Hive models section does not allow Patch 1.33 format models

[Retera]

Similar to the upgrade to Reforged, the new Patch 1.33 for Warcraft III has a new model format of its own. From what I can tell, when I tried to upload a model in that format I was not able to, and the Hive said that there was a server error.

The numeric versioning on these files is MDX version 1100.

Assuming that the Hive Workshop wishes to support the official version of Warcraft 3 going forward, the site should allow users to upload MDX model files from any patch.

 

[Archian]

Thank you for bringing this to our attention. We will look into it.

 

[Retera]

Thanks. I don't much like Patch 1.33 personally because it breaks the KMTF chunks in all versions of the MDX models -- with respect to how the game understands them -- and I have no ability to really anticipate if Blizzard is going to fix that. These are definitely hard times for the technology.

 

[Ralle]

What's the difference though?

 

[BogdanW3]

What's the difference though?

Meterial layers are a bit different, but I can imagine the issue of "A server error occurred. Please try again later." happens due to it trying to extract resources in use and failing. ideally, Hive's internal upload-time parser would only parse TEXS, CORN, and FAFX chunks as these are the chunks that contain file path references, and thus be unaffected by the v1100 changes. Maybe this is already the case and the parser just doesn't allow version == 1100

 

[Ralle]

If those layers have anything to do with textures, then it does. The parser parses all chunks mostly as a point of pride. Not a necessity.

 

[BogdanW3]

If those layers have anything to do with textures, then it does. The parser parses all chunks mostly as a point of pride. Not a necessity.

Layers do reference textures, but they aren't useful for determining which textures a model loads up, the Textures chunk is for that.
If v1100 models still fail to upload, it might be good to stop loading layers and materials for a quick sec. Here's also an example model from the game itself

 

[Retera]

The parser parses all chunks mostly as a point of pride.

Didn't a great druid once say, "If pride gives us pause, then perhaps we have lived long enough already."

Edit: I mean, I'm just kidding around here with that quote. But the cross-compatibility between v800, v900, v1000, and v1100 in ReteraModelStudio and trying to ensure it worked in every case killed a lot of time for me because I changed some of my inprocess representation of everything to match the v1100 format. So it left me feeling like supporting this takes longer than it might be worth. But really that's probably not true. For 99% of cases you can just read a repeating set of two ints and the parser is upgraded. The problem case is the one even Blizzard's people didn't get right -- hence the ingame SD models that no longer render correctly ingame -- when the two repeating integers inject a KMTF chunk between themselves. It's repeatable so you can have many KMTF per material layer instead of only one, but when you have those injected in then the numbers are intentionally set to 0 to confuse you, even though 0 is a valid value when there is no repeating KMTF.