D271 - An error code in a boolean reads as true

decided · 2026-08-25 · three failures, one shape

GuestError placeholders deliberately avoid the high bit so they can never be mistaken for an established firmware value. That choice makes them small positive integers, and a small positive integer is what true looks like.

So an unimplemented boolean answered yes:

• sceKernelIsCex and sceKernelIsDevKit both answered non-zero, and the platform claimed to be a retail unit and a development kit at once.
• posix_sigismember answered yes for every signal, so a set that sigemptyset had just cleared reported every signal still in it - and the failure was attributed to the function that did the clearing.

This is D125 in a boolean: a value the caller reads as data rather than tests against a table. It is the same defect as a failed open answering something that looks like a descriptor (D273), and the same one the stub policy already solves for pointer-returning functions by answering zero. The knowledge base records a return kind, so the policy could answer zero for a boolean too - that is the general fix, and implementing each boolean by hand is not it.