D502 - Ten outstanding measurements were never going to be claimed

decided - 2026-09-03

OUTSTANDING is documented as a work queue: "every entry in it should one day move up into a test", and OPAQUE exists for the ones that never will, because "keeping them there would leave the queue with permanent residents - at which point it stops being read as a queue."

Ten entries were in the wrong one.

OUTSTANDING 95 -> 85        OPAQUE 39 -> 49

The list's own example was in the queue

OPAQUE's doc comment names the case that forced the table to exist:

A module handle is the case that forced it. The console answered 0x15 and 0x14 for two /app0 modules, and both runs agree, but the number reflects how many modules that loader had already placed.

106-encoder/module-handle:loaded:handle was sitting in OUTSTANDING. So was 106-encoder/module-list:total:count, which is the same fact stated as a total.

What the ten have in common

None of them is a property of a call. Sorted by why:

• The check's own input, read back. 120-measure/identify-clocks:sceKernelUsleep:requested records the sleep the probe asked for. A platform that ignores the request entirely produces the same reading. This is D497's family one layer up - there it was an out-parameter the probe had initialised, here it is an argument the probe chose.
• A moment on one machine. The three clocks-advance absolutes are readings taken at an instant, and orbistoun answers from the host clock; it cannot reproduce the console's epoch and nothing should make it try.
• A rate on one machine. The three timer-ratio deltas. The frequency they calibrate to is claimable and is asserted separately, which is the useful half kept.
• That machine's configuration. kern.osrelease's length is the length of 0.0-prototype, a per-machine setting, empty by default.
• That loader's bookkeeping. The two module-handle entries above.

Why this is worth doing rather than tidy

A queue with permanent residents stops being read, and the count is quoted. "95 outstanding hardware measurements" was the headline for a day's planning, and eleven per cent of it could never have been worked. Moving them does not make the project more capable; it makes the number mean what people read it as.

The measurements are still recorded and their reasons are longer than they were - what changed is which list they are in, and therefore whether anyone will keep picking them up and putting them down.

What stays outstanding, and it is not obvious

120-measure/timer-ratio:tsc_hz_calibrated:hz stays, even though it sits among the deltas that moved: it is a frequency, orbistoun already claims the reported one, and the two agree to eight figures - so it is a claim that could be made rather than a fact about one machine.

110-modules/tier-probe stays too. "orbistoun models no tiers" is a scope decision, not an impossibility, and the moment it models them the reading is claimable. That is exactly what OUTSTANDING is for.