// This file is part of river, a dynamic tiling wayland compositor. // // Copyright 2020 The River Developers // // This program is free software: you can redistribute it and/or modify // it under the terms of the GNU General Public License as published by // the Free Software Foundation, either version 3 of the License, or // (at your option) any later version. // // This program is distributed in the hope that it will be useful, // but WITHOUT ANY WARRANTY; without even the implied warranty of // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the // GNU General Public License for more details. // // You should have received a copy of the GNU General Public License // along with this program. If not, see . const build_options = @import("build_options"); const std = @import("std"); const mem = std.mem; const os = std.os; const wlr = @import("wlroots"); const wl = @import("wayland").server.wl; const pixman = @import("pixman"); const server = &@import("main.zig").server; const util = @import("util.zig"); const Box = @import("Box.zig"); const LayerSurface = @import("LayerSurface.zig"); const Output = @import("Output.zig"); const Server = @import("Server.zig"); const View = @import("View.zig"); const ViewStack = @import("view_stack.zig").ViewStack; const log = std.log.scoped(.render); const SurfaceRenderData = struct { output: *const Output, /// In output layout coordinates relative to the output output_x: i32, output_y: i32, when: *os.timespec, }; /// The rendering order in this function must be kept in sync with Cursor.surfaceAt() pub fn renderOutput(output: *Output) void { var now: os.timespec = undefined; os.clock_gettime(os.CLOCK_MONOTONIC, &now) catch @panic("CLOCK_MONOTONIC not supported"); var needs_frame: bool = undefined; var damage_region: pixman.Region32 = undefined; damage_region.init(); defer damage_region.deinit(); output.damage.attachRender(&needs_frame, &damage_region) catch { log.err("failed to attach renderer", .{}); return; }; if (!needs_frame) { output.wlr_output.rollback(); return; } server.renderer.begin(@intCast(u32, output.wlr_output.width), @intCast(u32, output.wlr_output.height)); // Find the first visible fullscreen view in the stack if there is one var it = ViewStack(View).iter(output.views.first, .forward, output.current.tags, renderFilter); const fullscreen_view = while (it.next()) |view| { if (view.current.fullscreen) break view; } else null; // If we have a fullscreen view to render, render it. if (fullscreen_view) |view| { // Always clear with solid black for fullscreen server.renderer.clear(&[_]f32{ 0, 0, 0, 1 }); renderView(output, view, &now); if (build_options.xwayland) renderXwaylandUnmanaged(output, &now); } else { // No fullscreen view, so render normal layers/views server.renderer.clear(&server.config.background_color); renderLayer(output, output.getLayer(.background).*, &now, .toplevels); renderLayer(output, output.getLayer(.bottom).*, &now, .toplevels); // The first view in the list is "on top" so always iterate in reverse. // non-focused, non-floating views it = ViewStack(View).iter(output.views.last, .reverse, output.current.tags, renderFilter); while (it.next()) |view| { if (view.current.focus != 0 or view.current.float) continue; if (view.draw_borders) renderBorders(output, view, &now); renderView(output, view, &now); } // focused, non-floating views it = ViewStack(View).iter(output.views.last, .reverse, output.current.tags, renderFilter); while (it.next()) |view| { if (view.current.focus == 0 or view.current.float) continue; if (view.draw_borders) renderBorders(output, view, &now); renderView(output, view, &now); } // non-focused, floating views it = ViewStack(View).iter(output.views.last, .reverse, output.current.tags, renderFilter); while (it.next()) |view| { if (view.current.focus != 0 or !view.current.float) continue; if (view.draw_borders) renderBorders(output, view, &now); renderView(output, view, &now); } // focused, floating views it = ViewStack(View).iter(output.views.last, .reverse, output.current.tags, renderFilter); while (it.next()) |view| { if (view.current.focus == 0 or !view.current.float) continue; if (view.draw_borders) renderBorders(output, view, &now); renderView(output, view, &now); } if (build_options.xwayland) renderXwaylandUnmanaged(output, &now); renderLayer(output, output.getLayer(.top).*, &now, .toplevels); renderLayer(output, output.getLayer(.background).*, &now, .popups); renderLayer(output, output.getLayer(.bottom).*, &now, .popups); renderLayer(output, output.getLayer(.top).*, &now, .popups); } // The overlay layer is rendered in both fullscreen and normal cases renderLayer(output, output.getLayer(.overlay).*, &now, .toplevels); renderLayer(output, output.getLayer(.overlay).*, &now, .popups); renderDragIcons(output, &now); // Hardware cursors are rendered by the GPU on a separate plane, and can be // moved around without re-rendering what's beneath them - which is more // efficient. However, not all hardware supports hardware cursors. For this // reason, wlroots provides a software fallback, which we ask it to render // here. wlr_cursor handles configuring hardware vs software cursors for you, // and this function is a no-op when hardware cursors are in use. output.wlr_output.renderSoftwareCursors(null); // Conclude rendering and swap the buffers, showing the final frame // on-screen. server.renderer.end(); // TODO: handle failure output.wlr_output.commit() catch log.err("output commit failed for {s}", .{output.wlr_output.name}); } fn renderFilter(view: *View, filter_tags: u32) bool { // This check prevents a race condition when a frame is requested // between mapping of a view and the first configure being handled. if (view.current.box.width == 0 or view.current.box.height == 0) return false; return view.current.tags & filter_tags != 0; } /// Render all surfaces on the passed layer fn renderLayer( output: *const Output, layer: std.TailQueue(LayerSurface), now: *os.timespec, role: enum { toplevels, popups }, ) void { var it = layer.first; while (it) |node| : (it = node.next) { const layer_surface = &node.data; var rdata = SurfaceRenderData{ .output = output, .output_x = layer_surface.box.x, .output_y = layer_surface.box.y, .when = now, }; switch (role) { .toplevels => layer_surface.wlr_layer_surface.surface.forEachSurface( *SurfaceRenderData, renderSurfaceIterator, &rdata, ), .popups => layer_surface.wlr_layer_surface.forEachPopupSurface( *SurfaceRenderData, renderSurfaceIterator, &rdata, ), } } } /// Render all surfaces in the view's surface tree, including subsurfaces and popups fn renderView(output: *const Output, view: *View, now: *os.timespec) void { // If we have saved buffers, we are in the middle of a transaction // and need to render those buffers until the transaction is complete. if (view.saved_buffers.items.len != 0) { for (view.saved_buffers.items) |saved_buffer| { const texture = saved_buffer.client_buffer.texture orelse continue; renderTexture( output, texture, .{ .x = saved_buffer.surface_box.x + view.current.box.x - view.saved_surface_box.x, .y = saved_buffer.surface_box.y + view.current.box.y - view.saved_surface_box.y, .width = @intCast(c_int, saved_buffer.surface_box.width), .height = @intCast(c_int, saved_buffer.surface_box.height), }, &saved_buffer.source_box, saved_buffer.transform, ); } } else { // Since there are no stashed buffers, we are not in the middle of // a transaction and may simply render the most recent buffers provided // by the client. var rdata = SurfaceRenderData{ .output = output, .output_x = view.current.box.x - view.surface_box.x, .output_y = view.current.box.y - view.surface_box.y, .when = now, }; view.forEachSurface(*SurfaceRenderData, renderSurfaceIterator, &rdata); } } fn renderDragIcons(output: *const Output, now: *os.timespec) void { const output_box = server.root.output_layout.getBox(output.wlr_output).?; var it = server.root.drag_icons.first; while (it) |node| : (it = node.next) { const drag_icon = &node.data; var rdata = SurfaceRenderData{ .output = output, .output_x = @floatToInt(i32, drag_icon.seat.cursor.wlr_cursor.x) + drag_icon.wlr_drag_icon.surface.sx - output_box.x, .output_y = @floatToInt(i32, drag_icon.seat.cursor.wlr_cursor.y) + drag_icon.wlr_drag_icon.surface.sy - output_box.y, .when = now, }; drag_icon.wlr_drag_icon.surface.forEachSurface(*SurfaceRenderData, renderSurfaceIterator, &rdata); } } /// Render all xwayland unmanaged windows that appear on the output fn renderXwaylandUnmanaged(output: *const Output, now: *os.timespec) void { const output_box = server.root.output_layout.getBox(output.wlr_output).?; var it = server.root.xwayland_unmanaged_views.last; while (it) |node| : (it = node.prev) { const xwayland_surface = node.data.xwayland_surface; var rdata = SurfaceRenderData{ .output = output, .output_x = xwayland_surface.x - output_box.x, .output_y = xwayland_surface.y - output_box.y, .when = now, }; xwayland_surface.surface.?.forEachSurface(*SurfaceRenderData, renderSurfaceIterator, &rdata); } } /// This function is passed to wlroots to render each surface during iteration fn renderSurfaceIterator( surface: *wlr.Surface, surface_x: c_int, surface_y: c_int, rdata: *SurfaceRenderData, ) callconv(.C) void { const texture = surface.getTexture() orelse return; var source_box: wlr.FBox = undefined; surface.getBufferSourceBox(&source_box); renderTexture( rdata.output, texture, .{ .x = rdata.output_x + surface_x, .y = rdata.output_y + surface_y, .width = surface.current.width, .height = surface.current.height, }, &source_box, surface.current.transform, ); surface.sendFrameDone(rdata.when); } /// Render the given texture at the given box, taking the scale and transform /// of the output into account. fn renderTexture( output: *const Output, texture: *wlr.Texture, dest_box: wlr.Box, source_box: *const wlr.FBox, transform: wl.Output.Transform, ) void { var box = dest_box; // Scale the box to the output's current scaling factor scaleBox(&box, output.wlr_output.scale); // wlr_matrix_project_box is a helper which takes a box with a desired // x, y coordinates, width and height, and an output geometry, then // prepares an orthographic projection and multiplies the necessary // transforms to produce a model-view-projection matrix. var matrix: [9]f32 = undefined; const inverted = wlr.Output.transformInvert(transform); wlr.matrix.projectBox(&matrix, &box, inverted, 0.0, &output.wlr_output.transform_matrix); // This takes our matrix, the texture, and an alpha, and performs the actual // rendering on the GPU. server.renderer.renderSubtextureWithMatrix(texture, source_box, &matrix, 1.0) catch return; } fn renderBorders(output: *const Output, view: *View, now: *os.timespec) void { const config = &server.config; const color = blk: { if (view.current.urgent) break :blk &config.border_color_urgent; if (view.current.focus != 0) break :blk &config.border_color_focused; break :blk &config.border_color_unfocused; }; const border_width = config.border_width; const actual_box = if (view.saved_buffers.items.len != 0) view.saved_surface_box else view.surface_box; var border: Box = undefined; // left and right, covering the corners as well border.y = view.current.box.y - @intCast(i32, border_width); border.width = border_width; border.height = actual_box.height + border_width * 2; // left border.x = view.current.box.x - @intCast(i32, border_width); renderRect(output, border, color); // right border.x = view.current.box.x + @intCast(i32, actual_box.width); renderRect(output, border, color); // top and bottom border.x = view.current.box.x; border.width = actual_box.width; border.height = border_width; // top border.y = view.current.box.y - @intCast(i32, border_width); renderRect(output, border, color); // bottom border border.y = view.current.box.y + @intCast(i32, actual_box.height); renderRect(output, border, color); } fn renderRect(output: *const Output, box: Box, color: *const [4]f32) void { var wlr_box = box.toWlrBox(); scaleBox(&wlr_box, output.wlr_output.scale); server.renderer.renderRect(&wlr_box, color, &output.wlr_output.transform_matrix); } /// Scale a wlr_box, taking the possibility of fractional scaling into account. fn scaleBox(box: *wlr.Box, scale: f64) void { box.x = @floatToInt(c_int, @round(@intToFloat(f64, box.x) * scale)); box.y = @floatToInt(c_int, @round(@intToFloat(f64, box.y) * scale)); box.width = scaleLength(box.width, box.x, scale); box.height = scaleLength(box.height, box.x, scale); } /// Scales a width/height. /// /// This might seem overly complex, but it needs to work for fractional scaling. fn scaleLength(length: c_int, offset: c_int, scale: f64) c_int { return @floatToInt(c_int, @round(@intToFloat(f64, offset + length) * scale) - @round(@intToFloat(f64, offset) * scale)); }