Windows 11 - Human displays as shadow after render #206
Comments
rwbaer
changed the title
|
I have tried the same sequence on Ubuntu 22.04 and it works perfectly (no shadow at any point). Also no errors in the log. |
|
@kshepherd2013 thanks for the idea. Despite the logs being empty, there are a few lines on the console related to dict_keys and numpy that may make a difference (and I never remember seeing before), but I don't really know where in the program these might be coming from. There also are some QWidget errors that probably shouldn't be happening. I think the ICC profile issues are long standing and haven't made an impact before although it strikes me that it would be nice to take care of them too some time. Attribute Qt::AA_EnableHighDpiScaling must be set before QCoreApplication is created. |
|
Here is my complete log for running python3 makehuman.py --debugopengl --fullloggingopengl --debugnumpy PRE STARTUP SETTING: useHDPI = False *** MESSAGE TYPE:kernel_info_request*** kernel_info_request: {'header': {'msg_id': 'be7339ad-704fb091b1120d80ae0bf6e7_1715437_0', 'msg_type': 'kernel_info_request', 'username': 'kshepherd', 'session': 'be7339ad-704fb091b1120d80ae0bf6e7', 'date': datetime.datetime(2022, 7, 25, 15, 9, 53, 385373, tzinfo=tzutc()), 'version': '5.3'}, 'msg_id': 'be7339ad-704fb091b1120d80ae0bf6e7_1715437_0', 'msg_type': 'kernel_info_request', 'parent_header': {}, 'metadata': {}, 'content': {}, 'buffers': []} *** MESSAGE TYPE:history_request*** history_request: {'header': {'msg_id': 'be7339ad-704fb091b1120d80ae0bf6e7_1715437_1', 'msg_type': 'history_request', 'username': 'kshepherd', 'session': 'be7339ad-704fb091b1120d80ae0bf6e7', 'date': datetime.datetime(2022, 7, 25, 15, 9, 53, 395929, tzinfo=tzutc()), 'version': '5.3'}, 'msg_id': 'be7339ad-704fb091b1120d80ae0bf6e7_1715437_1', 'msg_type': 'history_request', 'parent_header': {}, 'metadata': {}, 'content': {'raw': True, 'output': False, 'hist_access_type': 'tail', 'n': 1000}, 'buffers': []} *** MESSAGE TYPE:execute_request*** execute_request: {'header': {'msg_id': 'be7339ad-704fb091b1120d80ae0bf6e7_1715437_2', 'msg_type': 'execute_request', 'username': 'kshepherd', 'session': 'be7339ad-704fb091b1120d80ae0bf6e7', 'date': datetime.datetime(2022, 7, 25, 15, 9, 53, 493593, tzinfo=tzutc()), 'version': '5.3'}, 'msg_id': 'be7339ad-704fb091b1120d80ae0bf6e7_1715437_2', 'msg_type': 'execute_request', 'parent_header': {}, 'metadata': {}, 'content': {'code': '', 'silent': True, 'store_history': True, 'user_expressions': {}, 'allow_stdin': True}, 'buffers': []} libpng warning: iCCP: known incorrect sRGB profile |
|
I do not see anything from "QWidget" in mine, so you may be right that it is related to the "shadow" error. |
|
Build f8ff31c5 It would be nice if one of the other devs could confirm that they are seeing the differing behaviors Kevin is experiencing on Linux and I am experiencein on Windows 11. Also, does it happen on Windows 10 as well? It would also be nice to know what happens if we rollback to the version before we started to fix the 22.04 version to know if it is related to the fixes or is just an previously undiscovered behavior because rendering in MH is relatively uncommon. I could use a git refresher on exactly what would be the proper way to do that if one of you git experts can help. |
|
I found a machine with Windows MH1.2.0 (20210426 final (HEAD:409e74d4) in which rendering is still working as expected. |
|
As of build 1.2.1 alpha (master:1c4fa878) there is no shadow problem on Windows after render |
|
1.2.1 alpha (master:d7e0c21a) 7/14/2022 2:04 AM |
|
While it does seem we need to move to QOpenGLWidget , we may need to study the subtle changes that have happened to buffer updating and clipping to make sure we accomodate our complex GUI.
This seems to channel some of the issuses that have plagued the "fix" Is the "shadow" an empty or non-updated buffer? |
|
I removed 2 lines added to glmodule.py by kshepherd2013 [glEnable(GL_DEPTH_TEST); ] that seemed to be responsible for the "shadow behavior" on windows. I left his other modifications. It will be important to find out if this breaks the Ubuntu 22.04 repair, or if they were unimportant to that repair. It would be great if one of you could try Ubuntu 22.04 with the changes. |
|
My workstation is still at ubuntu 20.04, I haven't had time to update it yet. And running a test in virtualbox isn't really fair. With ubuntu 20.02, I don't get a shadow but the zoom bug is there |
|
There is a small side effect of having removed these two lines in Windows. The clipping plane on the front of the face after a render is incorrect. However this is only in the viewer tab after a render. The modeling tab and other tabs show the human as expected and seem to function normally with the exception of the zoom issue #208. On Windows overall functionality is better with the lines gone because there is no seeing anything with the black shadow in front.
|
|
Same problem on my Linux-System, now. It looks like the shaded mesh is hidden behind the black object. Could it be a wrong buffer showing? |
|
@Aranuvir - Just for clarity, do youn see the shadow with Ubuntu 22.04 or 20.04 (or both)? Should we move the ksshepard2013 fix to a feature branch until we get this solved? Are we certain the old master code doesn't work on all Ubuntu 22.04 systems or is it just some of them? Either way, I think we do need to move forward on QOpenGLWidget (useful links in post above It's a questionof what master is least toxic to sourcce users. |
|
FWIW, with the current nightly I'm no longer getting the black shadow on Windows 11 when I render, but there is a new issue that mI describe in #225 along with a link to reports of back face culling issues by Ricardo2020 on the forum that may be related. If you think they are the same issue, you can close this and continue there from the Windows point of view. |
Nightly Build: 20220719 alpha (HEAD:f8ff3bc5)
Steps to Reproduce:
Start MH
Switch to Rendering | Scene and observe what it looks like for current lighting
Go toRendering | Render tab; click render
Return to Rendering | Scene and observe that the display is a shadow now
Is this related to #198 and/or #204?
EDIT: IMPORTANT ** Going to the modeling or geometry tabs produces a human that is also just a shadow. The human can be seen when changing the rotation dynamically, but when things come to rest the human is again a shadow.
The shadow honors "hide faces under clothers". If you uncheck this box, the "supposedly" hidden faces of the naked human show through the shadow. Could this be side-effect of fixing z-plane culling in the scene tab?
The text was updated successfully, but these errors were encountered: