The DNMD edit APIs don't account for/update the "referenced type system tables" number of rows for portable pdbs.
The existing "keep the image schema matching the savable format" logic assumes that all table references point to tokens that can exist given the current schema. For Portable PDBs, tokens can reference tables that exist in the "runtime metadata" image, so we need to account for that when editing. In particular, we likely need to enable "resizing" cell references on "set column value as token or cursor" operations when values don't fit and tracking the "highest referenced row" in various tables.
The DNMD edit APIs don't account for/update the "referenced type system tables" number of rows for portable pdbs.
The existing "keep the image schema matching the savable format" logic assumes that all table references point to tokens that can exist given the current schema. For Portable PDBs, tokens can reference tables that exist in the "runtime metadata" image, so we need to account for that when editing. In particular, we likely need to enable "resizing" cell references on "set column value as token or cursor" operations when values don't fit and tracking the "highest referenced row" in various tables.