Saving and loading gpencil was using different order for the individual list items.
On a 120 Mb gpencil project (yes, animators!) loading time went down from 1 minute
to a second or two.
Note that this to have effect, you need to save once.
Developer note: check this commit, it uses a new writelist function. You can
speedup stuff tremendously with keeping saved and read data in sync.
This is as close as I can get to keeping the old code intact. After this
commit, I will have to change existing code paths, making testing of
functionality harder.
Changes:
* Keep only projective texturing code in paint_image_proj.c
* Move 2D code to paint_image_2d.c. This needed the introduction of
allocation/cleanup functions for the relevant structures.
* Common code interface for both modes stays in paint_image.c (which
still includes all old code, system should work as it did with the
exception of non-projective 3D paint mode) and is made public. This is
not a lot of code, only rectangle invalidation and undo system.
* Changed the naming in the new code slightly: imapaint_ prefixed functions refer to
common functions used by both systems, paint_2d_ prefixed to 2d
painting. There will be an interface for the projection painting as
well. Probably there is some leftover naming conversions to do.
TODO:
* Move operator init/exec/modal to common interface file
* Get rid of old BKE_brush_painter_paint, now brush_painter_2d_paint.
All code uses stroke system for the stroke management
* Write space pressure management for the paint stroke system (for other
systems to access as well :) )
* Move texture paint tablet presssure exception code for old bugs to
stroke system (makes me wonder...aren't other systems also influenced by
these pressure issues?) or up in the function hierarchy inside texture
paint. This code is still not there so users with tablets may notice
some issues.
* possibly change other systems to pre-multiply pressure with the
relevant influenced attributes in the stroke function. This could get
tricky though and it's possible that it could backfire.
- activate from spacebar search (3D Ruler)
- ctrl-click adds new rulers
- clicking in the middle of a ruler, turns into protractor, dragging out of view snaps back to ruler.
Adding new file paint_image_proj.c which includes the projective texture painting part of texture
painting, using the stroke system. To access the new code path use Shift-LClick.
The new code path still is problematic with tablet pressure and I will be looking
into ways to unify this across paint systems next.
The old code is still present and can be accessed by regular Lclick as usual.
Also removed 3D (non-projective) painting from 3D viewport.
TODO:
* Add pressure influence code to stroke, remove from every other paint
system code, including texpaint.
* Put UnifiedPaintSettings update in PaintStroke code.
- Dopesheet need to be updated when adding or switching
between objects.
- After removing object it shall also be tagged for update,
otherwise crash will likely happen.
This was from initial BMesh merge, but should not have been added in since face normals are calculated and stored in the DerivedMesh.
Toggling editmode would remove poly-normals so its unlikely anything relies on this custom-data.
errors with the variables/targets, even if those errors are for variables which
aren't used (and are hence "harmless" errors)
This means that the filter can be truly useful for helping locate things that
need "cleaning up". For example, previously, there could still have been drivers
where there were some of these "harmless" errors would emit warnings, but would
otherwise appear perfectly functional.
The implementation here uses a slightly slower method of checking any errors in
these driver vars. However, it's no slower than what's done when these are
evaluated, and should be less error prone than introducing yet another type of
error tagging for this one case. The problem here is that the "driver invalid"
flag, which is usually set when a target has errors, gets cleared by the
pydrivers code if nothing went wrong when evaluating the expression. Removing
this clearing step will probably open a can of worms, so unless this method
proves to be far too slow, this simpler fix will do.
you to fix the RNA Path in-place
For F-Curves that are disabled or marked as having errors because their paths
are invalid (indicated with a red line underneath their names), it is now
possible to use the Ctrl-Click renaming functionality to bring up a textbox for
fixing the offending RNA Path "in place" (i.e. in the channels list) without
having to bring up the properties region first.
This makes it easier to fix the paths if you know what you're doing. However,
caution is still advised for most people. In particular, be aware that this uses
a separate "RNA Array Index" for indexing into array properties (i.e. location,
rotation, color) which will not be shown here, and can only be edited from the
panel (or datablocks editor/scripts).
modes, so we can show degrees rather than radians. Still refers to the same
DNA variable to keep backwards compatibility.
Patch #33807 by Gottfried Hofmann.
Fix for 2.66a
The new Copy/Paste objects feature could hang in eternal loop.
Only happens for objects that refer to another scene via linkage.
This fix then crashed Blender, needed to add a NULL check for screens.
Compositor relies on correctly tagged invalid links, but the nodeInternalRelink function used to replace reroute nodes in localization did not correctly take this flag into account. If a node replaces a link with an invalid upstream link the resulting link must also be flagged invalid.
from Lawrence D'Oliveiro (ldo)
Get rid of BLI_splitdirstring, replace with calls to BLI_split_dirfile, BLI_split_dir_part and BLI_split_file_part as appropriate.
from Lawrence D'Oliveiro (ldo)
- Better argument name for BLI_cleanup_path and BLI_cleanup_file
- remove redundant BLI_add_slash already done by BLI_cleanup_dir
from Lawrence D'Oliveiro (ldo)
- storage.c: Simplify BLI_dir_contents and make it and its internal subsidiary routines reentrant
- Moved common code for disposal of a struct direntry to new routine BLI_free_filelist in storage.c, and put calls to it in interface_icons.c and filelist.c
- Took out inclusion of BLI_fileops_types.h from BLI_fileops.h and put it explicitly into .c files that need it (which turned out to be only 7 of the 35 files that were including the former)
from Lawrence D'Oliveiro (ldo)
More use of bool type, necessitating adding inclusion of BLI_utildefines.h, or moving it up in the inclusion order if it was already included, in various places
- storage.c: make some variables only used in bli_builddir local to that
- storage.c: BLI_file_descriptor_size should allow 0 as a valid file descriptor
- path_util.c: make pointers to non-reentrant storage returned from folder routines const, necessitating making variables holding these returned pointers const elsewhere as well
- path_util.c: BLI_string_to_utf8 closes iconv context in case of conversion error
- blf_lang.c: fill_locales routine now has its own "languages" local variable to construct paths (was stealing internal storage belonging to BLI_get_folder before)
on mouse event coordinates. Every paint mode operated on and converted
back to region coordinates, texture paint does so too so this will help
in unifying the systems.
ALERT! POSSIBLE BREAKING COMMIT:
I have tested all paint systems that use the stroke code and they look
like they work well but I would appreciate it if more eyes could test
this.
Issue was caused by the way how pattern sampling happens in case of
anchored display: track offset is applying on search buffer which
means offset is rounding to an integer. Fractional pat of offset was
completely ignoring which lead to jumps in pattern buffer.
This was only a visualization issue in track preview widget.
