D043 - The platform is enumerated, not only asked about

Status: decided.

Every other section tests a list this program carries: someone wrote a name down and a check asks whether it works. That is only ever as complete as whoever maintained the list - and when an emulator's release notes were checked against the suite, two of the functions it had just implemented were not in the census at all. Not broken, not reported absent: never asked about.

110-modules asks the machine instead. sceKernelGetModuleList deals only in handles, so it needs no struct knowledge and is safe outright; sceKernelGetModuleInfo is given a large zeroed buffer with its leading size field set, and only the name at the front is read back.

On hardware this is the inventory. Against an emulator it is what that emulator claims to provide, and the difference between the two is the gap. That makes it the foundation for coverage: a name missing from the inventory is a different problem from a name present and broken, and until now this program could not tell them apart.

A guessed offset reports itself as a guess. When the assumed layout finds no name, the section reports the front of the structure word by word rather than trying other offsets until one produces something printable - a guess that lands on printable bytes reads as a name and is not one. The layout gets derived from what a platform actually wrote, which is the same argument that fixed the dynamic tags (D036).

First result, against an emulator: one module reported loaded, and EINVAL when asked to describe it. Found without a user report and without reading anyone's source.