Add a correction for this too please, might as well.

I just want feedback to make sure everyone is okay with the layout of the notes on the AO and Bevel pages before committing it.

Update to include notes on the AO and Bevel node pages.

• Rebase on master

I feel like dereferencing a null pointer isn't the best way to handle the problem. GLTexture::init_internal() is taking steps to verify that the GPU can handle the texture and explicitly returns null and cleans up after itself. It should be the caller's responsibility to check for a null pointer before using it. drw_texture_set_parameters(tex, flags) does, but its check silently returns. Everything is doing its job except for EEVEE_shadows_draw_cubemap() (I think).

No not that Windows. Only for Macintosh.

Can't confirm on windows

@Julien Kaspar (JulienKaspar) Thanks for the additional information and attention to this bug.

• Separate processing of attributes of different type and process their temporary data.

I currently do not have 3.4 installed due to an addon conflict but Im very certain the result is the same than with 3.1.2. Its just meant for comparison.(still includes the ugly issues that I am trying to push to get fixed in T72011)

Sorry, completely blanked out. Added it into the OP.

Thanks. Could you attach the file here?

@Hans Goudey (HooglyBoogly) @Philipp Oeser (lichtwerk) T104334

@Lukas Stockner (lukasstockner97) could you report a bug with a file to reproduce the issue? I think bug reports are preferred over the "Raise a concern" feature for this kind of thing.

camera.cpp - Further Refactoring, very similar approach

I'I have the same issue but not in all cases. Here my specs:

• Rebased
• Changed to keep the Anisotropic idname

This is a limitation mentioned in the manual:

• When a face of an N-gon is selected, the individual edges will be subdivided but the face will stay unsubdivided.

This seems to be too complex for the Triaging team to confirm if it really is a fake dependency cycle and therefore a bug.
But at first glance it seems so, so I'm confirming it.

Thanks for the detailed information.

The default color is (a specific) blue now, as @Julien Kaspar (JulienKaspar) requested. I'm not sure how I feel about it, perhaps a lighter shade would work better.
Also, I think it might work with Metal now, because I made some changes to the no_geom shader.
I think the issue was that I was adjusting gl_Position after the geometry shader alternative section; now it's before that, at the end of the regular vertex shader section, so it should be the same as the other vertex shaders.
I haven't been able to make it work in orthographic view yet, but I'll keep trying.

Or you mean attrbute id? I not sure baout this..

Hmm, I can't reproduce the difference in the file above. The file looks like this in 3.3:

• Don't find metadata for builtin attributes

Looks good to me, couldn't find any further issues with it.

• Rename.
• Move call in pointcloud_from_curves.
• Skip builtin attributes.

Hey Mr. Borhade:

Cannot reproduce in Blender 3.4 and latest master. Could give us more details?

The remaining thing I described badly in my last inline comment was ordering the selection input right after the geometry socket like other nodes. Then the retrieval of the specific socket in node_update needs to be updated.

This looks mostly good, thanks for looking into it. I'd change a few things:

• Check if the attribute is builtin on curves before transferring it.
• Call this from pointcloud_from_curves
• The function name move_generic_curve_attributes_to_point_domain might be clearer/more generic.

Thanks, always nice to remove code.

I think technical details for the OSL node is fine, it's for technical users.

This report does not contain all the requested information, which is required for us to investigate the issue.