- manage_runner.sh: fix RUNNER_DEFAULT_IMAGE clobbering by renaming
per-runner var to RUNNER_SECTION_IMAGE; .env fallback now works
- manage_runner.sh: render native runner config.yaml before registration
so act_runner can read it during --config flag
- manage_runner.sh: add SSH credential validation for remote hosts
(fail early with clear error instead of cryptic SSH failure)
- phase1/phase2: add UNRAID_DB_IP/FEDORA_DB_IP to conditional
require_vars when DB_TYPE != sqlite3
- cleanup.sh: only clear manifest when all actions for host succeeded;
failed actions are preserved for retry
- phase8_cutover.sh: strip empty environment: block from Caddy
docker-compose when TLS_MODE=existing
- phase5_migrate_pipelines.sh, phase5_teardown.sh, phase9_teardown.sh:
wrap cd+git in subshells so working directory is always restored
- phase3_post_check.sh: handle both string and numeric runner status
from Gitea API (offline vs 2)
- configure_env.sh: fix TOTAL_PROMPTS base count (63->64) and move
DB/repo count adjustments before their prompts are shown
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Identical copies of _set_db_vars() and _strip_block() existed in
phase1, phase2, and phase8. Moved to lib/common.sh as set_db_vars()
and strip_template_block() (public API, no underscore prefix).
Removed dead _strip_block definition from phase8 (defined but never called).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace host port vars with macvlan vars in require_vars. Add Step 2
to create the macvlan gitea_net network on Unraid. Update
docker-compose rendering to use GITEA_CONTAINER_IP and
DB_CONTAINER_IP instead of port mapping. Renumber steps accordingly.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add _set_db_vars() helper to configure DB-specific docker-compose
vars (image, env, healthcheck, data dir) for postgres/mysql/mssql.
Add _strip_block() to remove conditional template markers. Update
docker-compose and app.ini rendering to strip/populate DB blocks
based on GITEA_DB_TYPE. Conditionally require DB connection vars.
Create DB data directory for external databases.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- teardown_all.sh: replace `yes |` pipeline with `< <(yes)` process
substitution to avoid SIGPIPE (exit 141) false failures under pipefail
- phase6_teardown.sh: extract push mirror `.id` instead of `.remote_name`
to match the DELETE /push_mirrors/{id} API contract
- phase5_migrate_pipelines.sh: expand sed regex from `[a-z_]*` to
`[a-z_.]*` to handle nested GitHub contexts like
`github.event.pull_request.number`
- lib/common.sh: render_template now requires explicit variable list to
prevent envsubst from eating Nginx variables ($host, $proxy_add_...)
- backup scripts: remove MacBook relay, use direct Unraid↔Fedora SCP;
fix dump path to write to /data/ (mounted volume) instead of /tmp/
(container-only); add unzip -t integrity verification
- preflight.sh: add --skip-port-checks flag for resuming with
--start-from (ports already bound by earlier phases)
- run_all.sh: update run_step to pass extra args; use --skip-port-checks
when --start-from > 1
- post-checks (phase4/7/9): wrap API calls in helper functions with
>/dev/null redirection instead of passing -o /dev/null as API data
- phase8: replace GitHub archiving with [MIRROR] description marking
and disable wiki/projects/Pages (archived repos reject push mirrors)
- restore_to_primary.sh: add require_vars for Fedora SSH variables
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- phase1_gitea_unraid.sh: 9-step deploy (dirs, docker-compose, app.ini,
container start, wait, admin user, API token, save to .env, create org).
Every step has idempotency check — running twice changes nothing.
- phase1_post_check.sh: 5 independent verification checks
- phase1_teardown.sh: stop container + optionally remove data, with prompts
Also adds inline comments to lib/common.sh and preflight.sh explaining
WHY decisions were made (SSH flags, API tmpfile pattern, port checks, etc.)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
