One-shot services are reported as startup timeouts and failures can exit 0

[thromel]

Summary

Short-lived services that exit successfully are reported as startup timeouts, and some failed starts still cause container-compose up to exit with status 0.

This makes one-shot Compose jobs and CI usage hard to trust.

Environment

- OS: macOS 26.5.1 (25F80)
- Xcode: 26.5 (17F42)
- container: 1.0.0 (release, commit ee848e3)
- container-compose: 1.0.0 installed via Homebrew
- Architecture: arm64

Reproduction: successful one-shot service

compose.yml:

services:
  cmdonly:
    image: busybox:latest
    command: ["sh", "-c", "echo CMD_OK"]

Run:

container-compose --file compose.yml up -d cmdonly
container list --all | grep cmdonly
container logs feature-isolation-cmdonly

Actual behavior

container-compose reports:

Error Domain=ContainerWait Code=1 "Timed out waiting for container 'feature-isolation-cmdonly' to be running."

But the container actually ran successfully and exited:

feature-isolation-cmdonly  docker.io/library/busybox:latest  linux  arm64  stopped

Logs:

CMD_OK

The same happened for a locally built one-shot image:

build-built  local-compose-build-test:latest  linux  arm64  stopped

Logs:

from-build-arg

Expected behavior

If a service process exits before reaching a long-running running state, container-compose should inspect the exit result:

• exit code 0: treat as a successful completed one-shot service
• non-zero exit code: report failure and return a non-zero CLI exit status
• still booting/running: continue the existing wait behavior

This would match common Compose usage for migration jobs, probes, build checks, and test containers.

Related status-code issue

In a separate named-volume start failure, container-compose up -d printed an internal error but still returned shell exit code 0. That should return non-zero so scripts and CI can detect failure.

[ryan106]

This looks fixed by #119 (merged yesterday) — I tested both halves of the repro against current main:

Successful one-shot:

services:
  cmdonly:
    image: busybox:latest
    command: ["sh", "-c", "echo CMD_OK"]

container-compose --file compose.yml up -d cmdonly completes cleanly, no "Timed out waiting..." error, container reaches .completed state correctly.

Failing one-shot:

services:
  failer:
    image: busybox:latest
    command: ["sh", "-c", "echo FAILING; exit 7"]

container-compose --file compose.yml up -d failer prints Error: Service 'failer' failed to start (container run exited with status 7). and the process itself exits with status 1 (verified via echo $? directly on the binary, not through a pipe).

Both cases go through ServiceStartState/waitUntilServiceStarted in ComposeUp.swift, which now correctly distinguishes still-starting vs. .completed (exit 0) vs. a thrown ComposeError.containerRunFailed (non-zero exit). Looks like this can be closed.

[thromel]

Closing as completed by #119. Current main handles successful one-shot services as completed and returns a non-zero process status for failing one-shot services.