- User Since
- Apr 11 2017, 11:26 PM (102 w, 3 h)
Deactivating buttons makes sense. I'll have to check how to identify nodes that don't allow muting. I have an exam next Tuesday, so the update may have to wait for a bit. Perhaps we have a consensus on how the improvement should look like till then.
Currently the checkbox icon is displayed at all time, which may be distracting. Should it only be display when the node gets selected?
I think I'll try to implement this.
Looks like the same issue reported in T59583
@Brecht Van Lommel (brecht) You're right I was oversimplifying. What I meant was if $XDG_DATA_HOME is set we know where to move the files to (+handle metadata for restoring). The bigger challenge may be that some distributions don't follow the spec and $XDG_DATA_HOME isn't set. For instance on my Ubuntu VMs 16.04 and 18.04 both don't have $XDG_DATA_HOME set. Your linked code from Electron probably already solves this by using the cli tools, which will default to gio trash for Ubuntu with Gnome as far as I can tell.
Mon, Mar 25
@William Reynish (billreynish) I guess this should be addressed for Linux and MacOS as well. Currently remove() and rmdir() are used for Linux and MacOS which are equally destructive.
@Kris (Metricity) I agree, the separate function makes sense. It's a good idea to have this functionality for Linux and MacOS, too. I'm not sure if Linux has a unified way of handling the trash directory. If Linux and MacOS have the path set in an environment variable it should be a simple mv operation.
Corrected error handling. Only CoUninitialize() when CoInitializeEx() returns either S_OK or S_FALSE both of which have a positive return value.
- Only use backward compatible FOF_* flags
- Make progress silent, but show system dialogs when an error occurs, especially when a file would be permanently deleted because it can't be moved to the bin
- Removed legacy interface, since code should work for Windows Vista and newer versions
@LazyDodo (LazyDodo) That's what I thought. I wish they would properly document their API and give reason as to why certain recommendation are made (or even whether that is mandatory or a recommendation). I'll do some tests with the backwards compatible flags and if some others should be included as well, like FOF_WANTNUKEWARNING. It would be great if someone could test this on Vista, just in case something behaves differently.
I've tested multiple subsequent file deletions and there was no problem, even with how CoUninitialize() is currently used. Not sure how sound this solution is, but it's working in my test environment.
The COM handling seems to be a bit difficult with the CoInitializeEx() CoUninitialize(). As per the documentation I'm not exactly sure what they do on CoUnitialize that requires it to be called before closing the application (and not sooner). I'm not sure how we could satisfy this besides moving them to Blender's startup/exit code. Perhaps using the old interface is the better solution for now, if that causes too much of an headache.
Fixed header include, should now work on older platforms as well.
@Kris (Metricity) thanks for implementing this! I hope you don't mind my improvement suggestions.
It supports longer paths and is the current API. I'm not sure what Microsoft will do with the legacy SHFileOperationW in future versions.
I've got the patch done, doing some last tests and then I'll submit a diff based on the code of @Kris (Metricity)
Perhaps we could do the following: We could implement BLI_delete_recycle, but use #ifdef WIN32 in file_delete_exec to call this function only if it's Windows. This would avoid confusion about the Linux function call, which in your patch is just a wrapper for the regular BLI_delete.
@Kris (Metricity) I'm not sure that the BLI_delete_recycle is a good idea, since the Linux version does not recycle. Is there any particular reason why you'd want a new function?
Sun, Mar 24
@Kris (Metricity) I have implemented a first version that works, but I'll have to do more testing. The API's documentation is C++ only, which made it a bit difficult. I'm also not sure whether the choice between COINIT_APARTMENTTHREADED vs COINIT_MULTITHREADED makes a big difference. For simplicity I chose COINIT_APARTMENTTHREADED which is what the official examples use and by the [[ URL | https://docs.microsoft.com/de-de/windows/desktop/api/objbase/ne-objbase-tagcoinit ]] won't cause us any headaches with concurrency.
I'm doing some test to see if my idea with IFileOperation works.
Looks like Windows made this one a bit more difficult. The FOFX_RECYCLEONDELETE was introduced for Windows 8, so I'm not sure how that functionality could be achieved on Vista using the new API. Perhaps we should use SHFileOperation for Vista and IFileOperation for newer versions?
@Kris (Metricity) SHFileOperation is the legacy API which doesn't support paths longer than 255 characters. If XP support is required we could use the VersionHelper to determine the OS version running an use IFileOperation if it's Vista or later. Otherwise I'd go with IFileOperation , because that is the the current API.
Tue, Mar 12
Jan 30 2019
Jan 27 2019
@Joel Godin (FloridaJo) Have you tried building it from the console using make? This will give you a detailed output on what went wrong.
@Joel Godin (FloridaJo) That's just the hash of your local commit, Jacques won't be able to commit any code with that. It seems you have some trouble with the process so let me try to explain it to you. The reason this is done with git even for two lines of code is to have a consistent and predictable end result when you've got several people working on the project. So let me walk you through the steps (some of these may be obvious to you, but just I'll write them down anyway in case this is something you're unsure about).
This page explains the process of creating a diff. It's the same for Linux and Windows.
Jan 26 2019
@Dirk Richter (dirky2006) Have you tried running Blender with the --debug flag? It will keep the console open and hopefully displays information about the cause.
Marked @Brecht Van Lommel (brecht) comments as done.
The camera cull margin is now of subtype FACTOR and the cull distance of unit LENGTH.
So removing the subtype from camera_cull_margin and make distance_cull_margin a length?
You're right, I was following the remarks of T60824 without checking if the subtype and unit actually matches.
Jan 25 2019
I've fixed some inconsistent quotation marks in the file as well.
Jan 16 2019
@Marek Trejbal (marek3ball) Sounds more like an issue with your OS and drivers or even hardware then.
Jan 15 2019
@Marek Trejbal (marek3ball) On Windows Blender doesn't create crash report when a segfault occurs. You can start it from command line though and add the "--debug" flag, this will keep the console open with debug prints. More specific flags are available if detailed information for e.g. GPU rendering is required.
Jan 8 2019
Can you think of any work-around to get an albedo-like render pass for the hair shader?
Jan 4 2019
@Brecht Van Lommel (brecht) Alright makes sense.
Jan 1 2019
I believe this is a bug and unless use_property_split should not be used in certain situations. If that is the case the documentation should reflect this.
Dec 18 2018
I can't reproduce with cef41d0144e (Win10 64-Bit). Your .blend file has hair and importing the .obj and adding hair works fine as well.
You're right it takes a few samples and is not related to the hair shape.
There was a bug with undo in the version you're using. Please try the current build for Windows and see if the problem persists.
Bug report seems to be them same as T59560
Seems like the same bug reported in T59531
Debug output for 4fceaf3848b
Cannot reproduce on Ubuntu with latest build 7aeb24e0371
Can confirm on Windows.
Dec 3 2018
@Bastien Montagne (mont29) thanks for taking care of this. I would suggest using a venv for building the docs though, that makes it far easier to have the same versions as the main docs and perform upgrades as needed. No bleeding edge is certainly not required, but system packages may not be equal to the ones in available as pip package.
Dec 1 2018
@Bastien Montagne (mont29) replacing the fonts in .\site-packages\sphinx_rtd_theme\static\fonts would then be the easiest way. Sphinx only copies those files anyway, so as long as the replacement files have the same name everything will work just fine.
Nov 23 2018
Can reproduce on both Windows and Ubuntu 64-bit.
You have accidentally enabled a keying set, see: https://blender.stackexchange.com/questions/16531/keying-set-failed-to-insert-any-keyframes
Nov 22 2018
Apparently there was an error with the extended latin characters in the Lato font in case your sphinx_rtd_theme still uses the pre August 2014 version.
The sphinx build shouldn't be a problem, everything down to the font-face definition in the css is correct. I think it's either a weird bug on @Jean Da Costa (jeacom256)'s system or the font file has some obscure error. The easiest way is to replace the Lato .woff2 files with a current version. That definitively won't break anything and possible fixes the problem, since the current version of Lato seems to render just fine.
Could you try to load this page and see if it renders properly (and if it loads the .woff2 files or another format)?
That is a bit strange. If your browser wouldn't support, it should load the *.ttf as declared in the font-face definition.
The second GET-request to https://docs.blender.org/api/blender2.8/ in your .har-file returns all the data in the html. Perhaps this is a rendering bug in your browser?
It seems like the same problem reported in T57657
I can reproduce this error for both 2.79b and 2.8 on Ubuntu 64-Bit using a pip installed with ensurepip. This error doesn't seem to occur on Windows.
Nov 2 2018
Oct 30 2018
Increasing the feature strength mostly eliminates the artefacts, but this does potentially impact other areas and one might lose details there, right? If the current result is the expected behavior, please close the task.
Oct 29 2018
I apologize in advance in case this doesn't work as expected due to my lack of knowledge about the denoiser and its settings.
Aug 11 2018
The barbershop scene crashes every time so far using the GPU.
Aug 10 2018
This is likely a problem on the server, the benchmark json files look fine to me.
Feb 27 2018
This issue is discussed in T52507
scene = bpy.context.scene
Feb 26 2018
Feb 25 2018
Aug 9 2017
I understood that. An API to check thread status would still be very useful.
Aug 8 2017
Yes that is what I was doing already. ‘EXEC_DEFAULT’ is used when no arguments are given. So the solution is to have no responsive/interactive UI and no preview.
I will take a look at the code. Perhaps this can be changed in future versions. It is very counter intuitive that the callback claims everything is done, but it isn't.
What you described as root issue was pretty much what I imagined. Unfortunately the re-try idea doesn't work well, once it has returned 'CANCELLED' it's pretty much guaranteed to fail at subsequent attempts.