Draco compression will, in rare cases, require generating new accessors. That's not ideal — I'd much prefer that the I/O process not modify the input Document at all — but it is a current limitation in the encoding process.
Previously some other code in the general I/O pipeline assumed that the accessor list was known in advance, and caused failures when the accessor count changed later. This PR removes the assumption.
Draco compression will, in rare cases, require generating new accessors. That's not ideal — I'd much prefer that the I/O process not modify the input Document at all — but it is a current limitation in the encoding process.
Previously some other code in the general I/O pipeline assumed that the accessor list was known in advance, and caused failures when the accessor count changed later. This PR removes the assumption.
https://github.com/donmccurdy/glTF-Transform/blob/de19547de2659c605bba28531b26b7aa87d832bc/packages/extensions/src/khr-draco-mesh-compression/draco-mesh-compression.ts#L397-L413
Related:
Remaining:
possible to write a unit test for this?too difficult to reproduce, tracking instead at https://github.com/donmccurdy/glTF-Transform/issues/1386