feat: fresh Ubuntu testing#2312
Conversation
paul-nechifor
requested review from
leshy,
mustafab0 and
spomichter
as code owners
Greptile SummaryAdds a local VirtualBox-based test harness (
Confidence Score: 4/5Safe to merge as a local tooling addition; no production code is affected, but vmtest.sh has a logic flaw in ISO handling that will cause confusing build failures for anyone whose download was interrupted. The misc/fresh-ubuntu-tests/vmtest.sh — specifically the Important Files Changed
Sequence DiagramsequenceDiagram
participant User
participant vmtest.sh
participant VBoxManage
participant GoldenVM
participant CloneVM
participant suite/run.sh
Note over User,suite/run.sh: ./vmtest.sh build (once)
User->>vmtest.sh: build
vmtest.sh->>vmtest.sh: ensure_iso() — wget ISO
vmtest.sh->>VBoxManage: createvm + unattended install
VBoxManage->>GoldenVM: start headless, install Ubuntu
GoldenVM-->>vmtest.sh: SSH ready (/etc/dimos-vmtest-ready)
vmtest.sh->>VBoxManage: power off + snapshot "golden"
Note over User,suite/run.sh: ./vmtest.sh run (each time)
User->>vmtest.sh: run
vmtest.sh->>VBoxManage: clonevm golden --options link → CloneVM
VBoxManage->>CloneVM: startvm headless
CloneVM-->>vmtest.sh: SSH ready
vmtest.sh->>CloneVM: scp suite/
vmtest.sh->>CloneVM: ssh bash suite/run.sh
CloneVM->>suite/run.sh: setup.sh (apt-get, uv, git clone)
suite/run.sh->>suite/run.sh: 01-mypy.sh
suite/run.sh->>suite/run.sh: 02-pytest.sh
suite/run.sh->>suite/run.sh: 03-cyclonedds.sh
suite/run.sh-->>vmtest.sh: PASS / FAIL + exit code
vmtest.sh->>CloneVM: scp logs back
vmtest.sh->>VBoxManage: remove CloneVM (trap EXIT)
Reviews (1): Last reviewed commit: "feat: fresh ubuntu testing" | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| if [[ -f "$ISO_PATH" ]]; then | ||
| return | ||
| fi | ||
|
|
||
| wget --tries=3 --continue -O "$ISO_PATH" "https://releases.ubuntu.com/$RELEASE/$fname" \ | ||
| || die "download failed" |
There was a problem hiding this comment.
if [[ -f "$ISO_PATH" ]]; then return; fi guard fires as soon as the file exists — even if a previous wget was interrupted mid-download. The --continue flag on the wget call below is unreachable in this case, so the partial file is passed to VirtualBox as-is. The unattended install will then fail with a cryptic installer error rather than a clear "ISO is corrupt" message. The fix is to skip the early return and let wget --continue resume (or verify the file is complete after creation).
| if [[ -f "$ISO_PATH" ]]; then | |
| return | |
| fi | |
| wget --tries=3 --continue -O "$ISO_PATH" "https://releases.ubuntu.com/$RELEASE/$fname" \ | |
| || die "download failed" | |
| wget --tries=3 --continue -O "$ISO_PATH" "https://releases.ubuntu.com/$RELEASE/$fname" \ | |
| || die "download failed" |
| readonly RELEASE="24.04" | ||
| readonly VM="dimos-vmtest-base" |
There was a problem hiding this comment.
4) in the ISO filename is hardcoded separately from the RELEASE constant. When Ubuntu publishes 24.04.5, this script will silently keep downloading the older .4 image (or get a 404 if Canonical eventually removes it). Deriving the full filename from a single versioned constant keeps them in sync and makes future bumps a one-liner.
| readonly RELEASE="24.04" | |
| readonly VM="dimos-vmtest-base" | |
| readonly RELEASE="24.04" | |
| readonly RELEASE_POINT="24.04.4" # bump when Ubuntu publishes a new point release | |
| readonly VM="dimos-vmtest-base" |
Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
| ensure_iso() { | ||
| need wget | ||
| mkdir -p "$ISO_DIR" | ||
| local fname="ubuntu-24.04.4-desktop-amd64.iso" | ||
| ISO_PATH="$ISO_DIR/$fname" | ||
|
|
||
| if [[ -f "$ISO_PATH" ]]; then | ||
| return | ||
| fi | ||
|
|
||
| wget --tries=3 --continue -O "$ISO_PATH" "https://releases.ubuntu.com/$RELEASE/$fname" \ | ||
| || die "download failed" | ||
| } |
There was a problem hiding this comment.
SHA256SUMS and SHA256SUMS.gpg alongside every ISO at https://releases.ubuntu.com/24.04/. Without verifying the checksum, a corrupted download (e.g., from a flaky connection, CDN error, or partial wget resume misalignment) passes silently into VirtualBox. Adding a sha256sum -c step after wget completes — using the checksum fetched from Ubuntu — would catch bad downloads before the ~15-30 min install begins.
| readonly MEM=12288 # MiB | ||
| readonly CPUS=8 | ||
| readonly DISK=40960 # MiB (dynamic; only grows as used) | ||
| readonly SSH_PORT=2222 |
There was a problem hiding this comment.
SSH_PORT shared between golden VM and all clones. The golden VM's NAT rule binds 127.0.0.1:2222 → :22 at build time. Every linked clone inherits the same rule. If the golden VM is still running (e.g., left up after a previous build) when run starts its clone, both VMs compete for port 2222; VirtualBox doesn't error out but the SSH connection may reach the wrong VM or fail intermittently. A simple guard (vm_state "$VM" != "running") at the start of cmd_run — or stopping the golden VM if it is running — would prevent this.
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
1 participant
Problem
Closes DIM-968
Solution
How to Test
Contributor License Agreement