In T65843#702373, @Philipp Oeser (lichtwerk) wrote:

Can confirm.
current_character and select_end_character are directly pointing to text DNA curc, selc (and these can indeed inherit the "error" of unicode characters having a differrent length that one...)
Guess this could be resolved by making a rna property func that takes these into account using txt_utf8_offset_to_index or BLI_str_utf8_size...
Can check [might take a bit though...]

Confirmed, checking...

As far as I can tell the solution recommended by @Joseph Eagar (joeedh) works in this case. I will test it further but I was able to overcome the hard crashes I have been facing when ediitng the lattice in an operator.

Okay, looks like the problem is with the undo system; if you manually exit editmode then push the understack, e.g.:

It seems like this bug not just affecting undo delete but also it crashes Blender after undoing lattice naming and other Lattice objct related operations

Thanks for conforming the bug. Do you know a work around with the API so I can avoid the crash until this is fixed?

Backtrace is a bit misleading [since this is reproducable without animation_nodes enabled], but can confirm, checking...
Deleting from the UI has a valid Lattice->editlatt in lattice_undosys_poll / editlatt_object_from_context whereas doing this from python doesnt...

closing as archived. @nBurn (nBurn) feel free to double check.

Since it is not a bug (but a known limitation). Better close and continue on Devtalk.

Ah! I see...this explains why if I add all my code to the draw() function I see everything as it should be.Because the draw() function is what is getting the view3D OpenGL window context.
The problem is that I have no idea how to implement your suggested solution.

The problem with the code is that it disregards the fact that Blender works with more than one OpenGL context.
There is the OpenGL context offscreen and a context for each window.
The callback added to draw_handler_add runs in the offscreen context, but the rest of the code runs in the context of the window.
VBOs, Programs, Textures are shared between different contexts. But the VAO does not.
So a solution would be to create the VAO inside the callback with a lazy initialization.

alright, figured id make this easier to test, here's a .blend with the script already loaded and ready to run, just make sure you change the filepath variable or make sure you have a 'video.mpg' file on the root of your D drive

This is now addressed by the re-designed python API of the related calls. See the depsgraph python examples in the documentation.

The issue here is the access to the non-evaluated dependency graph. This is now not possible after rBe693918d407. So now you need to do context.evaluated_depsgraph_get() when you need to communicate to it.

To me this seems non-standard enough to not even have it as a preference, unless it's common in other REPL apps.

Hi, I'm aware of the possibility of adding emacs/bash/readline key shortcuts but would prefer to avoid this for the default keymap.

Sounds reasonable, will check with others though [and commit if agreed upon]

We discussed solutions to this type of problem, @Sergey Sharybin (sergey) will probably be the one to implement it.

Wow :D my first python fix! Thank you a lot, you are awesome!

also note there was very similar T61264

I just tested today build 4th of april, and this file no longer crashes. It seems bug was fixed.

I think that the root cause it probably the same as in T60766

@Jose Conseco (joseconseco), in the future, please use the bug reporting form:
https://developer.blender.org/maniphest/task/edit/form/1/

Backtrace:

Thread 1 "blender" received signal SIGSEGV, Segmentation fault.
0x00007fffeba89c17 in ?? () from /lib64/libc.so.6
(gdb) bt
#0  0x00007fffeba89c17 in ?? () from /lib64/libc.so.6
#1  0x00005555592b8755 in copy_m4_m4 (m1=0x7fffbc6a5a2c, m2=0x40) at /home/zed/prog/blender/source/blender/blenlib/intern/math_matrix.c:88
#2  0x0000555558015a95 in setTransformViewMatrices (t=0x7fffbc6a5708) at /home/zed/prog/blender/source/blender/editors/transform/transform.c:235
#3  0x00005555580762e9 in initTransInfo (C=0x7fffdc058188, t=0x7fffbc6a5708, op=0x7fffbc5dfec8, event=0x0)
    at /home/zed/prog/blender/source/blender/editors/transform/transform_generics.c:1681
#4  0x000055555801b1ac in initTransform (C=0x7fffdc058188, t=0x7fffbc6a5708, op=0x7fffbc5dfec8, event=0x0, mode=1)
    at /home/zed/prog/blender/source/blender/editors/transform/transform.c:2340
#5  0x0000555558086e01 in transformops_data (C=0x7fffdc058188, op=0x7fffbc5dfec8, event=0x0)
    at /home/zed/prog/blender/source/blender/editors/transform/transform_ops.c:370
#6  0x0000555558087092 in transform_exec (C=0x7fffdc058188, op=0x7fffbc5dfec8) at /home/zed/prog/blender/source/blender/editors/transform/transform_ops.c:463
#7  0x0000555557e63e48 in wm_operator_invoke (C=0x7fffdc058188, ot=0x7fffd6dfab88, event=0x0, properties=0x7fffffffcd90, reports=0x7fffb97fabb8,
    poll_only=false, use_last_properties=true) at /home/zed/prog/blender/source/blender/windowmanager/intern/wm_event_system.c:1353
#8  0x0000555557e6464b in wm_operator_call_internal (C=0x7fffdc058188, ot=0x7fffd6dfab88, properties=0x7fffffffcd90, reports=0x7fffb97fabb8, context=6,
    poll_only=false, event=0x0) at /home/zed/prog/blender/source/blender/windowmanager/intern/wm_event_system.c:1586
#9  0x0000555557e64899 in WM_operator_call_py (C=0x7fffdc058188, ot=0x7fffd6dfab88, context=6, properties=0x7fffffffcd90, reports=0x7fffb97fabb8,
    is_undo=false) at /home/zed/prog/blender/source/blender/windowmanager/intern/wm_event_system.c:1664
#10 0x00005555585af25d in pyop_call (UNUSED_self=<module at remote 0x7fffd7068688>, args=Python Exception <class 'gdb.error'> No type named PyUnicodeObject.:
)

Since last asking for information it has been 7 or more days, due to the policy of our bug tracker we will have to archive the report until the requested information is given.
(@Michael Demetriou (qwazix) : just reopen if you've checked and issues persist)

This was really fast. Thank you very much, Campbell!

Solved in rB22dc248f14be: Fix T56610: crash saving images with python API..

It's nice to hear you say that something is easy, related to node trees haha

@Jacques Lucke (JacquesLucke) I think this is fairly easy to add (just call the function adding sockets to group's nodetree instead in RNA code), trying it currently…

I'm pretty sure that this is not supposed to work.
I actually don't know the API to add sockets to groups, but I guess it has to be done "from the inside".

Confirmed, bpy.context.scene.node_tree.nodes['Group'].node_tree.inputs.new('NodeSocketColor','stuff') is working as expected (the direct access from Group node's inputs should be a shortcut to that one I think, or not allow this operation at all, checking code…)

Right, modal keymap handler isn't guaranteed to get all events - other modal operators and popups can swallow the events before they reach this modal operator.

I will have to check, but this might take a while.

Still should not crash I guess..

Is this still happening in blender 2.8?

I can confirm that there is something weird going on. If I open blender and paste in that script and run it and then use the Props Dialog, blender soft hangs for me as soon as I press the OK button.

Solved by Blender developers themselves in https://developer.blender.org/rBcae89a6889e0fc03130a56b6bf84e257eaff0fec

More than a week without reply or activity. Due to the policy of the tracker archiving for until required info/data are provided.

More than a week without reply or activity. Due to the policy of the tracker archiving for until required info/data are provided.

I just noticed when I add 'USE_EVAL_DATA' to operator bl_options
bl_options = {"REGISTER", "UNDO","USE_EVAL_DATA"}
Then crash won't happen. I'm not sure if this bug should be marked as fixed then.

This bug has been open a month without being addressed, so I'm going to use Phillip's suggested Python workaround. A fix before 2.80 ships would be appreciated!

The commit that caused this was reverted, another solution without this problem should be implemented.

CTX_wm_region(C) returns NULL in uiTemplateList(..) directly after the file selector is closed.
When I skip the function in this case, everything goes back to normal on the next redraw.

of course. Just wasn't sure what's easier to debug. I tried to strip it down as best as I could:

I can reproduce it. Can you try to remove everything from that script that is not necessary to reproduce the bug? That would make debugging a bit easier.

By the way, some functions (rna_Object_select_get, rna_Object_visible_get, ...) are missing the view_layer parameter and always get the selected/visible/etc. state from the current context's view layer, while other functions (rna_Object_holdout_get, indirect_only_get) require to specify the view layer explicitly. Is this an oversight or an intentional design?

In T61923#627800, @Jon Seigel (jseigelis) wrote:

Oh, my bad, I didn't mean to step on your toes.

Naah.

Oh, my bad, I didn't mean to step on your toes.