(C, type);
while() {
/* handler event, etc */
/* if something change. */
if(need_update) {
/* update the gesture data and notify about it. */
WM_gesture_update(C, data);
WM_event_add_notifier (.. WM_NOTE_GESTURE_CHANGE ..);
}
}
WM_gesture_end(C, type);
Another of the change is that now the gesture data is a link list
in the window struct, so we can have multiples "gestures" (but
of different type) at the same time.
Also take care that the "gesture data" is reusable, that mean that
only alloc it 1 time and use in all the place, that is
why don't support multiple gesture of the same type, but of course
that can be change.
This is a first implementation of the "gesture manager" system,
the idea is put the WM in a automatic draw mode so we can
implement different "Gesture types" to draw different class
of data (lasso, bound box, etc).
The gesture data is passed through the data field of the notifiers,
i think that we can change this to something like:
WM_gesture_init(C, data); /* put the data in the context. */
while() {
/* send WM_NOTE_GESTURE_CHANGED to update screen */
}
/* send event and free the data in the context. */
WM_gesture_end(C);
Also i add a new operator and event to test the gesture manager.
The new operator is the "border select" function, just press BKEY
in the window and LMB or ESCKEY to exit.
In the case of LMB you can see a print in the console about the
BORDERSELECT event.
All this still need a lot of work, comment are welcome.
add split area to screen manager
- RMB+ALT down on area edge activates
- mouse move interactively to area that is to be removed
- LMB to confirm, ESC to cancel
TODO:
- notifications for interactive drawing to mark area to remove
Some fixes
- uninitialized var warning.
- exit function of operators need to run before modal handler is removed if operator is used there.
- replaced MEM_mallocN with MEM_callocN to get rid of uninitialized mem for wmOperator.
- respect return value of wm_handler_operator_call() and break out of operator handling.
Now you can set/get: float, arrays (int and float) and string.
The only special function is OP_get_string, it is special
because return a pointer to the IDProperty data, so you can't
change/resize/free the string.
It's possible "fix" this with:
1) Return a "const char"
2) Return a duplicate string
All this new function are not in use yet, but i make a simple test
with the "move areas" operator (add a property of every type and then
print the result in the other size) and work fine, more test are welcome.
Other thing to check is the new OP_free_property function, because this
properties are only local to the operator, i choice free all this in the
"exit callback" of every operator (only move areas have property now),
so comment about this are welcome too :)
Also add some notes to the WM_api.h file.
This commit adds split area to the window/screen manager.
- RMB down on area edge activates
- mouse move interactively moves areas through new edge.
- RMB up to confirm action
- ESCKEY or LMB to cancel.
This still crashes in some situations, but I'm on it!
Also will start using new operator property system by bdiego
This is a simple API around IDProperty to store properties
in the Operator, it is really simple and this first commit
just add support for IDP_INT data type.
Take care that this "properties" are not save yet and you get
some "Error totblock" more with this.
I add some notes to the WM_api.h file, please check this,
comment and ideas are welcome.
- crash on start because of unitialized pointer
- Makefile had the python lib on weird location... I prefer all
blender/source/blender/ libs together on 1 place? The Make needs
cleaning... it assumes publisher stuff. :)
- on mouse-over edge, you can drag area borders around.
- note it's a handerized system now, so it updates UI while you
move mouse.
Feedback needed:
- read bottom part of the screen_edit.c file. It's the proposed
method for adding tools and handlers. I think it's close, but
might need some tweaks.
