SD shortfin CI fails when apt is locked

[renxida]

Should probably first check if cmake & ninja are installed & of the proper version before attempting to install. I don't think checking requires acquiring the apt lock.

[ScottTodd]

Workflow history: https://github.com/nod-ai/SHARK-Platform/actions/workflows/ci-sdxl.yaml
Sample logs: https://github.com/nod-ai/SHARK-Platform/actions/runs/11621536945/job/32366319347#step:2:13

Text:

Run sudo apt update -y

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.

Reading package lists...
E: Could not get lock /var/lib/apt/lists/lock. It is held by process 2271375 (apt)
E: Unable to lock directory /var/lib/apt/lists/

(screenshots aren't searchable of copy/paste-able ;P)

[monorimet]

Added a fix to #411, but this still will throw an error if we don't have those installed and cannot acquire the lock. May not be an issue for a while.

output logs:
https://github.com/nod-ai/SHARK-Platform/actions/runs/11637357380/job/32410398564?pr=411#step:2:1

[ScottTodd]

Added a fix to #411

Please keep changes small and send them as individual PRs. We shouldn't need to wait for a larger patch to be reviewed and landed to fix a broken CI build.

Something is suspicious here about the runner environment. Basic dependency setup / installation shouldn't be a sticky source of failures. Might make sense to run these jobs in containers or limit what other code (if any) is running on these machines.

[monorimet]

Added a fix to #411

Please keep changes small and send them as individual PRs. We shouldn't need to wait for a larger patch to be reviewed and landed to fix a broken CI build.

Something is suspicious here about the runner environment. Basic dependency setup / installation shouldn't be a sticky source of failures. Might make sense to run these jobs in containers or limit what other code (if any) is running on these machines.

Agree, will send up separate PR.

Containers are probably a bit heavy for this, but as long as we have env/machine setup flexibility validated within reason, it might be the most robust solution.

[ScottTodd]

I'm specifically wondering if apt update within a container is free from the hosting system's /var/lib/apt/lists/lock. We wouldn't need to use a container for packaging deps (we could)... just for isolation from other users on the runner machine.