Retype App::backend to dyn Backend — remove the concrete GtkBackend coupling that gates src/app.rs (unblocks #47)

[JDonaghy]

Problem

src/app.rs:1 describes itself as "struct App — the backend-neutral editor shell application". It is not neutral: src/app.rs:592 types the backend field as the concrete GTK struct.

pub(crate) backend: Rc<RefCell<backend::GtkBackend>>,

That single line is what keeps src/lib.rs:45-46 gating the whole 7,131-line module behind #[cfg(feature = "gui")], and it is the stated Stage-1 blocker on #47 (native macOS GUI).

The blocker is gone upstream. quadraui#699 (filed 2026-09-03 16:38Z, closed 17:11Z, PR#700 / 88345fb) put both handle accessors on the Backend trait, and that commit is already inside vimcode's pinned rev:

	Location
Backend trait	quadraui/src/backend.rs:745 fn modal_stack_handle(&self) -> Rc<RefCell<ModalStack>>, :750 fn drag_state_handle(&self) -> Rc<RefCell<DragState>>
GtkBackend	gtk/backend.rs:1439, :1447
TuiBackend	tui/backend.rs:967, :971
MacBackend	macos/backend.rs:695, :699
WinBackend	win/backend.rs:762, :766

App's 19 calls (modal_stack_handle ×12 at src/app.rs:1144, 1315, 2166, 2214, 2226, 3594, 3621, 3639, 3774, 4319, 4342, 4606; drag_state_handle ×7 at :2115, 3605, 3775, 3788, 3972, 4341, 4603) currently resolve to GtkBackend's inherent copies (gtk/backend.rs:560, :582) only because the field names the concrete type. Retype the field and all 19 compile unchanged.

Correction to the record: PLAN.md and GOALS.md both state "44 call sites" for this blocker. That number came from grep -n 'self\.backend\.' src/gtk/mod.rs — every use of the field, not these two methods. Today src/app.rs has 17 self.backend. lines. Do not size this work from the 44.

Scope

Retype App::backend and remove the GTK-concrete coupling that follows from it. Out of scope: any macOS code. #47's stages 2+ cannot be built or verified on any machine in this fleet (no Apple SDK / objc2 toolchain) — this issue is the buildable, testable part.

The residual GTK coupling in src/app.rs is 35 code lines (python3 scripts/native_lines.py gtk src/app.rs reports 64, but ~39 of those are doc-comments that merely mention GTK). Beyond the backend field:

Platform-typed fields — :281 settings_monitor: Option<gio::FileMonitor>, :425 window: Option<gtk4::Window>, :580 css_provider: Option<gtk4::CssProvider>. Also in the new_headless signature at :896-900.

Hook sites, with upstream status:

Hook	Site	Status
Exit	:1481, :1544 glib::idle_add_local_once(process::exit)	quadraui already has Reaction::Exit (quadraui/src/runtime.rs; GTK's ReactionSink::request_exit at gtk/run.rs:1074). app.rs uses it 0 times; TUI returns it at 7 sites. Pure adoption.
200 ms yank highlight	:1921 glib::timeout_add_local_once	No toolkit needed. TUI does it portably with yank_hl_deadline: Cell<Option<Instant>> polled in tick (src/tui_main/shell_app.rs:486, :2938).
Colorscheme / dark mode	:1934-1940, :5836-5839	Trait has set_theme (already called once by App).
Window maximize	:5521-5546	Trait has toggle_window_maximize — and App already calls it while also driving gtk4::Window::maximize directly at :5530-5536. Drop the direct call.
Window title / minimize / close	:2043-2044, :5521-5546	Small genuine gap — no set_title/minimize/close on the trait. File upstream if needed.
CSD capture	:4999-5027 (list_toplevels, set_decorated(false))	Needs a window-chrome flag in ShellConfig, not a toolkit call.
Settings-file monitor	:856-868	Needs a file watcher, not GTK (notify crate or mtime poll in tick). Parity note: the TUI has no settings hot-reload at all — zero hits.
Icon theme path / bundled icon font	:821-825	Only matters if native icon rendering remains.
Folder picker	:5592-5613 gtk4::FileDialog::select_folder	Genuine gap in PlatformServices (quadraui/src/backend.rs:2136 has file open/save only) — handled by the FolderPickerController adoption issue, not here.
Text measurement	:5971-5972 click::build_editor_click_context → set_pango_context	Genuinely toolkit-bound. Leave it; this plus folder-select are the only two real native needs in the file.

crate::gtk::{click, css, util} is less coupled than the module doc claims: of 19 helpers used, only click::build_editor_click_context (Pango), css::load_css (returns CssProvider), and util::install_bundled_icon_font/app_icon_image are toolkit-bound. pixel_to_click_target (11 uses), handle_mouse_click/double/drag, resolve_tab_right_click, make_theme_css, open_url and all 10 gtk/mod.rs helpers have no GTK types in their signatures.

Fix

1. Retype App::backend to a dyn Backend handle. Verify the 19 handle calls compile untouched.
2. Adopt Reaction::Exit; delete the two glib::idle_add_local_once exits.
3. Port the yank-highlight timer to the TUI's deadline-in-tick pattern.
4. Drop the direct gtk4::Window::maximize call in favour of the trait method already being called beside it.
5. Isolate what genuinely remains (Pango measurement context, window title/minimize/close, CSD, icon theme, settings watcher) behind a narrow trait or PlatformServices extension, so the gui gate covers only that.
6. Update the src/app.rs module doc — it currently lists blockers that no longer exist.

Do not close #47; it stays open for the macOS stages. Note in the PR which of the module doc's blockers are now retired.

Acceptance

• cargo test (GUI lane) green; cargo clippy -- -D warnings, cargo fmt pass.
• python3 scripts/native_lines.py gtk src/app.rs drops materially from 64.
• Behaviour-preserving refactor: exit, yank highlight, maximize and theme switching must be covered by existing or new GtkDriver black-box tests. If you add one, state in the PR that it fails against unfixed develop per CLAUDE.md.

[JDonaghy]

Before you start: re-verify the citations, do not trust them

Every file:line in this issue was measured on develop @ ee26268 (2026-09-04) by an audit, not by reading the tree you will check out.

They will have drifted by the time you pick this up. This issue sits in a serialised chain, and several entries ahead of it move code on purpose — #812 deletes ~1,300 lines from src/render.rs, #813 retypes fields in src/app.rs, #785 and #814 delete whole functions. Anything filed after those shifts.

So:

• Locate by symbol, not by line. grep -n "fn build_screen_layout_with_breadcrumb_row" src/render.rs, not "go to line 12886". GOALS.md already carries this rule for the same reason — it cites no line numbers on purpose.
• Re-measure any count before you use it. Line totals here came from python3 scripts/prod_lines.py and scripts/native_lines.py at a named revision. Re-run them rather than quoting the number in this issue. GOALS.md: "regenerate them with the script rather than trusting the table."
• Size the branch from what you find, not from the figure here. The counts are evidence that a problem is large, not estimates of a diff.

If a finding does not reproduce, say so — that is a useful result

This issue was generated from an automated audit. Some of its claims are wrong. Today alone, two confident findings in the same audit chain did not survive checking: a "1,671 byte-identical lines" figure measured out at ~85, and a "zero commits of pin lag" reading was taken against a stale local ref and was actually 42.

If the problem described here is already fixed, was never real, or is real but different from the description: comment with what you actually found and stop. Do not implement around a bad premise to make the issue closeable. A correction is worth more than a merged PR built on a wrong claim.