I disagree, because the original code fills cmd with null bytes first before filling in the size and the string. To get the same effect we need a null-padded string in the pack.

I imagine this code to be part of a symbol table with fixed-length cmd records, so the exact size could matter.