fix: harden repository refresh and server pull
ForgeFlow quality gate / quality (push) Canceled after 0s
ForgeFlow quality gate / quality (push) Canceled after 0s
This commit is contained in:
@@ -0,0 +1,10 @@
|
||||
# ForgeFlow 0.10.11
|
||||
|
||||
## Resilient repository refresh and consistent server pull
|
||||
|
||||
- Temporary Gitea list failures now use the in-session **last-known-good** repository inventory while local and server state continue to refresh. The UI clearly reports that remote data is stale.
|
||||
- A **closed output pipe** from a detached parent process is no longer treated as a fatal desktop-app exception.
|
||||
- Server pull, deploy-key verification, deployment, rollback and metadata now consistently prefer the verified **linked checkout origin** over a stale URL detected earlier on the server.
|
||||
- Repository-scoped **read-only deploy key** checks remain fail-closed; a changed SSH host still requires explicit trust and access reconfiguration.
|
||||
- The local **browser test server** now has a dedicated port and identity endpoint, preventing another localhost application from being mistaken for ForgeFlow.
|
||||
- All 42 responsive browser flows pass across dark/light, compact/wide and reduced-motion configurations.
|
||||
Reference in New Issue
Block a user