May Track T install qemu-system and a FreeBSD image on plexus?
- Type: decision (Track U — the owner's machine).
- Filed 2026-08-28 by Track T (pxx-a5) after completing step 1 of [[feature-t-freebsd-image-and-runner]]. This is a question, not a task someone forgot to do — the work is understood and small; only the consent is missing.
- Unblocks:
feature-t-freebsd-image-and-runner→feature-port-freebsd-native(top of Track A's ready queue, repeatedly picked up and put back down).
Why you are being asked rather than told
plexus is your workstation as well as the test box. Installing a system emulator and downloading a multi-gigabyte OS image changes its configuration and its disk, which is different in kind from the test-infra work Track T normally does on it unasked. The same instinct kept a 251s test sweep off the box earlier at load 27; this is that instinct applied to packages and disk.
Nothing has been installed, downloaded, created or booted. Step 1 was read-only.
What was established first, so this is not a speculative ask
- There is no FreeBSD image on this machine, and the belief that one existed
somewhere was a misreading — the phrase "pre-built qcow2 exists" in
feature-port-freebsd-nativeis a statement about what FreeBSD.org publishes, filed as a contrast with OpenBSD's "no pre-built qcow2" in the same commit. Not stale, not about another box. That lead is closed. - The real gap is narrow:
qemu-useris installed,qemu-systemis not. That is why the cross-target tiers work (foreign binaries, user-mode) and nothing here can boot a kernel. - Disk is a non-issue:
/has 96G free,/datahas 2.0T.
What is actually being asked for
- A package:
qemu-system-x86(Ubuntu). Tens of MB plus dependencies. - An image: the official FreeBSD/amd64 qcow2 VM image from
download.freebsd.org. Exact size not stated here because it was not fetched — the compressed VM images are in the hundreds-of-MB-to-low-GB range and expand to several GB. It would live under/data, not/. - Running it: a headless boot during the test job — RAM in the 1–2 GB range while running, nothing resident between runs.
The three answers, any of which is useful
- Yes, go ahead — Track T installs the package, fetches the image under
/data, and writes the runner. The ticket resumes at its step 2. - Not on plexus — then say which box, or that the FreeBSD port waits for
one.
feature-port-freebsd-nativemoves from "blocked on infra" to "blocked on a machine", which is a truer label and stops it being re-picked. - You install it yourself — name where the image lands and Track T writes the runner around it. Fine, and arguably better: it keeps package changes on your side of the line entirely.
Why it is prio 55 and not lower
Ranked to match what it blocks, not to its own size: it is the sole gate on the Track T ticket that is the sole gate on the top Track A item. Per [[decide-t-should-a-skip-close-an-open-regression]]'s note about decisions and ranking, a decision left unreached is an answer given by default — and the default here is "the FreeBSD port stays unstartable indefinitely", which is a real outcome nobody has chosen.
Not urgent
Nothing is broken and nothing regresses while this waits. It is ranked to be reached, not to interrupt.
APPROVED 2026-08-31 — install qemu-system, pull a FreeBSD image
Owner: "yes this box is dedicated to development. i think we have plenty disk space left. so yes, we can pull a BSD image."
Verified at ruling time rather than assumed: plexus root filesystem is 156G with 84G available (44% used). A multi-GB OS image is comfortable.
This was the owner's call for the reasons the ticket gives — it costs disk, it pulls a multi-GB image over the network, and installing a system emulator changes the box for every lane. All three are his to spend.
Build it on plexus, and this is now load-bearing
Two fleet facts the owner stated the same day, recorded here because they decide where this goes and are recorded nowhere else:
- borg is being repaired this week.
- seven is moving off-site to a work location — it is too noisy to keep running at home.
So plexus is the host for the FreeBSD VM, not seven. Anything built assuming seven stays reachable at home should be re-checked against that.
Not approved here
Nothing about what the FreeBSD port must then pass. This clears the blocker — that the port had no bootable kernel to test against — and no more.
Approved 2026-08-31 by the owner; disk verified by frank-user.