- User Since
- Dec 25 2013, 10:35 PM (344 w, 6 d)
May 28 2020
Oct 13 2019
Jul 6 2017
Aug 26 2016
Jul 17 2016
Here is an updated draft. Can you update the revision if appropriate?
extend and only_active mode work for all mentioned cases. In the latter case it uses a single element list as suggested, not sure if the implementation is ok, it feels a bit hacky.
Jun 25 2016
The changes about the selection should be very simple to make. One way would be D2068, but i am sure there might be better ways. That is the great thing about open source: Everyone can participate, it allows people to invest their free time to help where they can (there are only a handful devs paid from donations and funds by Blender Foundation atm). However those with more experience with the Blender code base are trusted with making design decisions, which is good because it really promotes quality.
This is off topic but i thought it would be worth sharing in case you are interested.
Jun 23 2016
Totally agree! :)
Jun 22 2016
D2064 is a possible fix. Waiting for review for now.
Can't redo the crash, can you upload a blend file or give precise steps to reproduce the crash?
Jun 21 2016
Similar to T23288.
Have no graphics card, but can confirm identical strong circular bands on i5-2557M with 4 GB RAM running Blender 2.77a on Arch Linux.
Only thing i can guess is that Blender needs some specific functions related to GLX >1.2 and then crashes.
Found this possible workaround but haven't looked into it: http://wiki.x2go.org/doku.php/wiki:development:glx-xlib-workaround
Since X2Go only supports GLX until 1.2 and your client has 1.4 that could be it.
Did you try any other OpenGL apps?
147 should be about GLX (OpenGL Extension to the X Window System).
What version of OpenGL / GLX does your system implement? Check via glxinfo | grep -i version.
Could be that your client requires a more recent version not compatible with what X2Go supports: It seems that X2Go is based on nxagent which supports GLX only up to 1.2.
It also says that Cinnamon 2.2 (Linux Mint 17) is not compatible and reverts to fallback mode.
Jun 17 2016
There are dependency graph cycles: The children of Slot-0-LEADER depend on it as a parent, and Slot-0-LEADER depends on its children via Boolean Modifiers.
This alone should not be a reason for a crash though. Crash only happens for Slot-0-LEADER with Slot-2 or Slot-3 which all use the same mesh.
Jun 2 2016
Oh well Campbell did fix it already ;)
Seems like the rendered region is defined at float precision whereas the border is positioned at int precision via sboxdraw which uses glVertex2i. Apparently the int precision was introduced a few years ago when (i guess) the rendered region was defined differently as well (rBS37ff3).
Can confirm, also on Archlinux with 2.77a.
Does it work if you use OpenGL Occlusion Query (from Preferences -> System -> Selection) instead of OpenGL Select?
May 17 2016
One more thing i found remarkable: If material indices are assigned to face indices as 0->0 and 1->1 it crashes, however if assigned 1->0, 0->1 it does not.
Could not agree more would be great to see this solved! Wish i had more time and wisdom at the moment (currently i am pretty much an apprentice). Would love to look into this as soon as i can but i hope someone experienced can look at this at the moment, as currently i must admit i have a lot of things to do regarding studies and personal things, so at least i could help a little bit here.
May 16 2016
Thanks for the video!
This is really interesting... Basically steps 1–3 seem to be enough to make it crash on double click, in other words:
- Take a mesh of two faces (dont need to be joined, also works with more faces).
- Assign a new material to each face.
-> Crash on double click.
Can reproduce as described by Daniel for
Version 2.77 sub 0 date 2016-04-05 18:12 hash abf6f08
Apr 17 2016
ok makes sense then!
@Bastien Montagne (mont29) thanks for pointing this out about the priorities, will keep that in mind!
Apr 16 2016
Guess this should not be considered a bug, since in fact everything is working fine.
It's more of a design issue, that likely has already been considered by the maintainers.
Assumably a lot of people might get blender via a package manager (at least on Linux) where it might come without python, and unless using Arch Linux it might be an outdated version not fit for Blender. That's why the addon copies the python lib into the executable bundle for Linux and Windows (for Mac OS it seems that its already contained in the Blenderplayer.app), but it tries to copy from a local path assuming python always comes with blender.
Apr 15 2016
What exactly is the error? And please always instead of showing the error (the .blend) describe the exact steps that led to the error.
Does this only happen with Blender?
I must admit i have no experience with multiple monitors maybe someone might have an idea.
So basically all i can guess is that it might just be too much to ask from your GPU.
At least it does not appear to be Blender specific.
It seems the path to the python lib is assumed to be relative to the location of the blenderplayer, which might not be a valid assumption on Linux if you installed via packages without bundled python.
What exactly do you mean by multiple windows? How many, at what resolution, what else is running in the background, does it get better if you close some of the windows?
Could you upload a .blend file to reproduce the issue?
What exactly is the issue?
Please attach a .blend file and describe step by step how to reproduce the issue.
Apr 5 2016
Lets wait for what reviewers say then :)
Apr 4 2016
it seems the idea of the midlevel offset was not integrated in the vertex groups mode, as it was with the texture mode.
Could you check if the points described in T47997 fix the issue?
e.g. from Preferences -> System -> Selection choose "OpenGL Occlusion Queries".
Mar 30 2016
What is your Python version?
e.g. Help -> Save System Info, then just drag the file into the comment text field here to upload.
Still can not reproduce this.
Could you please provide your system information?
Help -> Save System Info, then just drag the file into the comment text field here to upload.
T47997 possibly related?
can not reproduce this on Arch Linux for 2.77 22a2853, please provide the information as mentioned above.
I can not reproduce this issue so far on Arch Linux, Blender 2.77.
What action do you have assigned to the right click?
Is it selection (check whether Preferences -> Input -> Select With is Left or Right) ?
Mar 27 2016
Can't reproduce this either with recent build or build from the specified time, on Arch Linux though.
Mar 19 2016
Can confirm this. Still broken on Arch Linux, currently Blender 2.77 6aeb1f7f.
If you leave the flt_flags return value empty, i.e. remove [self.bitflag_filter_item] * 3, or use a different factor, then it works. But yep since the flag value is the same its likely an issue with its interpretation.
Mar 18 2016
No problem :)
The issue seems to be fixed in the latest build i just did of 2.77 3d59be30.
Until this gets into your package for Manjaro Linux, you might use a build from here: https://builder.blender.org/download/
Or you can use the official 2.76b version where it also seems to work.
Can confirm this on Arch Linux with Blender 2.76.11.
Mar 3 2016
As the operator description says "Select region of faces inside of a selected loop of edges" it currently does not work for loops of faces yet. Selecting an edge loop and then using Select Loop Inner-Region surely is just as comfortable.
Feb 26 2016
[Desculpe, eu só sei espanhol não Português, mas presumo que o Inglês é ok. Traduz Google dá um bom resultado para o texto abaixo, se necessário.]
Feb 25 2016
Seems to work fine with 2.76.11 on Arch Linux using Intel HD 3000 (using Mesa 11.1.2 OpenGL implementation supporting at most GL 3.3 core).
Can confirm, but this is not a bug.
After you have scaled the whole curve object down to 0 along y, trying to scale 0 via the S manipulator will only multiply the value of 0 and not change anything.
You can set the scale manually from the Properties toolbar (press N) to something higher than 0.
Still an issue in current master, this really has been around a long time :)
are you sure you attached the .blend file?
Feb 24 2016
Feb 20 2016
Feb 18 2016
In the construction of the new coordinate system with one axis aligned to the vector, the use of a parallel unit vector should be prevented, however this was not the case due to a floating point precision issue (subtraction of values with too far apart exponents).
The most evident symptom is a division by zero (epsilon margin MT_fuzzyZero comparison) in the normalization of the base vectors of the new aligned coordinate system, in KX_GameObject::AlignAxisToVect.
Aug 25 2015
Aug 14 2015
cant reproduce this from your description, for Blender 2.75 or 2.75.4 on Arch Linux.
Aug 10 2015
thanks for the comments guys :)
Aug 9 2015
Sounds like a very convenient and concise workflow ! :)
Aug 7 2015
does "u f <Enter>" not work for you? that works here on Arch Linux using 2.75a.
Jul 26 2015
Can't confirm for 2.75a on Arch Linux.
Jul 15 2015
can reproduce this on Arch Linux with Blender 2.75.2 dfd383ca.
worked with release 2.75a c27589e.
Feb 26 2015
seems like i lost the momentum too early...
thanks for the fix, i will study the solution carefully :)
Feb 25 2015
does the timeline cursor move when you start the preview?
can confirm this for 2.73a on Ubuntu 14.04.1
afaik mesh structures that are unintuitive are not considered a feature.
however when you select any curve vertex you can toggle 'Cyclic' in the 'Active Spline' panel, so im not sure if it can really be considered unintuitive either, since one might actually explicitly want it to be cyclic, even in such an extreme case. deviations from the value of the 'Cyclic' checkbox in special cases might be more confusing.
btw. sorry for thinking out loud, i guess thats rather unappropriate...
though this issue seems too involved for me atm. im learning a lot in all this process, but i shouldnt go into such detail i feel.
ok so i see that the number of edges to use in CDDM_from_template is
(int)((numEdges * stride) + newEdges + newVerts)
in case of do_shell stride is 2 and newVerts is 0, since it just takes a translated copy of the original mesh and adds the newEdges of the rim.