Same fixes as the previous commits (run no egui pass at all when
nothing will be shown), but the mechanisms are shared instead of
duplicated:
* `Context::run_frame(input, show_ui, f)` is the one entry point for
integrations: one `FramePhase::Logic` (always, outside any pass),
then one `FramePhase::Ui` per pass (none when `show_ui` is false).
`Context::run_logic` is now a thin wrapper around it, and the
logic-outside-of-pass sequencing lives in egui, not in each backend.
* egui itself buffers the input of pass-less frames
(`ViewportState::pending_raw_input`) and prepends it to the next
pass. This replaces `EpiIntegration::pending_raw_input` and the web
backend's `input.raw.append`, so an integration cannot lose input.
* `FullOutput` is now `{ platform_output, viewport_commands,
pass_output: Option<PassOutput> }`, and `LogicOutput` is gone.
One-shot viewport commands (imperative) are separated from
`ViewportOutput` (which viewports should exist - declarative), so
each backend has exactly one command-handling path, shared by frames
with and without a pass. `pass_output: None` encodes "no pass ran:
paint nothing, leave the viewports alone" in the type.
* The glow/wgpu `!show_ui` early-return blocks are gone: a hidden root
viewport flows through the same tail as a visible one, with the
paint and viewport-structure steps gated on `pass_output`.
Behavioral fixes that fall out:
* `App::logic` now sees the current window state in visible frames
too (it was one frame stale outside the hidden path).
* Auto-save keeps working while a window is minimized or occluded.
* Commands sent to a freshly created viewport apply in the same frame.
* The Wayland resize workaround now also covers commands from
pass-less frames.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
When an agent screenshots the app using the mcp and then interacts with
the app by clicking at coords, they see the coords in the native image
coords. Since the mcp does everything else in logical coordinates, it
helps if the image they see is also in logical resoltution, so we always
downscale it to 1.0.
I've added this here to avioid having to decode and re-encode the image
in the mcp.
Unfortunately it only does downscaling for now, since adding some way to
upscale the image just for the screenshot would add a lot of complexity,
and might be invasive from a plugin.
I've also changed the submit call to take a closure, to make it easier
to use other transport channel (makes the implementation for reruns mcp
nicer).
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Lucas Meurer <lucas@rerun.io>
Introduces live inspection for running egui apps over a small TCP
request/response protocol, plus the `egui::Plugin` that serves it.
This is the minimal surface to get the egui mcp in, we may want to
extend this in the future to add support for the inspection gui.
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
