It still crashes with TdrDelay of 8 after reboot.
I just noticed that notifications pop up a bit before the process terminates, saying that access had been blocked to the graphics hardware for blender.exe:

Use active bone of active object for all objects.

See my previous comment https://developer.blender.org/D3330#78857

In D3394#78884, @Nick Milios (semaphore) wrote:

The sugested way is to report only if the same error occurs for all objects(with selected bones).

@Fulk33 You lost me. Now the operator does nothing if action/type is one of the following:

• SRT_VIEW_ZAXIS
• SRT_VIEW_XAXIS
• SRT_CURSOR_DISTANCE
• SRT_MATERIAL
• SRT_RANDOMIZE
• SRT_REVERSE

According to Google Translate, it is about changing/converting units to millimeter (but?) Blender uses centimeters.

@Dalai Felinto (dfelinto) I don't understand why SRT_SELECTED would be an exception.
There is at least one selectmode, and if there is nothing selected for any of the active modes, does sort_bmelem_flag() actually change something under certain circumstances?

Same problem here, Windows 10 Pro 64bit, build 1709 (not fully up to date, Windows Update broken for unknown reasons) and an Nvidia Geforce GTX 970 with latest drivers.

In D3379#78669, @Fulk33 wrote:

If you look at T54643 under Differential Revisions you can see which operators are already being worked on. I this case, there already is a diff D3330.

Submitted a patch for MESH_OT_sort_elements: https://developer.blender.org/D3379

Has anyone thought of allowing users to color-code (groups of) objects?

@carlos (c17vfx) It's also pretty common to place (at least the) visibility toggles to the left of the entries, generally speaking (like checkboxes).
That applies to RTL languages only obviously. In an LTR layout, I would expect them to be on the right side instead.

I would interpret the lack of a good visual solution as a general design issue.
Why do we need a fake user toggle in the first place? To stop content from being removed if it's not referenced by anything?

People complain about truncated bpy access paths in tooltips - like bpy.data.screen...use_manipulators, is that what you mean Jonathan? If no, please add it as a to-do. It should wrap if necessary, preferably at periods, but shouldn't matter much.

Error occurs at least on Windows 8, 64bit in every version since the introduction of sort(), which was 2.64 (had to adapt the test script, because update_edit_mesh() didn't exist back then). I tried 32 and 64 bit releases.

All it prints to console is "Error: EXCEPTION_ACCESS_VIOLATION".

Don't get me wrong, your patch is great! It's just that I've seen ready patches lurking around for months or even a year. So the emphasis was on "soon"!

Fingers crossed for this to land soon!

I would argue that deleting files does not need to be at your fingertips (literally), users should be able to figure out another shortcut (delete key, backspace too if you will).

Great, thanks! Wasn't immediately apparent that the description attribute is used for the abbreviated names.

IIRC, it casts a lot of rays to determine intersections, with the ability to select intersecting faces. I wonder if it could utilize BVHTree to increase performance (overlap?)

Can Python scripts access the abbreviated key name too? (PAGE_DOWN > PgDown)

What about the 3D print toolbox addon? Shouldn't it utilize this new feature?

Sounds sweat, thanks Severin!
I wished one could also use letter keys for quicksearch in file names. But that require to remove all existing key bindings like X to delete files (not the worst step to take however).

What's the status of the patch?
The parameter could also be called propagate_values IMO.
Also interesting would be auto-increment as default value, so assign unique numbers (IDs) to vertices and disable value propagation...

Why not register_module? Where else should register_class be preferred?

If split normals are not available (yet), that's fine. Should be made clear in the description of the bmesh function however. Vertex normals work correctly as far as I can tell.

Better silly than sorry ;-)

Checked the curve code and must say, I'm not very eager to implement it. Curve code is really alien to me.

@Joshua Leung (aligorith): are you planning on rolling custom tessellation code to allow PacMan? There are two algorithms used elsewhere, scanlines and ear-clipping (ideasman probably knows best)

The GridCell is only used to give the 3 properties it has names, but in fact, a plain list could be used if you used numeric indices instead.

quite early for spring cleanup

coll_item is indeed passed to the callback. I wonder if a reference to pfloat should be passed instead however, because that is the property the callback is registered for.

I wonder if FPS walk mode + on-screen instructions and navigation controls would be better. A single small indicator in the right upper corner of the 3d view could start it. Editing and all keyboard shortcuts other than for navigation should be disabled while that operator is running (maybe add *File > Open (View only)* option?)

So it's the ray_cast() method and I believe it is caused by object being in edit mode - there's no data for raycasting in that mode.

Is this the error you are getting?

No error in 2.71 RC2 on Windows 7, was probably solved by Python 3.4 upgrade.

Re-opened, because macro operator arguments can't be accessed.

"when next used" means when it's displayed in the sequencer preview? Or is there anything like the refresh_all operator that can trigger the update?

Re-opened, because it's not only a problem if called from another operator, but rather any call to it (Python console / Text editor) does nothing, and Redo support is broken as well.

The solution is to set an amplitude and a period, doh!

I quickly went over all official addons, and none seems to check the return value (boolean) of keyframe_insert(). Thus, I don't expect any problems on the python-side by this change.

The scene stats should be moved to another location to be honest, or there could just be a button in the info bar to open a popup. In many UI proposals, they are shown as text overlay in the 3D View (either of the corners), that's definately another good place, although some othe editors may need to overlay it as well (e.g. UV editor I guess?).

Makes sense, as the input mode is shown, it's absolutely correct!

I'm not familiar with the easing itself, but the Bounce and Back icons feel odd - I expect them to be mirrored (left and right flipped for Bounce, top and bottom for Back).