Blender Git Commits

operator ED_OT_undo_push, needed for editmode undo/redo glitch fix, (coming next).

Two minor fixes:

* Fixed a bug in MaterialF0D::operator() that an error condition was ignored due
to a wrong return value.

* Fixed a typo in the docstring of StrokeVertex::setPoint().

Consolidation of stroke drawing.

Fixed a complicated bug that caused a failure of CurvePoint::getFEdge()
which had affected a number of C/Python API functions such as MaterialF0D.

The current view map building procedure may generate ViewEdges whose
two-dimensional (2D) length is almost or exactly zero. Such a zero-length
ViewEdge is possibly chained with other ViewEdges to form a stroke. When
the stroke is finally generated by Operators::create(), an attempt to remove
redundant vertices at the same 2D point is made. This possibly breaks the
links of ViewEdges on top of which the stroke has been built, and eventually
result in a fatal error of CurvePoint::getFEdge() when API functions that
rely on this method are called from within a style module.

The present fix addresses this issue by automatically removing zero-length
ViewEdges (and Chains of them) before stroke drawing is started and after
splitting is performed (e.g., using Operators::sequentialSplit()).

Bugfix #25803 (LibLoad'ed scene objects don't have timer property updated)

The KX_Scenes' SCA_TimeEventManagers were not being merged. These handle the timer prop

Fix for bug #25804 (LibLoad'ed scene objects material CRASH when doing mat.getShader())

X_Material's mScene member variable wasn't properly set on merge if it didn't have an mBlenderShader defined. This patch always makes sure it gets set on merge.

Editbone length was still using "normalize()" instead of "normalized()"
for setting length, which was thus failing and erroring out.

Accidentally left in a print for commit 34912

Fixed do_versions for node groups which contain nodes that have changed sockets. do_versions is running before the lib_verify_nodetree function updates socket lists of nodes. This means that when adding unlinked sockets in do_versions to restore older node groups, the new sockets are not taken into account. To fix this a temporary node tree flag has been introduced, which delays actual group socket do_version until the group tree internal nodes have been updated. After that the unlinked group sockets are exposed (like old node groups did), then the external sockets look up the new index, so that external links to group instances are preserved.

Bugfix #26116

Node materials with 'ray transparency' set now work again.
Fix provided by Ervin Weber, thanks a lot!

Finally got around to fixing priorities, so cleanup in stubs.c can be done. blenderplayer builds again with cmake.

patch [#26146] debian/rules: Avoid unnecessary downloading
--- from the tracker
It's annoying that build_debian.sh downloads the source tree to create tarball every time we build a deb package.

fix_debian_rules_avoid_downloading.patch modifies debian/rules so that the tarball will be created from local repository if possible.

Big node groups improvement patch. Node group trees now have their own lists of input/output sockets. Those can be linked to internal nodes just like links between regular nodes. In addition group sockets can be renamed and have a defined order, which can be modified, and they can be removed again.
More details can be found in the patch tracker description (#24883) and on the code.blender.org development blog.

build python module without binreloc, add dummy argv[0] to initialize bprogname.

And the last one. hopefully.

And another one.

Missed this in battle against BLO vs MEM storm.

I swear, it was just an innocence change in guardedalloc!

The butterfly wing flap, causing a nice storm in the rest of blender.

Now all dependencies should point ok again. CMakers, do double-test.

add includes to cmake files

doxygen: some code shuffle to prevent clashing file names, ensure MEM_guardedalloc.h also shows in docs

remove BGE's PyObjectPlus.isA method docs, was removed in favor of built-in isinstance(), report [#26080].