Skip to content

test(conformance): Run the suite on hull on Linux - #680

Merged
ananos merged 1 commit into
mainfrom
feat/conformance-hull-linux
Nov 20, 2026
Merged

ananos merged 1 commit into
mainfrom
feat/conformance-hull-linux

Conversation

@ananos

@ananos ananos commented Nov 19, 2026 •

Copy link
Copy Markdown
Member

Summary

The conformance suite (#264) runs on hull with hvi on macOS and, since #513, on nerdctl on Linux. hull on Linux is the third path, and the one we want to make the Linux default, so it needs the same suite and the same kind of record before that flip.

This adds a hull-linux backend. It boots through brig with BRIG_RUNTIME=hull, probes through brig sh as on hvi, and files its record under the runtime bundle's tag: the bundle builds hull from a commit, so hull's own version names no release, and hull's tags read like the bundle's.

#513 and #675 are on main. A real run of the full suite also wants #682, so no record is committed here yet.

Related issues

Refs: #264

Changes

  • test/conformance/backend.go: the hull-linux backend. The run is refused unless the commit in hull --version is the bundle's HULL_REF; the stack rows come from the bundle's pins.env (hull, urunc, hvi, hull-assets, containerd, rootlesskit, slirp4netns), the boot assets from provenance.json, the image digest from hull inspect. nerdctl's rootless check and bundle row are shared helpers now.
  • test/conformance/runner.go, cases.go, main.go, record.go: the case, BRIG_RUNTIME=hull for this backend only, the metadata reason as on hvi (hull's gateway resets 169.254.0.0/16).
  • script/egress-conformance.sh -backend hull-linux, refused off Linux.
  • Docs: releasing.md, manual-tests/egress-policy-linux.md, policies.md.
  • Tests in test/conformance/hulllinux_test.go, negative ones included: another hull commit, a hull tag that reads like the bundle's, no commit, nerdctl in hull's place, no HULL_REF, no bundle. Each is refused, boots nothing and writes no record.

Checklist

  • make all passes (vet, test, build)
  • script/smoke.sh passes
  • I have added or updated tests covering the change
  • I have run go test ./... -race, if the change touches concurrency, subprocesses or the daemon
  • I have exercised the change against a real runtime (brig run <agent>), if it touches the run, exec or credential path
  • I have updated the affected docs (see docs/README.md for the map)

Not run on a real host yet: it needs #675 and #682 first. The record from the first run lands in a follow-up commit here.

@ananos
ananos marked this pull request as ready for review November 19, 2026 08:36
@margkeka
margkeka force-pushed the fix/264-conformance-nerdctl branch from 38d3029 to 582655d Compare November 19, 2026 12:40
Base automatically changed from fix/264-conformance-nerdctl to main November 19, 2026 12:48
@ananos
ananos force-pushed the feat/conformance-hull-linux branch 2 times, most recently from d5c39d0 to 65c1d1d Compare November 19, 2026 15:49
@ananos
ananos force-pushed the feat/conformance-hull-linux branch from 65c1d1d to 10532cd Compare November 20, 2026 10:17
hull on Linux now boots its guest behind hull's own gateway, the one
that enforces a policy on hvi, but the suite booted only hvi and
nerdctl. So there was no record a release could gate on for it.

The hull-linux backend boots the hull the runtime bundle ships. Its
record is filed under the bundle's tag, not hull's version: the bundle
builds hull from a commit, so hull reports a pseudo-version, and a hull
tag such as v0.1.0-rc22 reads like the bundle's own. The stack rows
come from the bundle's pins: hull, urunc, hvi, hull-assets and
containerd. The run is refused, boots nothing and writes no record
when the hull that boots is not built from the commit the bundle pins,
or when there is no bundle at all.

The bundle's environment says BRIG_RUNTIME=nerdctl unless the shell
said hull first. So the suite runs brig with BRIG_RUNTIME=hull for this
backend, through the brigRuntime field the nerdctl backend uses, and the
script asks brig-env.sh for the bundle's hull. The metadata case gets hvi's reason for an unproven target,
since the gateway is the same. No verdict rule changes.

Refs: #264
Signed-off-by: Anastassios Nanos <ananos@nofire.ai>
@ananos
ananos force-pushed the feat/conformance-hull-linux branch from 10532cd to cfbd9ea Compare November 20, 2026 10:21
@ananos
ananos merged commit 41b422b into main Nov 20, 2026
9 checks passed
@ananos
ananos deleted the feat/conformance-hull-linux branch November 20, 2026 10:25
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants