cd4a6bd3a8
The 5 binary downloads used a bare `curl -fSL` with no retries. The assets are served from velxio.dev, which has brief unavailable windows during a deploy (the app container is recreated -> a few seconds of 502). A single blip mid-download failed the whole Backend E2E job even though nothing was wrong with the change under test. Add a shared retry policy (--retry 5 --retry-delay 10 --retry-all-errors --retry-connrefused --connect-timeout 15) so the step rides out a container recreate (~50s of headroom) instead of failing hard. |
||
|---|---|---|
| .. | ||
| workflows | ||
| FUNDING.yml | ||