VM campaign — the whole catalog and every profile on Pop!_OS 24.04¶

Run 2026-09-04 with scripts/vm_campaign.py against a Pop!_OS 24.04 LTS VM (ID=pop, ID_LIKE="ubuntu debian", x86_64). Pop is not a declared target and has no CI container; docs/contributing/vm-testing.md says how to read its results — as evidence for a decision about declaring pop, not as pass/fail against current claims. This page is that evidence. It was the last of the six VMs to be run, and the README had said "queued on a VM that does not exist yet" until it did.

The engine under test was a local integration branch of the day's work; the branches it merged landed on main as PRs #21 (cmake joins the apt set of every cmake build), #22 (wsjtx declares libhamlib-dev) and #23 (yaac re-pinned). The first two were found by this campaign.

Standing¶

Campaign	Result
Every unit, 243, from the clean snapshot	224 installed+confirmed, 17 refused at plan time, 1 stopped at a consent gate, 1 failed — the failure re-pinned and confirmed the same day
Every profile whole, from the clean snapshot	14 of 16 installed+confirmed; `digital-modes` failed twice, was fixed, and confirmed alone; `editors` and `rf-research` were not run (consent gates, D-021)

"Confirmed" is the engine's own bar (D-031): every command completed and the re-probe found the software. Raw reports on the host under ~/.local/state/hammunition-campaigns/: units-pop2404-2026-09-04, units-pop2404-yaac-2026-09-04, profiles-pop2404-2026-09-04, profiles-pop2404-rerun-2026-09-04, profiles-pop2404-digital-modes-2026-09-04 (.md and .log).

The seventeen refusals¶

All at plan time, nothing changed on the machine, each naming its reason. Pop's archive is Ubuntu 24.04's plus Pop's own suites, so this is mostly the 24.04 rung's list seen again (docs/reference/vm-campaign-ubuntu.md).

Unit	Reason
`aethersdr`, `gtk-meshtastic-client`, `m2kcli`, `mlat-client-adsbfi`, `odr-audioenc`, `python3-meshtastic`, `readsb`, `rtl-ais`, `satdump`, `sdrangel`, `voacapl`	apt has no candidate on this release
`pythonprop`	no candidate, and depends on `voacapl`
`arduino-cli`, `soapysdr-module-plutosdr`	their only install block is `distro: [kali]`; no block declares this target
`openhamclock`	the archive's `nodejs` is 18.19; the manifest's floor is 20.19 and Node is never fetched (D-037)
`wsjtx-improved`	its vendor `.deb` collides with the installed `wsjtx-data` — refused before dpkg could half-do it
`noaa-apt`	retired: the satellites it decodes were taken out of service 2025-11-09

That is Ubuntu 24.04's list less two. code and codium were refused there for an engine gap that D-040 has since closed; here codium stopped at the D-040 fingerprint prompt, which the campaign never affirms, and code installed from the archive in 16 seconds — Pop!_OS carries Microsoft's build in its own pop-os-apps suite, so no third-party repository was offered or needed. That is the one result on this page that is Pop's own, and it is D-040's first rule ("added only when the archive offers nothing") holding on a target nobody designed it for.

The two failures, and what each taught¶

yaac — the fetch of YAAC.zip did not match the manifest's sha256; the download was discarded and nothing ran. Upstream had re-cut the beta under the same URL. Re-pinned to build #230 (PR #23) and confirmed alone the same day. This is the fetcher doing its one job. The author's own account of how YAAC is versioned and distributed arrived later as issue #31.

digital-modes — failed twice, at 386 s and 656 s, in the same place: wsjtx's cmake configure could not find Hamlib's development headers. The wsjtx manifest had never declared libhamlib-dev; on every other target it had been building in js8call's wake, which declares it and sorts earlier. Pop was the first machine where the order fell the other way. Declared (PR

22), then the profile ran alone from clean and confirmed in 513 s.¶

The related finding, from the same day's wsjtx build: a cmake build had been relying on cmake already being present, so PR #21 made the engine put it in the apt set of every cmake block rather than hope.

Every profile¶

Seconds are the harness's wall clock from plan to the re-probe's last answer.

Profile	Result	Seconds
`antenna`	confirmed	50
`digital-modes`	confirmed on the third run, after PR #22	386 ✗, 656 ✗, 513
`editors`	not run — consent gate (D-040)	—
`electronics`	confirmed	113
`listening`	confirmed	150
`logging`	confirmed	589
`morse`	confirmed	27
`packet`	confirmed — see below	54
`propagation`	confirmed	527
`rf-research`	not run — consent gate (D-021)	—
`rf-security`	confirmed	18
`rfid`	confirmed	114
`satellite`	confirmed	55
`sdr`	confirmed	170
`station`	confirmed	26
`workstation`	confirmed	4

The row that was true of packages and false of capability¶

packet installed whole and confirmed every check. It could not have worked: the VM's kernel, 7.1.5, carries no ax25 module — Linux 7.1 removed AX.25 — so kissattach had nothing to attach to. The engine had no way to know, because the capability matrix is per target and this is a fact about the running kernel. That finding is D-041 and docs/reference/kernel-ax25.md: the plan now reads /lib/modules/$(uname -r) and refuses or defers by name. Confirming an install by its packages is necessary and not sufficient.

Left with the maintainer¶

Declare pop a target, or not. Nothing on this page refused for being Pop rather than Ubuntu 24.04: every refusal is the archive's or the manifest's, and the ID-only selector caveat in vm-testing.md did not bite: the one manifest that gates on distro: [ubuntu] (sdrangel) also gates on 26.04, so a Pop 24.04 machine refuses it for the release, not the ID. The decision is still open; this is the evidence it did not need to be urgent.
COSMIC menus. Pop!_OS 24.04 is the release that ships COSMIC, so the machine D-036's third mechanism was waiting on now exists. Which desktop this VM booted has not been checked, and nothing about COSMIC has been measured.