Part of #43927. The unfill concept introduced in #45222 as a mirror to fill is ultimately misguided; the actual use case is backing up the received key fields while zeroing a packet struct (to initialize the cache), which amounts to temporarily moving those values out of the struct – this is better achieved by bitcopying the data, i.e. a straight assignment (for non-arrays at least). In contrast, fillclones the data, which is not what we want here (and which could cause problems if and when we start to support types that need to be destroyed explicitly).
alienvalkyrie 
Part of #43927. The unfill concept introduced in #45222 as a mirror to fill is ultimately misguided; the actual use case is backing up the received key fields while zeroing a packet struct (to initialize the cache), which amounts to temporarily moving those values out of the struct – this is better achieved by bitcopying the data, i.e. a straight assignment (for non-arrays at least). In contrast, fill clones the data, which is not what we want here (and which could cause problems if and when we start to support types that need to be destroyed explicitly).