run.sh: fail fast when venv package install fails

[ideaship]

Problem

The manager bootstrap run.sh venv seed path installed uv and the
Python/ansible environment without checking exit codes:

pip3 install uv
uv pip install -r requirements.txt
uv pip install "ansible==$ANSIBLE_VERSION"

Unlike the surrounding lines (ansible-galaxy, osism.manager.*), which use || exit $?, these three swallowed any failure. When uv could not install — e.g. it failed to initialize its cache (Failed to initialize cache at ~/.cache/uv: Permission denied) — run.sh proceeded with a half-built environment: requirements.txt (which pulls in ansible plus collection dependencies such as ansible.posix) never installed, ansible-galaxy still installed the osism.* collections, and the operator playbook aborted much later with a misleading couldn't resolve module/action 'ansible.posix.authorized_key' — No module named ansible_collections.ansible.posix, tens of lines removed from the real cause.

Observed live in an autodrive/tbng deploy on regiocloud.

Fix

Guard all three install lines with || exit $? so a uv/pip failure aborts run.sh immediately with uv's own error message instead of producing a broken venv and a confusing downstream failure.

A targeted || exit $? was chosen over set -euo pipefail deliberately: these exits occur before the trap cleanup ERR EXIT is registered (further down the script), so the fix cannot trigger cleanup's rm -rf "$VENV_PATH" on the failure path.

Impact on downstream copies

This file is also vendored by cloud-in-a-box, metalbox, the generics-drift-* repos, and tbng, which pick up the fix on their next sync.

🤖 Generated with Claude Code