Bugfix: commit of last week that moved object_handle_update() out
of the view3d code into event loop caused render to crash. This
update should not be called during render. In future it'll even
be thread-locked or better; solved by giving objects an 'owner'
for storage of updates.
This solves a lot of crashes on render.
Useful goodies: most buttons for Image window back.
Not every button works! But what you can do:
- press Nkey to show/hide options
- use curves, with realtime updating
- image properties panel, load, browsing layers,
setting types
- paint panel
- plus new paint color picker panel! (why it wasn't
there in 2.4x is probably obvious, but now it can!)
Hrm... radial control should be added here too, and a nice
paint size cursor?
- Mesh Rip back (Vkey). For those who keep wondering how
it works: just put mouse cursor somewhere close to the
selection, press V and mouse mouse away from it. Feels
like real rip!
- Made extrude sorta work, no menu/options yet though.
But it does transform!
- Added an short event->mval[2] with region coords,
easier coding for the guys :)
- Fill operator standard delivers 'beauty' now
Several things in one commit; could not split this up easily,
one job invoked another, and so on. :)
- Added pulldowns for save/load .blend file in top bar.
- To enable "Save" without further popups (save over)
I've added a signaling function in window header to
indicate a succesful save.
- On any undo push it now signals 'file changed'. This
goes by notifiers nicely, but now registers only the
undopushes, which is quite unreliable. "Changed" state
shows in header as "Blender*" and for OSX with the
standard close button black dot.
- Made screencast show a button in top bar indicating such,
and allowing quit. No hotkey for quit yet... but ESC will
keep casting now.
- Fixed new BLF_init(), which should be in WM_init() and not
on any .B.blend read.
- Fixed CTRL+F3 "Save Screenshot", which was still using old
fileselect code.
Some small fixes:
- handler operator exec gave too quick a warning for invalid area
(set area context to NULL is OK :)
- 4-split 3D view with clip option didn't reset clip when 4-split
was disabled
- creating a new 3d view didn't initialize buttons region correct.
Proper integration of File-selecting in WM. The communication
flow was flawed. :) Main problem was that filewindow can change
the screen context entirely, and should not do this directly on
a call inside an operator. Another problem was that the operator
ownership was handed over to SpaceFile, which is asking for
problems if you want to execute the operator with proper context
later on.
Solution is simple; window handlers already are valid owners of
operators and can manage context, so instead of directly talking
to the 'file space', you give the operator to a new handler this
way:
WM_event_add_fileselect(C, op);
This handler then listens to events (OPEN, EXEC, CANCEL) sent
by the WM or by the filewindow itself. This way local context
operators (like "open new image in imagewindow") will survive
a full-window fileselector fine, and in future also secondary
windows browsing files.
Two bugfixes included in this commit too:
- Add sequence menus in Sequencer used wrong context.
- When handler executes operators, it sets stored context now
by first checking if this is still valid.
Smaller jobs, all in one commit!
- Moved object_do_update out of view3d drawing, into
the event system (currently after notifiers).
Depsgraph calls for setting update flags will have to
keep track of each Screen's needs, so a UI showing only
a Sequencer doesn't do objects.
- Added button in "Properties region" in 3D window to set
or disable 4-split, including the 3 options it has.
(lock, box, clip)
- Restored legacy code for UI, to make things work like
bone rename, autocomplete.
- Node editor now shows Curves widgets again
- Bugfix: composite job increased Viewer user id count
- Bugfix: Node editor, not "Enable nodes" still called
a Job, which didn't do anything
- Various code cleaning, unused vars and prototypes.
- Added depsgraph tag for object-change in AnimData, so the
new animsys doesn't have to all objects anymore.
(Still WIP, depsgraph has to do this much better)
- Bugfix in notifiers; only 1 notifier was handled for
frame updates
Result: 2 windows, displaying 2 scenes, now can be edited
independently, and play independent. Not when they share
data, of course. :)
Scene browse button now works.
Note that animsys currently executes animation for the entire
database, so multiple scenes in multiple windows don't work yet.
Various fixes:
- crash on invoking filewindow when mouse outside (active) window
- removed obsolete error prints (set screen error, copy data error)
- displist fix for loading files with curve/nurbs, the select-outline
then didnt draw immediately.
- outliner allows scene activating
Render back! And not only back, even full threaded now. :)
Current state is unfinished, but too much fun to not to
commit for review and test!
WARNING: because render is in a threaded job, it will
use data as can be edited in the UI. That'll crash in many
cases of course... the idea is to limit UI usage to viewing
stuff, especially for the Image Window to inspect layers
or zoom in/out.
What works now;
- F12 render (no anim)
- ESC from render
- ESC pushes back temporary Image Window
- Render to ImageWindow or full-screen.
- Executing composites, and edit composites after render.
Note that the UI is 100% responsive in a render, you can
switch screens, slide area dividers around, or even load
a new file during render. :) It's quite stable even.
I'll collect all crash reports especially to get a good
picture of where the protection is required at least.
Also added: XKey "Delete Objects", to get things crash...
unfortunately it didn't for me.
Sanitized the 'tweak' event.
Original idea was to have WM event system generating it
automatically. However, I first tested it via a handler
and operator, to check what kind of configurations would
be useful. It appeared to not work nice, also because
that inserting a tweak operator in a keymap is confusing.
Now 'tweaks' are generated automatically, and can be
catched by keymaps as any event. The current definition
of tweak is:
- if Left/Middle/Rightmouse pressed
if event wasn't handled by window queue (modal handlers)
start checking mousepositions
- while mousepositions are checked
- escape on any event other than mouse
- on mouse events:
- add tweak event if mousemove > 10 pixels
- stop checking for tweak if mousebutton released
- Tweak events have a define indicating mousebutton used
EVT_TWEAK_L, EVT_TWEAK_M, EVT_TWEAK_R
- In keymap definitions you can use _S or _A to map to
action or select mouse userdef.
- Event value in keymap should be KM_ANY for all tweaks,
or use one of the eight directions:
EVT_GESTURE_E, _SE, _S, _SW, _W, _NW, _N, _NE
- And of course you can add modifier checks in keymaps for it.
- Because tweaks are a result of mouse events, the handlers get
both to evaluate. That means that RMB-select + tweak will work
correctly.
In case you don't want both to be handled, for example the
CTRL+LMB 'extrude' and CTRL+LMB-tweak 'lasso select', you will
need to set the first acting on a EVT_RELEASE, this event only
gets passed on when tweak fails.
The current system allows all options, configurable, we had in 2.48,
and many more! A diagram of what's possible is on the todo. :)
Also in this commit: lasso select editmesh failed with 'zbuffer
occluded select'. Also circle-select failed.
Moved 'redo last operator' stuff to WM level, with a
WM_operator_repeat()
Code in screen_ops was not checking all operator stuff
correctly, so repeat menu (F3 now) crashed in cases.
Big commit, but mainly adminstration.
- Enabled ot->flag OPTYPE_UNDO to work.
- Removed all redundant ED_undo_pushes, but I'd
recommend everyone to check it while testing. :)
- Added view manipulations as OPTYPE_REGISTER,
although this will flood the redo stack a bit...
Nevertheless; for a "redo last action" panel we
can simply check if both flags are set for redo.
- Bugfix in editmode undo: selectmode was cleared,
so you couldn't select after undo
- Bugfix in mixing tweaks and keymaps... solution
works but is weak, need to think over a while.
Edit Mesh
Added the two other 'select similar' ops, for edge and face.
For testing fun I've put these ones as first keys using the
modifier order feature. So it's now:
SHIFT+G : similar vertex
CTRL+SHIFT+G: similar edge
SHIFT+CTRL+G: similar face
Not very useful here; but I think this can work well to map
a double amount of options under mouseclicks. Power user
config stuff!
Fake commit, but for history I have to write the log!
Fixed CTRL+Q quit-blender. It was using ancient code
still, and since I always use APPLE+Q I never noticed :)
Edit Mesh:
- Added back "Edge Shortest Path select"
It now also does regular selection, more fun!
It's mapped to CTRL+click now, and makes or clears selections
between current and previously activated edge.
Seam/Sharp/etc marking is a toolsetting mode still. These
options cannot become properties easily, because the tool
uses the properties of selected edge to clear...
- Removed a whole bunch of G.f flags, related to mesh drawing.
It's all now local in me->drawflags. Here's the list of
removed old globals:
G_DRAWEDGES
G_DRAWFACES
G_DRAWNORMALS
G_DRAW_VNORMALS
G_ALLEDGES
G_HIDDENEDGES
G_DRAWCREASES
G_DRAWSEAMS
G_DRAWSHARP
G_DRAWBWEIGHTS
G_DRAW_EDGELEN
G_DRAW_FACEAREA
G_DRAW_EDGEANG
- Edit mesh: Add ctrl+click add vertex or extrude.
I've made it not move the 3d cursor in that case.
Also found out tweak events conflicted with existing
keymap definitions; on tweak failure (= no mousemove)
it now passes on the mouse event as 'mouse down' for
the remaining keymaps to check.
These then actually respond to mouse-up instead of down...
The location in the keymaps where tweaks get generated
remains important. Examples:
1 - 'select' mouse-handler, operator return pass-through
2 - tweak handler checks, and makes tweak event
3 - grabber responds to tweak event
1 - ctrl+mouse tweak handler checks, makes tweak event,
or passes event on
2 - if tweak event, it runs lasso
3 - else when passed on, ctrl+click extrude happens
In the first case, select works on mouse-down, immediate.
In the second case, extrude happens on mouse-release, even
though the keymap defined mouse-press.
This will make designing nice balanced keymaps still not
simple; especially because you can't tell operators to
pass on the key... although we can add the convention that
select-mouse operators always pass on to enable tweaks.
Still a good reason to wait with custom keymaps
when this is fully settled!
* Cleaned up UI_interface.h a bit, and added some comments to
organize things a bit and indicate what should be used when.
* uiMenu* functions can now be used to create menus for headers
too, this is done with a uiDefMenuBut, which takes a pointer
to a uiMenuCreateFunc, that will then call uiMenu* functions.
* Renamed uiMenuBegin/End to uiPupMenuBegin/End, as these are
specific to making popup menus. Will convert the other
conformation popup menu functions to use this too so we can
remove some code.
* Extended uiMenu functions, now there is is also:
BooleanO, FloatO, BooleanR, EnumR, LevelEnumR, Separator.
* Converted image window headers to use uiMenu functions, simplifies
menu code further here. Did not remove the uiDefMenu functions as
they are used in sequencer/view3d in some places now (will fix).
* Also tried to simplify and fix bounds computation a bit better
for popup menus. It tried to find out in advance what the size
of the menu was but this is difficult with keymap strings in
there, now uiPopupBoundsBlock can figure this out afterwards and
ensure the popup is within the window bounds. Will convert some
other functions to use this too.
Made 'select-tweak-grab' work in 3d window, fcurve and nodes.
For nodes it works with both left/right, the others follow
user preset for select.
Tweak is a WM event, which also follows user preset for
tablet-style tweaks (release = apply).
You enable tweak events with an operator WM_OT_tweak_gesture.
Keymap entries can assign to left/middle/right button or to
action/select button tweaks.
File reading of 2.50 files fixed.
- it now exits nicely from queues without using free'd memory
(checks CTX_wm_window(C) to be NULL)
- code that matched current (ghost) window with new situation
lacked setting a pointer good.
Overall cleaned some parts too; like context clears now all
wm related pointers when you set a new wm, window or area.
Font object + editing back.
Was quite some work due to a myriad of globals all over!
Works nicely 100% local now.
To enable a single textedit operator, I've added a new
keymap entry KM_TEXTEDIT, which gives all keyboard events
to the handler. Also had to add a new keymap-add function
to force a keymap handler in beginning of region handlers.
In future this can be used to prioritize handlers.
Also: split off the arrow keys (frame change) to a separate
region level handler. Can be set with default flag in
regiontype->keymapflag ED_KEYMAP_FRAMES
* Triple Buffer is now more complete:
- Proper handling of window resize, duplicate, etc.
- It now uses 3x3 textures (or less) if the power of two sizes
do not match well. That still has a worst case wast of 23.4%,
but better than 300%.
- It can also use the ARB/NV/EXT_texture_rectangle extension
now, which may be supported on hardware that does not support
ARB_texture_non_power_of_two.
- Gesture, menu and brushe redraws now require no redraws at all
from the area regions. So even on a high poly scene just moving
the paint cursor or opening a menu should be fast.
* Testing can be done by setting the "Window Draw Method" in the
User Preferences in the outliner. "Overlap" is still default,
since "Triple Buffer" has not been tested on computers other than
mine, would like to avoid crashing Blender on startup in case
there is a common bug, but it's ready for testing now.
- For reference "Full" draws the full window each time.
- "Triple Buffer" should work for both swap copy and swap exchange
systems, the latter still need the -E command line option for
"Overlap".
- Resizing and going fullscreen still gives flicker here but no
more than "Full" drawing.
* Partial Redraw was added. ED_region_tag_redraw_partial takes a
rect in window coordinates to define a subarea of the region.
On region draw it will then set glScissor to a smaller area, and
ar->drawrct will always be set to either the partial or full
window rect. The latter can then be used for clipping in the 3D
view or clipping interface drawing. Neither is implemented yet.
Added WM Jobs manager
- WM can manage threaded jobs for you; just provide a couple
of components to get it work:
- customdata, free callback for it
- timer step, notifier code
- start callback, update callback
- Once started, each job runs an own timer, and will for
every time step check necessary updates, or close the
job when ready.
- No drawing happens in jobs, that's for notifiers!
- Every job stores an owner pointer, and based on this owner
it will prevent multiple jobs to enter the stack.
Instead it will re-use a running job, signal it to stop
and allow caller to re-initialize it even.
- Check new wm_jobs.c for more explanation. Jobs API is still
under construction.
Fun: BLI_addtail(&wm->jobs, steve); :)
Put Node shader previews back using wmJobs
- Preview calculating is now fully threaded (1 thread still)
- Thanks to new event system + notifiers, you can see
previews update even while dragging sliders!
- Currently it only starts when you change a node setting.
Warning: the thread render shares Node data, so don't delete
nodes while it renders! This topic is on the todo to make safe.
Also:
- bug in region initialize (do_versions) showed channel list in
node editor wrong.
- flagged the channel list 'hidden' now, it was really in the
way! This is for later to work on anyway.
- recoded Render API callbacks so it gets handlers passed on,
no globals to use anymore, remember?
- previewrender code gets now so much nicer! Will remove a lot
of stuff from code soon.
* Drawing code from wm_event_system.c split into separate wm_draw.c file.
Now there's 3 different draw methods implemented, not sure what survives
or will be added but is useful for debugging.
* Draw All: redraws everything each time, for reference.
* Draw Overlap All: what the code did before this commit, only draw
regions marked for redraw, and anything that overlaps them.
* Triple Buffer: copies/retores all area regions into a texture, and
blits that before drawing. Menus, brushes, gestures, etc are redrawn
always on top of that.
Currently "Draw Overlap All" is set hardcoded to be used still. Triple
Buffer code is not complete, it doesn't handle window resize yet. Cards
that don't support non power of two textures can need quite large
textures as well, this could be split into multiple smaller ones.
Python operator api was using WM_operator_name_call() which was confusing things too much.
Added WM_operator_call_py() which ended up being a very small function and split out operator creation into wm_operator_create()
Python operator now runs the poll() function and raises an error if it fails.
Eventually there should be error messages for poll that python can use to give the exact reason for failing (eg - library linked data, no active object...)
(when in debug "-d" mode only)
copy & paste duplicate and transform operations can now be copied from user input and pasted into ./test.py and run with the Pkey (fixed some minor bugs preventing this)
Would be nice if the "mode" setting used a proper RNA Enum rather then an int.
# example, duplicate and transform
bpyoperator.OBJECT_OT_add_duplicate(mode=1)
bpyoperator.TFM_OT_transform(mode=1, options=0, values=(-1.23989, 0.570745, 0, 0), constraint_orientation=0, constraint_mode=0, constraint_matrix=(0, 0, 0, 0, 0, 0, 0, 0, 0))
- Added shift+d duplicate for object and editmode mesh.
Note it uses WM_operator_name_call(), which is fine now,
but in future might put again 2 undo's and operators on
the stack.
Will have to spend some time on how Macros will work!
- added itterator CTX_selected_editable_objects()
(named it first "edible" but that was too funny!)
Also cleaned object_edit.c to use this correctly.
- added CTX_wm_view3d(), especially for hybrid tools
that *can* use view3d, but don't have to.
- moved debug -d print for operators to the real invoke call
Testing commit; this puts back support for swap-exchange graphics
cards, which I had hoped to have faded out... but it appears it
still does it for intel and some atis.
This only swap-exchanges properly for areas/regions, not for
menus or the 'action zone triange'. Let's see if it works!
You enable it with starting with commandline option -E
Vertex Paint back!
Added WM level "paint cursor" system, which manages a custom painting
cursor for tools or modes.
- Activate it with WM_paint_cursor_activate(). That function wants two
callbacks, a poll(C) to check whether there's a cursor in given context
and ARegion, and a draw(C, x, y) which gets called when appropriate.
- While paintcursor is active, the WM handles necessary redrawing events
for all regions, also to nicely clear the cursor on region exit.
- WM_paint_cursor_activate returns a handle, which you have to use to
end the paint cursor. This handle also means you can register as many
custom cursors as you want.
At the moment, vertex paint mode registers only a mousemove handler,
all other events are still normally handled. This is stuff for the
future todo.
Bugfix: the area level Refresh callback was not getting Context set,
it was NULLed still. This fix allows bone selecting and seeing updates
in the action editor. Not other way around though, will investigate.
Small cleanup in region-based cursor handling.
- callback was in spacetype, now in regiontype
- made screen listener catch ND_MODE notifier and
call the active region cursor callback, if it's there.
WM: added area-listener, which can be used to tag refreshes for
either the drawing call, or use the new ED_area_tag_refresh()
function which will automatically call, after all notifiers
were handled, an the spacetype->refresh() you provided.
Added for Joshua, after reviewing Action/Dopesheet requirements.
Joshua: I've made two dummy functions in space_action.c:
- action_listener()
- action_refresh()
Wich now does a printf on activating a new object.
* errors in python called operators are raised as errors
* Python defined operators errors are reported as errors (not full traceback yet)
* added BKE_reports_string, same as BKE_reports_print but it returns a string rather then printing it.
* WM_operator_name_call optionally takes an initialized report struct
* Object has some more properties wrapped, mostly game related.
* Scene frame changes now send a notifier.
* Added functions to create/free operator properties for calling
operators. This also simplifies some duplicated code that did
this. Ideally though this kind of thing should use the properties
pointer provided by buttons and keymap items. Example code:
PointerRNA ptr;
WM_operator_properties_create(&ptr, "SOME_OT_name");
RNA_int_set(&ptr, "value", 42);
WM_operator_name_call(C, "SOME_OT_name", WM_OP_EXEC_DEFAULT, &ptr);
WM_operator_properties_free(&ptr);
- Put back vertex/edge/face select
- Mode menu allows going in/out editmode
- Tested some tweak event stuff in WM, made tweak event follow
user preset for 'action' or 'select' mouse.
(Wanted to try this for transform, but better get advise from
Martin first :)
- Made scrollwheel behave uniform for zooming. It uses the UserDef
to map what's in our out. Use keymap entries WHEELINMOUSE and
WHEELOUTMOUSE to trigger this. Also removed the feature to store
in event->val the direction, it was conflicting with KM_ANY.
- Added more default poll callbacks, use for generic ops things
like ED_operator_object_active.
- Edit mode Mesh undo/redo back
(undo for editmodes needed recode, todo for curve/lattice/etc)
- Added border select for edit mesh
- Added Akey (de)select all for edit mesh
- Added notifiers for mode changes.
This is also the first trial to dynamically add/remove
keymap handlers, based on modes. For that reason the Object
keymap was split in 2, modal and non-modal. On TABkey, the
view3d listener removes and adds maps.
The view3d keymap still handles generic mouse/border selection.
Internally it will verify modes.
The modes are not re-implemented still... have to move this
to scene context.
Editmesh code cleaned and compiling/linking. A whopping
20k lines back! :)
Not that it does stuff... editmode in/out has to be done,
and loads of operators. Also linking/exporting editmesh
calls has to be reviewed.
Also: added a blender_test_break() mechanism in BKE.
* Added a report list to operator, to which they can report errors and
warnings. When the operator ends, it will display them with a popup. For
python these should become exceptions when calling operators.
* Added a function to make a popup menu from a report list.
* Also added a utility function to prepend a string before the reports to
indicate what they relates to. Also made the report functions used
BLI_dynstr to simplify the code.
* Made file reading and writing report errors to the user again using this
system, also replacing the left over uncommented bad level error() calls.
Anim playback part 1 (needs more test, will do after commit)
- added the update_for_new_frame() back
- proper evaluation of time change notifier in WM level
- fixed redraw flushes for menus while animation plays.
Notifier system upgrade, based on Brecht's doc.
Implementation notes:
http://wiki.blender.org/index.php/BlenderDev/Blender2.5/DataNotifiers#Implementation
In short: let's try to strictly communicate in a notifier what happened.
The listeners then can act themselves. It also means that a frame-changed
notifier doesn't send out redraws, the editors themselves can decide.
- Added 3d cursor operator (no exec, waiting for vector property!)
- Added default editor callbacks, for the poll() function.
Example:
ot->poll= ED_operator_view3d_active;
I'll add later more, per module, such as OBJECT_active, which
can be called without WM active
Brought back 'smooth view'. Took some hours to untangle it all,
code was spread all over, instead of localized in 1 call. Tsk!
Still not perfect, but at least more in control. For the hackers;
check void smooth_view() in view3d_view.c, here all 3d view
stuff should be handled, so it can optionally use animating.
For the users: 'smooth view' now plays at a maximum of 30 hz,
and doesn't block anymore. So even slow animated views remain
responsive if you press many numpad keys.
