davsclaus opened a new pull request, #29653: URL: https://github.com/apache/camel/pull/29653
The second batch of [CAMEL-26864](https://issues.apache.org/jira/browse/CAMEL-26864): bugs 1, 2, 3, 8 and 9, the ones that needed a live reproduction. Each one was reproduced on a TUI built from `main`, then checked on a TUI built from this branch. Both ran in tmux against the integration from the issue and were driven through the TUI's MCP server. The first batch is #29639. The UX items (10, 12, 15, 17) come next. **Diagram: the screen did not follow a drill-down, and showed another route's nodes (1, 2)** Both had one cause. After Esc up to the Architecture level, `tui_navigate(route=hello, node=choice1)` drilled into hello, but the Architecture view stayed active. While it is active, `render` draws it and never the route, so: - the screen kept showing the Architecture while `tui_get_state` said `diagramMode: route` (2) - the route's node boxes stayed those of the last route drawn (`[seven, log2]`), so `choice1` was "not found" (1) Drilling into a route, or selecting one in the topology, now closes the Architecture view. `diagramMode` reports the level the screen shows, including `architecture`. **Inspect: the Diagram view showed an old exchange (3)** The Diagram view (`d`) draws the exchange the history had when the view opened. Opening the tab again, or F5, reads the history again, but the diagram kept drawing the old exchange. When the history now holds another exchange, the diagram is drawn for it. It is not redrawn when the exchange is the same, so following (`f`) doesn't flicker. **Endpoints: out 0, but the avg body chart said out 7 B (8)** #29510 deliberately leaves out of the flow the hand-offs between the routes (direct, seda): hello → direct:seven → seven is neither an out nor an in. The counts and the msg/s chart did that, but the avg body chart still took the largest mean body size of every endpoint. It now leaves out the hand-offs as well (`EndpointFlow.meanBodySize`). Out 0 stays, as the help page explains. **`tui_send_keys` had no effect while the AI panel was open (9)** Keys from an MCP client went through the normal key handling, and the open AI panel typed them into its input. Keys from `tui_send_keys` are now marked as MCP keys and go to the active tab. The AI panel's input gets only what the user types, and a tape still types into it as it did. The result says where the keys went: "Queued 1 key(s) … for the Diagram tab", and `activeTab` when called with `wait`. Tests: `DiagramTabArchitectureNavigationTest` (fails without the fix), `EndpointFlowTest`, `ConfirmationsTest`. Found along the way, not in this PR: `tui_get_diagram` has returned "No diagram available" since #23772 removed the text diagram (`DiagramTab.getTableDataAsJson` returns null). _Claude Code on behalf of Claus Ibsen_ 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
