You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Fixed: proxied fetch failures carry the upstream's real status and a reason. Every failure path in the indexd /v2 proxy answered a bare empty 502, which told a client nothing: an upstream 404 (missing ref), a 429 rate limit, and a dead relay thread were indistinguishable from outside. The upstream registry's own status now reaches the client as itself, with its Retry-After forwarded and a one-line reason in the body; transport and token-dance failures stay 502, but say which. This matters more as machine consumers arrive (root-driven --assumeyes pulls from long-lived machines need the 429 to back off), and it is the fix for the hosted remote's opaque breakage: the empty 502 was indexd's own shape, not the reverse proxy swallowing a body. The hosted remote at apps.orchard.rocks was redeployed onto this release, which closed the outage: its stale binary predated the v0.5.3 Accept fix and every manifest fetch through the remote failed.
The store page says what the remote serves. The app list states up front that the catalog covers OCI-image artifacts only: a bundle-only .flatpak app never appears through a remote.