# IR-59 instrument — `xtask disk-floor`, the run-start free-space reading The log-the-floor half of [[IR-59]] in `docs/INFRA-REGISTER.md`. Call it as the FIRST step of any rig or gate script, before the first cargo invocation. ``` cargo run -p xtask -- disk-floor --label ``` It prints one line whether it passes or fails: ``` disk preflight: drive=C:\ free_bytes=155184713728 floor_bytes=34359738368 label=gate-w2 ``` ## Why a verb and not a line in a script The entry asks for the reading in "the rig/gate preflight path". There is no such path in this tree to amend — the gate legs that hit this ran as **plain per-run scripts**, which is the entry's own second face. There was nothing to add a line to, so the verb is the template: one command, carrying the floor with it, that a rig script calls at run start. ## Flags | Flag | Default | Means | | --- | --- | --- | | `--path ` | cwd | the volume to read | | `--floor-gb ` | `32` | refuse below this; **`0` is print-only and never refuses** | | `--label ` | none | attributes the reading on a box carrying several lanes | `32` is the golden runner's number, so a local rig and a golden leg refuse at one floor rather than at two. The keys are the golden runner's too (`disk preflight: drive= free_bytes= floor_bytes=`, and `RESOURCE=disk` on refusal), so one grep reads a local rig log and a CI log alike. ## What it is answering A full volume **reds as a linker defect that names no disk**: `LNK1318: Unexpected PDB error` (the parenthesised code varies — do not key recognition on it), often beside `LNK4209: debugging information corrupt`. It has also arrived as a `traceable-reqs check` PANIC carrying `os error 112`, so even a registry red can be the disk's. Nothing in those texts says "disk". Two agents spent hours on a mechanism for one of them and neither read free space. **A run that starts under the floor produces reds that belong to nothing. They are re-run, never re-read.** ## The limit, and it is the important sentence **One reading is an INSTANT, not headroom** (IR-46). The danger window is the TAIL of a cold build, not a steady state you can check once: the measured instance died linking the last and largest artifact of a 74 GB pool while that same build was draining the volume out from under itself, and a reading taken before it started looked fine. A green preflight says the run was allowed to start. It is not a guarantee about the run. The lever this leaves alone is pool accumulation — a pool re-accumulated to 152–169 GB within one day of a clean, and a reap is an instant too. Free space at run start is the falsifier for a red in hand; `xtask pool-sweep` is what keeps the number healthy.