D169
- The provenance ladder gains a rung: implementations
D166 corrected a check that asserted the opposite of what
sceKernelClearEventFlag does, and the correction had
nowhere honest to go. What supports the corrected version is shadPS4's
Clear (m_bits &= bits, C++) and PS5PCEM's
clearEventFlag (object.bits &= mask, Zig,
carrying the comment "The PS5 ABI supplies the bits to retain, not the
bits to remove") - two implementations, two languages, no shared
codebase.
ASSUMED says this project reasoned it out, which throws
away that somebody's working code says so. DOCUMENTED
claims a citation nobody here can produce, which is the mistake that
caused D166 in the first place. The check sat at ASSUMED
for as long as it took to conclude the ladder was missing a rung rather
than that this was a guess.
ASSUMED -> IMPLEMENTATIONS -> SPEC -> DOCUMENTED -> DERIVED -> HARDWARE
Below
SPEC, and the placement is the entire caveat
Stronger than this project's own guess. Weaker than a
document anyone can check, because implementations are not independent
witnesses. These projects read each other's source -
obscene-tool consensus prints exactly that in its own
output, "agreement is evidence, not four witnesses" - and two sharing an
ancestor agree about their ancestor.
The line to hold: at least two implementations that do not share a
codebase, read directly, and named in the check's own
comment. One implementation is not this; ASSUMED
describes a single opinion accurately whoever holds it.
This reopens something declined earlier
The sibling project proposed a value for "measured, but not on the target" and it was declined on the argument that the origin field carries which-machine as data. That argument is sound and it does not transfer: there is no origin field on the documentation axis. Declining a second time would have been a position held for symmetry rather than for a reason.
Which is worth saying to them plainly, since the decline is on the record between the two projects.
Contract impact
A new value in the provenance field of res
records, so a consumer matching exhaustively on provenance needs the one
line. Same shape as unauthorised, and it goes on the bridge
for the same reason. counts reports it,
OUTPUT.md describes it with the caveat attached, and
exactly one check carries it today.