TSTATE — Track T watcher index (generated by tools/twatch.py)

host last tested date verdict wall full through
borg 24455e065275 2026-09-22T09:31:11Z RED (native) 299.2s 41f94503bb89 RED
plexus (retired 2026-09-11T20:19:53Z → borg) 27424c927b65 2026-08-30T10:24:09Z RED (full) 1370.0s 27424c927b65 RED
seven (retired 2026-09-11T16:29:49Z → plexus) 120eeb39fd48 2026-09-11T16:28:30Z GREEN (native) 188.5s ae2280f1aa46 RED
xeon (retired 2026-08-07T16:44:07Z → plexus) 0db7276f06a0 2026-08-04T23:13:51Z RED (native) 124.5s 7d8929633721 GREEN

Cross-target currency — which host's map to read

A host's jobs map is only as current as that host's own last FULL tier. quick, native and limited run no cross target, so every i386 / arm32 / aarch64 / riscv32 / xtensa entry in a host's state dates from its last full run — however recently that host published something else.

Newest full tier in the fleet: 41f94503bb89 on borg, 2026-09-22T08:54:44Z (36m ago).

host full through verdict age behind the newest by
borg 41f94503bb89 RED 36m — (newest)

Reading a staler host's map for a cross-target job answers a question about an OLDER tree, and it is what makes an already-fixed job still read fail.

...and which TOOLCHAIN measured it

host toolchain fp
borg kernel=7.0.0-31-generic gcc=13.3.0 qemu=8.2.2(6 of 6) espqemu-riscv32=9.2.2 espqemu-xtensa=9.2.2 git=2.43.0 wasmtime=48.0.1 892e942692d7

Two hosts with different fingerprints did not measure the same thing, and a job that disagrees between them may be disagreeing about the EMULATOR rather than about the tree. Check this before filing a cross-target red against the compiler: bug-t-tstate-fingerprints-the-code-and-the-hardware-but-not-the-emulator-toolchain is the incident that cost an afternoon for want of this row.

Open regressions