How this was tested
The install, deploy, rollback, failure and backup paths have been run end
to end on fresh Debian 12 and Ubuntu 24.04 machines with systemd. The test
site was https://localhost with Caddy's internal certificate authority,
and a local S3-compatible server stood in for object storage. That was
everything except a real domain, a real mail provider and real EU object
storage. docs/status/H-operations.md summarizes what each run found.
On every CI run (docs/status/P7-ops-tests.md and Q3-quality-ops.md list
what each proves):
install.sh,install-bugsink.sh,deploy.sh,rollback.shandreportlogrun against a pretend server in a temporary directory, with systemd, apt, pip, Caddy and the build as stand-ins, answering the installers' questions on standard input (apps/api/test/unit/deploy-install.test.ts,deploy-release.test.ts,deploy-bugsink.test.ts);- the secret settings against
.env.example, the units and the installers (secrets.test.ts), and the secret scan against planted keys (secret-scan.test.ts); - this Caddyfile, in a real Caddy (
deploy-caddy.test.ts), and with Bugsink's site added (deploy-bugsink.test.ts); - the storage check, against moto (
deploy-storage-check.test.ts); - the doctor's migration check, against real databases in each drift state
(
apps/api/test/db/deploy-schema-drift.test.ts).