- User Since
- Mar 15 2015, 2:34 PM (200 w, 5 d)
Thu, Jan 17
I have only made a few formatting adjustments and rebase to master. Do whatever you want with it :-)
The mode is global in Blender …
Yeah, I already saw the discussion about that.
Sorry not native writer here:) I meant the global state of the active tool. I think it feels strange to use your pointer in different editor types with one global state (or in this issue, the global status of the active tool changes by changing an editor type in a split view). For example you can so easily get broken UIs like this
@William Reynish (billreynish) I edited my comment before I saw yours. Play 10 minutes with different tools in different windows and different modes.
I think it feels broken in many cases. For example, he's talking about changing one window from Image Editor to Shader Editor …
Switch left side window from Image Editor to Shader Editor. TEXTURE SLOT feature on PROPERTIES-Tool AREA clears and you are unable to return to it even after going back to Image Editor.
Tue, Jan 15
Mon, Jan 14
Fri, Jan 11
It's broken since switching to OpenSubdiv (rB9abcf56fa88dd)
Mon, Jan 7
@Howard Trickey (howardt) by the way, when I look at your patch, with RNA_def_property_ui_range (prop, 0.0f, 100.0f, 5.0, 2); I would expect, that dragging over the value-bar should change it with the step value. It's mentioned as a mystery in source/blender/makesrna/intern/rna_define.c:1444. Is this intended or something that could be fixed?
Sat, Jan 5
Now I get a segfault instead of a black render when opening a new file, activating freestyle, then rendering. Still macOS.
This can be a driver problem, submit a new bugreport?
Works perfectly also in Linux:
However, it's easily possible to sneak-in incompatible shaders everywhere. The world shader nodes have a similar problem T59747.
I can confirm that. Looking at the file modifications times, exiting an already saved Blender writes the correct btx file to the hard disk. CTRL-S just writes the old unmodified displacement version.
Wed, Jan 2
Tue, Jan 1
there's something I'm not doing right
Yes, you've set the render border, clear it with ctrl+alt+b
It crashes with eevee by trying to create the material thumbnail, a simpler file
It's also a mystery to me why anyone would implement it that way.
After deselecting all objects, I would expect to select the top object first, the second one behind by a second click and so on.
However, it seems to work the same as in v2.79 (it selects the last active object, now with alt-a and left-click):
Mon, Dec 31
The mesh cache is here unhappy with a texture and non-existent UV geometry.
A simpler file:
Sun, Dec 30
If technically possible, if there is enough room for the text on the left side of the checkbox, this would be nicer (more consistent):
Appreciate the help. So that's because of the blank material thingy ? How do I get rid of it ?
Sat, Dec 29
This and similar issues are caused by trying to access an empty element inside the material array:
Fri, Dec 28
Worked: Before the new "Quality" setting was introduced in the Subdivision Surface modifier
It was introduced after the enabling of OpenSubdiv (rB9abcf56fa8)
it was still working on 2.80 Alpha 2. (hash: 3b7daa5bf46) (2018-11-22)
I can confirm that.
there was a big slowdown by translating or rotating similar objects. The viewport movement is still fast.
It's unrelated to the armature modifier.
It works, but every attempt to – Deselect all (alt-a) / Click on any object in front of another – cycles between the involved objects.
(current 2.8 factory default)
Not sure if this is considered a bug or feature.
Thu, Dec 27
I have just found a "workaround":
temporally disable the Does F-curve contribute to error [ ] box in the graph editor side-bar and change/add a keyframe with the modified value.
I've seen the same issue in the GP object. After adding,
for example a keyframe for thickness or color, the input fields become unusable:
Tue, Dec 25
duplicate of T59799
duplicate of T59799
But the color are on the line, instead of under the line
Take a look at https://youtu.be/pywbPQD9vYU?t=549
Apparently, the layer draw order in alpha2 was backwards, now the top layers are drawn on top.
(If your drawing isn't in-plane, there's a known 3D depth-ordering bug T59417)
Mon, Dec 24
Well, you generally get artworks :-) if you don't use the background shader:
@Brecht Van Lommel (brecht) as far I understand, that isn't fix for this crash.
I can 100% reproduce that e.g. with
- start Blender with the factory settings
- delete the default cube
- add a new cube or another mesh
- switch to the texture paint mode
Sun, Dec 23
It is already known issue, see T58317
Sat, Dec 22
@Philipp Oeser (lichtwerk) Some related notes:
Fri, Dec 21
Thu, Dec 20
In particular, adding the "pen" to the whitelist causes the CURSOR to be erroneously recognized as a stylus in the device name check in tablet_mode_from_name():
Dec 20 2018
Removing pen from the whitelist fixes the issue for this tablet:
Dec 17 2018
Hi guys, I'm trying to understand what problems you want to solve.
Looks like you're talking about two different issues.
Dec 12 2018
@Clément Foucault (fclem) so far I haven't got any edit-mode crash after 10eb92035c9b.
I just got another related? crash:
- works on commit(s) before ae1f563899de458b94d315823fd077ea4002dd86
- crashes on current blender2.8 git tip, commit eedd679ef274ed1bbb65420483ff5f23afc07f21
Dec 11 2018
It opens fine here. All the commit you bisected are from yesterday.
bad: [d415b5c7b85b07ce4e30e4dbe33491df23202cdc] Fix crash in do-versions after recent changes Date: Tue Dec 11 12:44:29 2018 +0100
git bisect start # good: [168a6a4bfc13cbe5c2fb510607b9565de3eb9472] Merge branch 'master' into blender2.8 git bisect good 168a6a4bfc13cbe5c2fb510607b9565de3eb9472 # good: [168a6a4bfc13cbe5c2fb510607b9565de3eb9472] Merge branch 'master' into blender2.8 git bisect good 168a6a4bfc13cbe5c2fb510607b9565de3eb9472 # bad: [d415b5c7b85b07ce4e30e4dbe33491df23202cdc] Fix crash in do-versions after recent changes git bisect bad d415b5c7b85b07ce4e30e4dbe33491df23202cdc # bad: [0c3535df9f5bb91045d85298ebc4697fec75eda1] Edit Mesh: Fix edges poking through the verts points in xray/seethrough mode git bisect bad 0c3535df9f5bb91045d85298ebc4697fec75eda1 # good: [c2b0d8b6d6c771da70b8fd9da5ac9cc53d04184d] Fix T57633: Particle texture update problem git bisect good c2b0d8b6d6c771da70b8fd9da5ac9cc53d04184d # good: [a99eb0ca689c8116964032faf8425cfba16759bc] GPU: Cleanup: Rename vbo to ibo for index buffers git bisect good a99eb0ca689c8116964032faf8425cfba16759bc # bad: [ae1f563899de458b94d315823fd077ea4002dd86] DRW: Batch Cache: Mesh: Port more batches to batch request method git bisect bad ae1f563899de458b94d315823fd077ea4002dd86 # good: [1d92888ee8e8abf7ba40c1875dedc84e0138be2c] DRW: Batch Cache: Add batch request system git bisect good 1d92888ee8e8abf7ba40c1875dedc84e0138be2c # first bad commit: [ae1f563899de458b94d315823fd077ea4002dd86] DRW: Batch Cache: Mesh: Port more batches to batch request method
Mesh with Subdivision Surface or Mirror modifier -> edit mode
( Linux d415b5c7b85b07ce4e30e4dbe33491df23202cdc )
Dec 8 2018
Looks like a mesa 18.2.x driver issue.
Works on the same hardware with Linux mesa 18.0.5 or Windows.
Dec 7 2018
I can reproduce that with this file
Dec 4 2018
Note: Loading your file without UI (i.e. with factory settings) fixes the issue.
Dec 1 2018
with just one multires level, the model disappears completely when rotating in sculpt mode.
(Linux, git blender2.8 tip) It depends on the option "Number of subdivisions in sculpt mode".
For >1, it shows only one division during the view-port movement. The automatic depth positioning for the center point works only when sculpting the 'visible' subdiv #1 mesh points.
Nov 15 2018
@Campbell Barton (campbellbarton) from a graphic-design point of view and the front/back consistency it doesn't look very good. My (female) co-worker and I still prefer variant 1 :
testing here, axes which are in-front are displayed brighter - which is the intent.
I mean the aligned axes (Y in the example 1,2) that worked on the previous commit
now they are always bright.
@Campbell Barton (campbellbarton) That's hard to understand and doesn't work as intended for the first paragraph:
+ /* Flip the faded state when axis aligned, since we're hiding the front-mode axis + * otherwise we see the color for the back-most axis, which is useful for + * click-to-rotate 180d but not useful to visualize. + * + * Use depth bias so axis-aligned views show the positive axis as being in-front. + * This is a detail so primary axes show as dominant. + */ + const bool is_pos_color = ( + ((axis_order[axis_index].depth > (axis_depth_bias * (is_pos ? -1 : 1))) == (axis_align != axis))); +
Nov 14 2018
Anyways, will leave open for now...
Thanks, I still don't understand how this CMS works ;)
Nov 13 2018
To evaluate this requires more than just recoloring a single screenshot …
You're right, we need a prototype to fine-tune the details.
Personally, however, I believe that you only construct problems that aren't a problem.
I think the negative signs are more disturbing. What if you only change the opacity?
It becomes clear which axis points forwards or backwards:
Nov 9 2018
Oh, this null pointer is actually the cause of T57639.
Nov 7 2018
I'm not familiar with the code, but it looks like some timing/concurrency problem.
I think this breaks the build
Nov 6 2018
Unfortunately I don't know any pattern to reproduce.
Happens here randomly 2-3x / day during a quick work on low-poly game assets.
Nov 5 2018
I marked it as invalid because it's not a bug and definitely not a crash (the only kind of reports allowed at this moment). But then I felt bad about it and spent my Friday night fixing it.
Nov 3 2018
Although Pablo very kindly marked this task as Invalid, the problem was IMO resolved in:
UI: Darken the backdrop of navigation gizmos.
UI: Soft drop shadow on 3D Viewport info text.
Nov 2 2018
Actually, the left mock-up looks good, the outlines are annoying.
I think it does not matter if the backdrop disappears. As long as the (white) text is still readable and the symbol is visible, the grey backdrop loses its original function.
Looks like we have a transparent (white) circle surrounding gizmos already (text doesnt). Unsure what a good solution would be here.
Well, I think the most important option in the entire Blender history would actually be [x] Position Lock ;-)
or maybe just [«] [»] for position undo/redo.
Nov 1 2018
+ Add [Rotation x y z] – here and in the sidebar?
Setting the cursor position from the sidebar works here (Linux, current tip)
Apr 21 2017
clinfo 2.1.16.01.12 (mesa), this seems to be a related bug/error:
Unfortunately you are right, GCN1 arch ;-(
Even after upgrade to the latest AMD driver still the same issue.
Apr 3 2017
Apr 2 2017
The warning is confusing / probably outdated. Should be something like
'Enable Auto Run Python Scripts in User Preferences > File'
I can confirm the same error message with Ubuntu LTS 16.04 + proprietary AMDGPU-PRO Driver 16.6:
Jun 5 2015
Turns out, there is also a regression in the 'freehand' drawed strokes with the tablet pen. The first screenshot is from v2.74,
the second one from current tip at 94e7ac5b97dd0b15a53d47c78672b8ed1e8a4bf1
Even with my previous 'imrovement' the artist keeps calling
the GP pen pressure support a crap :(
Jun 4 2015
At least it works better for me with both tested Wacom tablets on Linux and Windows7.
Jun 3 2015
gpencil_paint.c:362 /* if this is just the second point we've added, increment the buffer size * so that it will be drawn properly... * otherwise, just leave it alone, otherwise we get problems */ if (gpd->sbuffer_size != 2) gpd->sbuffer_size = 2;
is kind of funny ;-)
Xinput events coming from the tablet pen (a quick rise from full pressure):
(blender probably pick-ups the last a=0 events for the line thickness)
A similar problem has the GP_PAINTMODE_DRAW_POLY mode.
Jun 2 2015
Apr 5 2015
Btw. almost everywhere in blender, the key-repeat doesn't make sense
and should be therefore disabled. I have been teaching already many kids
(using blender). For some reason this is causing big problems
for their fingers/brains. I am curious to see how it is with stickies ;)
(beware: I can't decipher the code and may completely misunderstand the stickies)
this enables the keyboard-repeat and makes the numpad keys behave as
before - here is it expected. But i think, this shouldn't be the default for shortcut-keys
set as hold aka sticky?