On 18/07/2025 2:28 pm, Jan Beulich wrote: >>>>>> uint16_t model; >>>>> Whereas the model is strictly limited to 8 bits. >>>> There is space in here, if we need it, but you can't shrink it without >>>> breaking the check for the NULL entry (going back to the first >>>> obfuscation). >>> Breaking? Or merely affecting code generation in a negative way? >> Shrinking model without adding (and checking) a new field would mean the >> loop condition no longer covers all metadata. > And it doesn't strictly need to. It needs to check enough to not mistake a > valid entry for a sentinel one.
I've found a nicer way of doing this, but it needs another prereq patch. I'm preparing a v2 with the remainder. It also addresses the horrible code generation. ~Andrew
