innerSeams configuration for internal faces of hollow cube model

[muzikbike]

Currently developing a resource pack which contains a custom model for glass blocks containing inner faces. I've attached minimal resource packs containing these models for testing.

When innerSeams is not enabled, the back faces of pretty much any shape constructed from glass blocks render connected like the front faces do.

I, however, want to use innerSeams, and that's where the problem arises: turning it on causes the texture connections on the back faces to just give up. The front still works fine, the back faces are the issue here.

The two test resource packs can be found as follows: hollow-glass-continuity-no-seams.zip hollow-glass-continuity-with-seams.zip

The same issue is also present in OptiFine 1.19.4, but is blocked by a more significant bug (https://github.com/sp614x/optifine/issues/7338).

[PepperCode1]

From my understanding of innerSeams, this is intended. Continuity does not care what depth the face is at or what its cull face is.

Assume connected textures are being applied to the inner face of an outer center block $B_O$ at position $P_O$. To make explanation easier, also assume that right now the up direction is being checked for a connection. Block $B_A$ at position $P_A$, the position offset in the relative up direction, is checked for a connection. If the predicates in the properties file match $B_A$, the connection succeeds. However, if innerSeams=true, block $B_B$ at position $P_B$, the position "in front" (offset in the direction of the quad's light face) of position $P_A$, is also checked. If block $B_B$ also passes the connection predicates, the original up connection fails; otherwise, it succeeds.

I would like to know how you would like this issue to be resolved, since right now I do not imagine how it could be done without special-casing. Did innerSeams work differently in MCPatcher and would such functionality resolve this issue? If so, how did it work?

[muzikbike]

Here's glass from Misa's resource pack in 1.7.10, which is the look that I'm going for in the pack I'm developing, model-wise:

Checking the properties file for glass CTM reveals that innerSeams are set to true . 47 files are specified instead of 46, and the properties file specifies this accordingly, but in terms of connection geometry 0-46 are the same as in vanilla Continuity and 47 appears to duplicate 26 (the images are identical), so I'm not sure what it represents.

Since this is 1.7.10, a version that predates models, the effect seen in these screenshots would likely have been achieved by simply not culling the rear faces of the blocks, which is admittedly a far simpler approach than defining a completely new model with extra planes explicitly specified as stand-ins for those faces. I'm not sure if this is something that would actually be possible to implement in Continuity at all.

I assume this is what "renderPass" (https://github.com/sp614x/optifine/issues/5084) is referring to? I myself have never actually developed resource packs until many years after MCPatcher stopped being a thing, so I wouldn't know. The properties file for Misa's glass defines renderPass to be 3 for the glass itself, and 2 for the overlaid edge frame texture.

[muzikbike]

According to the following forum post, renderPass was for changing the render type of the target block: https://www.minecraftforum.net/forums/mapping-and-modding-java-edition/resource-packs/resource-pack-discussion/1256343-a-texture-artists-guide-to-mcpatchers-features

Its functionality appears to have since been moved into a new block.properties file, if I'm understanding it correctly: https://github.com/sp614x/optifine/blob/master/OptiFineDoc/doc/block.properties

This is what the latest version of Misa's for 1.19 seems to use. I've compared the glass geometry, and most things match up, with the only difference is that innerSeams only appear to render from the outside, and not on the inside. I don't know if this is down to the custom model or not.

Does Continuity also support block.properties and other such files?

[PepperCode1]

Yes, Continuity supports block.properties, but this file has never had an option to disable back face culling. Disabling back face culling is not something Continuity can do because it does not control rendering and the API that it uses does not have an option for this.

[muzikbike]

I've done some more experimentation. Here's a minimal-ish implementation of the partial workaround from Misa's pack: hollow-glass-continuity-partial-seam-fix.zip

This works by having the inside and outside of the model used point to separate texture files. That way, it's possible for CTM to apply different types of connected texture specifications to the outside faces and inside faces. This almost resolves our issue.

Here's the original demo without innerSeams:

Here's the original demo with innerSeams:

And here's the new partial workaround:

The reason why this is only a "partial" workaround is that this fix only enables innerSeams on the outer faces: it's disabled on the inner faces, which we can see from the L-shape above. If we enable it on the inner faces, we just run into the original issue again.

It seems entirely possible to fix this issue properly now that we define the inner and outer faces separately. While Continuity doesn't care about what depth a given face is in a model, we can still explicitly define what counts as an "inner" face and what counts as an "outer" face in the resource pack itself and configure their CTM settings independently.

My proposal for fixing this is as follows: the CTM .properties file would accept a new optional parameter - let's call it "seamType". This new "seamType" would define the way that innerSeams would detect adjacent glass blocks and how it would change the texture used. By default (and if omitted), it would be set to "outer", which is the current innerSeams behaviour. It could also be set to "inner", in which case the checks would instead be performed in a way that would be appropriate for inner faces instead, and the chosen textures would be applied accordingly. So, if "seamType" was implemented, we could set innerSeams to true for both the inner and outer textures, and set "seamType" to "outer" for the outer textures and to "inner" for the inside textures. With this, we would be able to recreate MCPatcher's backface functionality perfectly.

Of course, that's just off the top of my head - I wouldn't be surprised at all if it turned out to not be that simple, but one can hope.

[muzikbike]

For the record, here's the same arrangement as above in 1.7.10 with MCPatcher. This is the visual result we're looking for. Note the vertical white bar inside the L-shaped cavity from the inside - we want to be seeing this in 1.19.4 as well.

Also, since you said that Continuity doesn't touch the rendering, I have a few questions:

• Is this why the proposed glass pane culling fix (https://github.com/PepperCode1/Continuity/issues/74) was said to be fixed by a separate mod at some point in the future, rather than the fix being included in Continuity itself?
• Does this mean that there isn't a way to make normally non-translucent blocks translucent in Continuity? I've been using MoreTranslucency (https://github.com/Benonardo/MoreTranslucency) in order to make glass, glass panes and beacons render translucently. I know that OptiFine has the ability to make non-translucent blocks translucent in resource packs via the block.properties file, but this seems like something Continuity wouldn't be able to do if it can't change anything rendering-related.
• On a similar topic to the above, I assume that porting MCPatcher's "blend" translucency mode to Continuity is also out of the question since it's rendering related?

[PepperCode1]

• 74 is not fixed by Continuity because it is not directly related to connected textures, but it can be fixed without touching rendering.

• As I said, Continuity supports block.properties exactly like Optifine. You can use it to change the block layer. Changing the layer is not part of rendering, but what how the layer looks like is part of rendering.
• Correct. Continuity cannot support this currently.

[muzikbike]

Not to get too off topic, but is block.properties expected to be formatted differently somehow compared to that of Optifine? I have the following test resource pack with a block.properties file formatted correctly for Optifine to my knowledge, but it doesn't work in Continuity and the glass remains completely opaque. Is there something I should be doing differently to make the glass translucent without relying on an extra mod? block-properties-test.zip

[PepperCode1]

block.properties works exactly like in Optifine. block.properties must be located in the minecraft namespace, just like in Optifine.