Apologies, still learning how to use arc properly and accidentally pulled in a change to object_dupli.c. Let me see if I can undo that.

If the option leads to a more consistent results, why is not enabled by default and why is this even an option?

In D8445#205916, @Sergey Sharybin (sergey) wrote:

If the option leads to a more consistent results, why is not enabled by default and why is this even an option?

It produces more consistent results for instancing; unfortunately it also changes the way quads (and other ngons) get triangulated which would be undesirable in other situations.

If there's a way to control triangulation separately (is it precalculated?) I agree it would be better if this happened invisibly.

Rebased these changes to master so I believe this should be up to date now.

This patch has an approved status however it was not committed yet. So I'm assuming the other reviewers are blocking. Updating the reviewers list to reflect this. This way the patch status still show as Need Review.

Also removing modifier since this is not a module that someone can talk on behalf of anyways.

In D8445#279243, @David Swift (rococo) wrote:

Would it be fair to say at this point that Geometry Nodes provide the generalized solution to this kind of problem and the old way of doing instancing will slowly go away? If so this "fix" would therefore be unnecessary.

On the other hand the same problem should be considered if creating a node that converts faces to points/instances. My recommendation would be to use the texture coordinates instead which are more easily visualized and operated upon.

I disagree on this quite entirely. The beauty of face instancing for me has always been that I don't need to manage modifiers on multiple objects to get the desired results. However the downside has then been that mirroring & using deform modifiers have always been difficult because of the issue described here.

I see no reason not to commit this. At lesat I will continue using face instancing at least until we get to see something like "collection modifiers".

In D8445#292725, @Mr Chocolate Bar (WillyWonka) wrote:

However the downside has then been that mirroring & using deform modifiers have always been difficult because of the issue described here.

I feel like the real solution here is to extend face instancing to make it behave more consistently over mirroring (e.g. by computing transforms based on texture coords like I suggest above).

Benefits:

• Only adds options where they make sense contextually: in the face instancing options
• Improves face instancing
  • More easily visualized and edited using existing tools
  • Makes it easier to drive instances procedurally e.g. through generated texture coords
• Provides a more natural transition to geometry nodes in the future

Downsides:

• Patch doesn't exist yet! Although I did implement a basic prototype
• Deciding which way round the coordinate system belongs will probably be a big discussion

In D8445#293039, @David Swift (rococo) wrote:

In D8445#292725, @Mr Chocolate Bar (WillyWonka) wrote:

However the downside has then been that mirroring & using deform modifiers have always been difficult because of the issue described here.

I feel like the real solution here is to extend face instancing to make it behave more consistently over mirroring (e.g. by computing transforms based on texture coords like I suggest above).

Benefits:

• Only adds options where they make sense contextually: in the face instancing options
• Improves face instancing
  • More easily visualized and edited using existing tools
  • Makes it easier to drive instances procedurally e.g. through generated texture coords
• Provides a more natural transition to geometry nodes in the future

Downsides:

• Patch doesn't exist yet! Although I did implement a basic prototype
• Deciding which way round the coordinate system belongs will probably be a big discussion

Alright, I will be waiting impatiently 😁