Add missing exec to non-root branch of svc-qbittorrent s6 run script

[jhogendorn]

---

• I have read the contributing guideline and understand that I have made the correct modifications

---

Description:

Add the missing exec keyword to the non-root branch of the svc-qbittorrent s6 run script. Without it, bash spawns s6-notifyoncheck -d ... as a child; the -d flag causes s6-notifyoncheck to daemonize, the bash script exits, and the s6-rc supervisor tears down the service group — SIGTERMing the orphaned qBittorrent process within ~1s of every start.

The root branch has used exec since c828bd5c (2022-09-20, "Exec notify"). The else branch was added in 016be7b8 (2024-12-20, "Support nonroot operation") without the matching exec. The bug was latent for ~17 months, likely made deterministic by 5152b908 (2026-05-04, "Switch to static builds for v2") which changed startup timing.

Closes #435

Benefits of this PR and context:

Restores non-root operation (k8s securityContext.runAsUser, Podman rootless, any environment that sets LSIO_NON_ROOT_USER). Currently-affected users have to pin to 5.1.4-r2-ls443 or earlier.

Distinct from #430 / #432 / #433 / #434 (upstream qBittorrent lockfile race, resolved by deleting the lockfile) — those reports describe "WebUI not accessible" or "container hangs" without exit signal. This bug terminates the container with exit 137 within ~1s and tight-loops.

How Has This Been Tested?

Root-cause established by static analysis:

1. Read current root/etc/s6-overlay/s6-rc.d/svc-qbittorrent/run on master — confirmed asymmetric exec (root branch has it, else branch doesn't).
2. Traced introduction to 016be7b8 via git log -- root/etc/s6-overlay/s6-rc.d/svc-qbittorrent/run — the else branch was added without the matching exec.

Triggering symptom observed in production: qBittorrent termination initiated log line within ~1s of startup, exit 137, tight pod restart loop on 5.2.0_v2.0.12-ls455 in a k8s cluster with securityContext.runAsUser set to a non-zero UID. Workaround currently in place is pinning to 5.1.4-r2-ls443.

End-to-end validation of the patch (build the fixed image, deploy to the affected k8s cluster, watch a non-root pod run cleanly) has not yet been performed — submitting on the strength of the static analysis since the asymmetry is unambiguous and the fix is the minimal change to restore symmetry with the working root branch. Happy to test against a CI build of the patched branch if that's preferred before merge.

Source / References:

• Asymmetric introduction: 016be7b8
• Likely race-exposing commit: 5152b908
• Original exec notify on root branch: c828bd5c
• Related (but distinct, lockfile-race) issues: [BUG] 5.2.0_v2.0.12-ls454 container hangs on startup with no error message #430, [BUG] qb 5.2.0 webui not started (KNOWN BUG NEED TO DELETE LOCKFILE) #432, [BUG] 5.2.0_v2.0.12-ls455 image is not starting #433, [BUG] WebUI inaccessible after latest update #434

[github-actions]

Thanks for opening this pull request! Be sure to follow the pull request template!

[thespad]

Thanks for catching this.