3f58f12617
We use the headless backend to create a special fallback output used when no other output is connected. However this messes up the "real" headless output names users have come to expect (e.g. currently the first headless output will be named "HEADLESS-2" instead of "HEADLESS-1"). Fix this by setting the output name with [1]. [1]: https://gitlab.freedesktop.org/wlroots/wlroots/-/merge_requests/3395
1012 lines
30 KiB
C
1012 lines
30 KiB
C
#define _POSIX_C_SOURCE 200809L
|
|
#include <assert.h>
|
|
#include <stdlib.h>
|
|
#include <strings.h>
|
|
#include <time.h>
|
|
#include <wayland-server-core.h>
|
|
#include <wlr/backend/drm.h>
|
|
#include <wlr/backend/headless.h>
|
|
#include <wlr/render/wlr_renderer.h>
|
|
#include <wlr/types/wlr_buffer.h>
|
|
#include <wlr/types/wlr_drm_lease_v1.h>
|
|
#include <wlr/types/wlr_matrix.h>
|
|
#include <wlr/types/wlr_output_damage.h>
|
|
#include <wlr/types/wlr_output_layout.h>
|
|
#include <wlr/types/wlr_output.h>
|
|
#include <wlr/types/wlr_presentation_time.h>
|
|
#include <wlr/types/wlr_surface.h>
|
|
#include <wlr/util/region.h>
|
|
#include "config.h"
|
|
#include "log.h"
|
|
#include "sway/config.h"
|
|
#include "sway/desktop/transaction.h"
|
|
#include "sway/input/input-manager.h"
|
|
#include "sway/input/seat.h"
|
|
#include "sway/layers.h"
|
|
#include "sway/output.h"
|
|
#include "sway/server.h"
|
|
#include "sway/surface.h"
|
|
#include "sway/tree/arrange.h"
|
|
#include "sway/tree/container.h"
|
|
#include "sway/tree/root.h"
|
|
#include "sway/tree/view.h"
|
|
#include "sway/tree/workspace.h"
|
|
|
|
struct sway_output *output_by_name_or_id(const char *name_or_id) {
|
|
for (int i = 0; i < root->outputs->length; ++i) {
|
|
struct sway_output *output = root->outputs->items[i];
|
|
char identifier[128];
|
|
output_get_identifier(identifier, sizeof(identifier), output);
|
|
if (strcasecmp(identifier, name_or_id) == 0
|
|
|| strcasecmp(output->wlr_output->name, name_or_id) == 0) {
|
|
return output;
|
|
}
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
struct sway_output *all_output_by_name_or_id(const char *name_or_id) {
|
|
struct sway_output *output;
|
|
wl_list_for_each(output, &root->all_outputs, link) {
|
|
char identifier[128];
|
|
output_get_identifier(identifier, sizeof(identifier), output);
|
|
if (strcasecmp(identifier, name_or_id) == 0
|
|
|| strcasecmp(output->wlr_output->name, name_or_id) == 0) {
|
|
return output;
|
|
}
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
struct surface_iterator_data {
|
|
sway_surface_iterator_func_t user_iterator;
|
|
void *user_data;
|
|
|
|
struct sway_output *output;
|
|
struct sway_view *view;
|
|
double ox, oy;
|
|
int width, height;
|
|
};
|
|
|
|
static bool get_surface_box(struct surface_iterator_data *data,
|
|
struct wlr_surface *surface, int sx, int sy,
|
|
struct wlr_box *surface_box) {
|
|
struct sway_output *output = data->output;
|
|
|
|
if (!wlr_surface_has_buffer(surface)) {
|
|
return false;
|
|
}
|
|
|
|
int sw = surface->current.width;
|
|
int sh = surface->current.height;
|
|
|
|
struct wlr_box box = {
|
|
.x = floor(data->ox + sx),
|
|
.y = floor(data->oy + sy),
|
|
.width = sw,
|
|
.height = sh,
|
|
};
|
|
if (surface_box != NULL) {
|
|
memcpy(surface_box, &box, sizeof(struct wlr_box));
|
|
}
|
|
|
|
struct wlr_box output_box = {
|
|
.width = output->width,
|
|
.height = output->height,
|
|
};
|
|
|
|
struct wlr_box intersection;
|
|
return wlr_box_intersection(&intersection, &output_box, &box);
|
|
}
|
|
|
|
static void output_for_each_surface_iterator(struct wlr_surface *surface,
|
|
int sx, int sy, void *_data) {
|
|
struct surface_iterator_data *data = _data;
|
|
|
|
struct wlr_box box;
|
|
bool intersects = get_surface_box(data, surface, sx, sy, &box);
|
|
if (!intersects) {
|
|
return;
|
|
}
|
|
|
|
data->user_iterator(data->output, data->view, surface, &box,
|
|
data->user_data);
|
|
}
|
|
|
|
void output_surface_for_each_surface(struct sway_output *output,
|
|
struct wlr_surface *surface, double ox, double oy,
|
|
sway_surface_iterator_func_t iterator, void *user_data) {
|
|
struct surface_iterator_data data = {
|
|
.user_iterator = iterator,
|
|
.user_data = user_data,
|
|
.output = output,
|
|
.view = NULL,
|
|
.ox = ox,
|
|
.oy = oy,
|
|
.width = surface->current.width,
|
|
.height = surface->current.height,
|
|
};
|
|
|
|
wlr_surface_for_each_surface(surface,
|
|
output_for_each_surface_iterator, &data);
|
|
}
|
|
|
|
void output_view_for_each_surface(struct sway_output *output,
|
|
struct sway_view *view, sway_surface_iterator_func_t iterator,
|
|
void *user_data) {
|
|
struct surface_iterator_data data = {
|
|
.user_iterator = iterator,
|
|
.user_data = user_data,
|
|
.output = output,
|
|
.view = view,
|
|
.ox = view->container->surface_x - output->lx
|
|
- view->geometry.x,
|
|
.oy = view->container->surface_y - output->ly
|
|
- view->geometry.y,
|
|
.width = view->container->current.content_width,
|
|
.height = view->container->current.content_height,
|
|
};
|
|
|
|
view_for_each_surface(view, output_for_each_surface_iterator, &data);
|
|
}
|
|
|
|
void output_view_for_each_popup_surface(struct sway_output *output,
|
|
struct sway_view *view, sway_surface_iterator_func_t iterator,
|
|
void *user_data) {
|
|
struct surface_iterator_data data = {
|
|
.user_iterator = iterator,
|
|
.user_data = user_data,
|
|
.output = output,
|
|
.view = view,
|
|
.ox = view->container->surface_x - output->lx
|
|
- view->geometry.x,
|
|
.oy = view->container->surface_y - output->ly
|
|
- view->geometry.y,
|
|
.width = view->container->current.content_width,
|
|
.height = view->container->current.content_height,
|
|
};
|
|
|
|
view_for_each_popup_surface(view, output_for_each_surface_iterator, &data);
|
|
}
|
|
|
|
void output_layer_for_each_surface(struct sway_output *output,
|
|
struct wl_list *layer_surfaces, sway_surface_iterator_func_t iterator,
|
|
void *user_data) {
|
|
struct sway_layer_surface *layer_surface;
|
|
wl_list_for_each(layer_surface, layer_surfaces, link) {
|
|
struct wlr_layer_surface_v1 *wlr_layer_surface_v1 =
|
|
layer_surface->layer_surface;
|
|
struct wlr_surface *surface = wlr_layer_surface_v1->surface;
|
|
struct surface_iterator_data data = {
|
|
.user_iterator = iterator,
|
|
.user_data = user_data,
|
|
.output = output,
|
|
.view = NULL,
|
|
.ox = layer_surface->geo.x,
|
|
.oy = layer_surface->geo.y,
|
|
.width = surface->current.width,
|
|
.height = surface->current.height,
|
|
};
|
|
wlr_layer_surface_v1_for_each_surface(wlr_layer_surface_v1,
|
|
output_for_each_surface_iterator, &data);
|
|
}
|
|
}
|
|
|
|
void output_layer_for_each_toplevel_surface(struct sway_output *output,
|
|
struct wl_list *layer_surfaces, sway_surface_iterator_func_t iterator,
|
|
void *user_data) {
|
|
struct sway_layer_surface *layer_surface;
|
|
wl_list_for_each(layer_surface, layer_surfaces, link) {
|
|
struct wlr_layer_surface_v1 *wlr_layer_surface_v1 =
|
|
layer_surface->layer_surface;
|
|
output_surface_for_each_surface(output, wlr_layer_surface_v1->surface,
|
|
layer_surface->geo.x, layer_surface->geo.y, iterator,
|
|
user_data);
|
|
}
|
|
}
|
|
|
|
|
|
void output_layer_for_each_popup_surface(struct sway_output *output,
|
|
struct wl_list *layer_surfaces, sway_surface_iterator_func_t iterator,
|
|
void *user_data) {
|
|
struct sway_layer_surface *layer_surface;
|
|
wl_list_for_each(layer_surface, layer_surfaces, link) {
|
|
struct wlr_layer_surface_v1 *wlr_layer_surface_v1 =
|
|
layer_surface->layer_surface;
|
|
struct wlr_surface *surface = wlr_layer_surface_v1->surface;
|
|
struct surface_iterator_data data = {
|
|
.user_iterator = iterator,
|
|
.user_data = user_data,
|
|
.output = output,
|
|
.view = NULL,
|
|
.ox = layer_surface->geo.x,
|
|
.oy = layer_surface->geo.y,
|
|
.width = surface->current.width,
|
|
.height = surface->current.height,
|
|
};
|
|
wlr_layer_surface_v1_for_each_popup_surface(wlr_layer_surface_v1,
|
|
output_for_each_surface_iterator, &data);
|
|
}
|
|
}
|
|
|
|
#if HAVE_XWAYLAND
|
|
void output_unmanaged_for_each_surface(struct sway_output *output,
|
|
struct wl_list *unmanaged, sway_surface_iterator_func_t iterator,
|
|
void *user_data) {
|
|
struct sway_xwayland_unmanaged *unmanaged_surface;
|
|
wl_list_for_each(unmanaged_surface, unmanaged, link) {
|
|
struct wlr_xwayland_surface *xsurface =
|
|
unmanaged_surface->wlr_xwayland_surface;
|
|
double ox = unmanaged_surface->lx - output->lx;
|
|
double oy = unmanaged_surface->ly - output->ly;
|
|
|
|
output_surface_for_each_surface(output, xsurface->surface, ox, oy,
|
|
iterator, user_data);
|
|
}
|
|
}
|
|
#endif
|
|
|
|
void output_drag_icons_for_each_surface(struct sway_output *output,
|
|
struct wl_list *drag_icons, sway_surface_iterator_func_t iterator,
|
|
void *user_data) {
|
|
struct sway_drag_icon *drag_icon;
|
|
wl_list_for_each(drag_icon, drag_icons, link) {
|
|
double ox = drag_icon->x - output->lx;
|
|
double oy = drag_icon->y - output->ly;
|
|
|
|
if (drag_icon->wlr_drag_icon->mapped) {
|
|
output_surface_for_each_surface(output,
|
|
drag_icon->wlr_drag_icon->surface, ox, oy,
|
|
iterator, user_data);
|
|
}
|
|
}
|
|
}
|
|
|
|
static void for_each_surface_container_iterator(struct sway_container *con,
|
|
void *_data) {
|
|
if (!con->view || !view_is_visible(con->view)) {
|
|
return;
|
|
}
|
|
|
|
struct surface_iterator_data *data = _data;
|
|
output_view_for_each_surface(data->output, con->view,
|
|
data->user_iterator, data->user_data);
|
|
}
|
|
|
|
static void output_for_each_surface(struct sway_output *output,
|
|
sway_surface_iterator_func_t iterator, void *user_data) {
|
|
if (output_has_opaque_overlay_layer_surface(output)) {
|
|
goto overlay;
|
|
}
|
|
|
|
struct surface_iterator_data data = {
|
|
.user_iterator = iterator,
|
|
.user_data = user_data,
|
|
.output = output,
|
|
.view = NULL,
|
|
};
|
|
|
|
struct sway_workspace *workspace = output_get_active_workspace(output);
|
|
struct sway_container *fullscreen_con = root->fullscreen_global;
|
|
if (!fullscreen_con) {
|
|
if (!workspace) {
|
|
return;
|
|
}
|
|
fullscreen_con = workspace->current.fullscreen;
|
|
}
|
|
if (fullscreen_con) {
|
|
for_each_surface_container_iterator(fullscreen_con, &data);
|
|
container_for_each_child(fullscreen_con,
|
|
for_each_surface_container_iterator, &data);
|
|
|
|
// TODO: Show transient containers for fullscreen global
|
|
if (fullscreen_con == workspace->current.fullscreen) {
|
|
for (int i = 0; i < workspace->current.floating->length; ++i) {
|
|
struct sway_container *floater =
|
|
workspace->current.floating->items[i];
|
|
if (container_is_transient_for(floater, fullscreen_con)) {
|
|
for_each_surface_container_iterator(floater, &data);
|
|
}
|
|
}
|
|
}
|
|
#if HAVE_XWAYLAND
|
|
output_unmanaged_for_each_surface(output, &root->xwayland_unmanaged,
|
|
iterator, user_data);
|
|
#endif
|
|
} else {
|
|
output_layer_for_each_surface(output,
|
|
&output->layers[ZWLR_LAYER_SHELL_V1_LAYER_BACKGROUND],
|
|
iterator, user_data);
|
|
output_layer_for_each_surface(output,
|
|
&output->layers[ZWLR_LAYER_SHELL_V1_LAYER_BOTTOM],
|
|
iterator, user_data);
|
|
|
|
workspace_for_each_container(workspace,
|
|
for_each_surface_container_iterator, &data);
|
|
|
|
#if HAVE_XWAYLAND
|
|
output_unmanaged_for_each_surface(output, &root->xwayland_unmanaged,
|
|
iterator, user_data);
|
|
#endif
|
|
output_layer_for_each_surface(output,
|
|
&output->layers[ZWLR_LAYER_SHELL_V1_LAYER_TOP],
|
|
iterator, user_data);
|
|
}
|
|
|
|
overlay:
|
|
output_layer_for_each_surface(output,
|
|
&output->layers[ZWLR_LAYER_SHELL_V1_LAYER_OVERLAY],
|
|
iterator, user_data);
|
|
output_drag_icons_for_each_surface(output, &root->drag_icons,
|
|
iterator, user_data);
|
|
}
|
|
|
|
static int scale_length(int length, int offset, float scale) {
|
|
return round((offset + length) * scale) - round(offset * scale);
|
|
}
|
|
|
|
void scale_box(struct wlr_box *box, float scale) {
|
|
box->width = scale_length(box->width, box->x, scale);
|
|
box->height = scale_length(box->height, box->y, scale);
|
|
box->x = round(box->x * scale);
|
|
box->y = round(box->y * scale);
|
|
}
|
|
|
|
struct sway_workspace *output_get_active_workspace(struct sway_output *output) {
|
|
struct sway_seat *seat = input_manager_current_seat();
|
|
struct sway_node *focus = seat_get_active_tiling_child(seat, &output->node);
|
|
if (!focus) {
|
|
if (!output->workspaces->length) {
|
|
return NULL;
|
|
}
|
|
return output->workspaces->items[0];
|
|
}
|
|
return focus->sway_workspace;
|
|
}
|
|
|
|
bool output_has_opaque_overlay_layer_surface(struct sway_output *output) {
|
|
struct sway_layer_surface *sway_layer_surface;
|
|
wl_list_for_each(sway_layer_surface,
|
|
&output->layers[ZWLR_LAYER_SHELL_V1_LAYER_OVERLAY], link) {
|
|
struct wlr_surface *wlr_surface = sway_layer_surface->layer_surface->surface;
|
|
pixman_box32_t output_box = {
|
|
.x2 = output->width,
|
|
.y2 = output->height,
|
|
};
|
|
pixman_region32_t surface_opaque_box;
|
|
pixman_region32_init(&surface_opaque_box);
|
|
pixman_region32_copy(&surface_opaque_box, &wlr_surface->opaque_region);
|
|
pixman_region32_translate(&surface_opaque_box,
|
|
sway_layer_surface->geo.x, sway_layer_surface->geo.y);
|
|
pixman_region_overlap_t contains =
|
|
pixman_region32_contains_rectangle(&surface_opaque_box, &output_box);
|
|
pixman_region32_fini(&surface_opaque_box);
|
|
|
|
if (contains == PIXMAN_REGION_IN) {
|
|
return true;
|
|
}
|
|
}
|
|
return false;
|
|
}
|
|
|
|
struct send_frame_done_data {
|
|
struct timespec when;
|
|
int msec_until_refresh;
|
|
};
|
|
|
|
static void send_frame_done_iterator(struct sway_output *output,
|
|
struct sway_view *view, struct wlr_surface *surface,
|
|
struct wlr_box *box, void *user_data) {
|
|
int view_max_render_time = 0;
|
|
if (view != NULL) {
|
|
view_max_render_time = view->max_render_time;
|
|
}
|
|
|
|
struct send_frame_done_data *data = user_data;
|
|
|
|
int delay = data->msec_until_refresh - output->max_render_time
|
|
- view_max_render_time;
|
|
|
|
if (output->max_render_time == 0 || view_max_render_time == 0 || delay < 1) {
|
|
wlr_surface_send_frame_done(surface, &data->when);
|
|
} else {
|
|
struct sway_surface *sway_surface = surface->data;
|
|
wl_event_source_timer_update(sway_surface->frame_done_timer, delay);
|
|
}
|
|
}
|
|
|
|
static void send_frame_done(struct sway_output *output, struct send_frame_done_data *data) {
|
|
output_for_each_surface(output, send_frame_done_iterator, data);
|
|
}
|
|
|
|
static void count_surface_iterator(struct sway_output *output,
|
|
struct sway_view *view, struct wlr_surface *surface,
|
|
struct wlr_box *box, void *data) {
|
|
size_t *n = data;
|
|
(*n)++;
|
|
}
|
|
|
|
static bool scan_out_fullscreen_view(struct sway_output *output,
|
|
struct sway_view *view) {
|
|
struct wlr_output *wlr_output = output->wlr_output;
|
|
struct sway_workspace *workspace = output->current.active_workspace;
|
|
if (!sway_assert(workspace, "Expected an active workspace")) {
|
|
return false;
|
|
}
|
|
|
|
if (!wl_list_empty(&view->saved_buffers)) {
|
|
return false;
|
|
}
|
|
|
|
for (int i = 0; i < workspace->current.floating->length; ++i) {
|
|
struct sway_container *floater =
|
|
workspace->current.floating->items[i];
|
|
if (container_is_transient_for(floater, view->container)) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
#if HAVE_XWAYLAND
|
|
if (!wl_list_empty(&root->xwayland_unmanaged)) {
|
|
return false;
|
|
}
|
|
#endif
|
|
|
|
if (!wl_list_empty(&output->layers[ZWLR_LAYER_SHELL_V1_LAYER_OVERLAY])) {
|
|
return false;
|
|
}
|
|
if (!wl_list_empty(&root->drag_icons)) {
|
|
return false;
|
|
}
|
|
|
|
struct wlr_surface *surface = view->surface;
|
|
if (surface == NULL) {
|
|
return false;
|
|
}
|
|
size_t n_surfaces = 0;
|
|
output_view_for_each_surface(output, view,
|
|
count_surface_iterator, &n_surfaces);
|
|
if (n_surfaces != 1) {
|
|
return false;
|
|
}
|
|
|
|
if (surface->buffer == NULL) {
|
|
return false;
|
|
}
|
|
|
|
if ((float)surface->current.scale != wlr_output->scale ||
|
|
surface->current.transform != wlr_output->transform) {
|
|
return false;
|
|
}
|
|
|
|
wlr_output_attach_buffer(wlr_output, &surface->buffer->base);
|
|
if (!wlr_output_test(wlr_output)) {
|
|
return false;
|
|
}
|
|
|
|
wlr_presentation_surface_sampled_on_output(server.presentation, surface,
|
|
wlr_output);
|
|
|
|
return wlr_output_commit(wlr_output);
|
|
}
|
|
|
|
static int output_repaint_timer_handler(void *data) {
|
|
struct sway_output *output = data;
|
|
if (output->wlr_output == NULL) {
|
|
return 0;
|
|
}
|
|
|
|
output->wlr_output->frame_pending = false;
|
|
|
|
struct sway_workspace *workspace = output->current.active_workspace;
|
|
if (workspace == NULL) {
|
|
return 0;
|
|
}
|
|
|
|
struct sway_container *fullscreen_con = root->fullscreen_global;
|
|
if (!fullscreen_con) {
|
|
fullscreen_con = workspace->current.fullscreen;
|
|
}
|
|
|
|
if (fullscreen_con && fullscreen_con->view && !debug.noscanout) {
|
|
// Try to scan-out the fullscreen view
|
|
static bool last_scanned_out = false;
|
|
bool scanned_out =
|
|
scan_out_fullscreen_view(output, fullscreen_con->view);
|
|
|
|
if (scanned_out && !last_scanned_out) {
|
|
sway_log(SWAY_DEBUG, "Scanning out fullscreen view on %s",
|
|
output->wlr_output->name);
|
|
}
|
|
if (last_scanned_out && !scanned_out) {
|
|
sway_log(SWAY_DEBUG, "Stopping fullscreen view scan out on %s",
|
|
output->wlr_output->name);
|
|
output_damage_whole(output);
|
|
}
|
|
last_scanned_out = scanned_out;
|
|
|
|
if (scanned_out) {
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
bool needs_frame;
|
|
pixman_region32_t damage;
|
|
pixman_region32_init(&damage);
|
|
if (!wlr_output_damage_attach_render(output->damage,
|
|
&needs_frame, &damage)) {
|
|
return 0;
|
|
}
|
|
|
|
if (needs_frame) {
|
|
struct timespec now;
|
|
clock_gettime(CLOCK_MONOTONIC, &now);
|
|
|
|
output_render(output, &now, &damage);
|
|
} else {
|
|
wlr_output_rollback(output->wlr_output);
|
|
}
|
|
|
|
pixman_region32_fini(&damage);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static void damage_handle_frame(struct wl_listener *listener, void *user_data) {
|
|
struct sway_output *output =
|
|
wl_container_of(listener, output, damage_frame);
|
|
if (!output->enabled || !output->wlr_output->enabled) {
|
|
return;
|
|
}
|
|
|
|
// Compute predicted milliseconds until the next refresh. It's used for
|
|
// delaying both output rendering and surface frame callbacks.
|
|
int msec_until_refresh = 0;
|
|
|
|
if (output->max_render_time != 0) {
|
|
struct timespec now;
|
|
clockid_t presentation_clock
|
|
= wlr_backend_get_presentation_clock(server.backend);
|
|
clock_gettime(presentation_clock, &now);
|
|
|
|
const long NSEC_IN_SECONDS = 1000000000;
|
|
struct timespec predicted_refresh = output->last_presentation;
|
|
predicted_refresh.tv_nsec += output->refresh_nsec % NSEC_IN_SECONDS;
|
|
predicted_refresh.tv_sec += output->refresh_nsec / NSEC_IN_SECONDS;
|
|
if (predicted_refresh.tv_nsec >= NSEC_IN_SECONDS) {
|
|
predicted_refresh.tv_sec += 1;
|
|
predicted_refresh.tv_nsec -= NSEC_IN_SECONDS;
|
|
}
|
|
|
|
// If the predicted refresh time is before the current time then
|
|
// there's no point in delaying.
|
|
//
|
|
// We only check tv_sec because if the predicted refresh time is less
|
|
// than a second before the current time, then msec_until_refresh will
|
|
// end up slightly below zero, which will effectively disable the delay
|
|
// without potential disastrous negative overflows that could occur if
|
|
// tv_sec was not checked.
|
|
if (predicted_refresh.tv_sec >= now.tv_sec) {
|
|
long nsec_until_refresh
|
|
= (predicted_refresh.tv_sec - now.tv_sec) * NSEC_IN_SECONDS
|
|
+ (predicted_refresh.tv_nsec - now.tv_nsec);
|
|
|
|
// We want msec_until_refresh to be conservative, that is, floored.
|
|
// If we have 7.9 msec until refresh, we better compute the delay
|
|
// as if we had only 7 msec, so that we don't accidentally delay
|
|
// more than necessary and miss a frame.
|
|
msec_until_refresh = nsec_until_refresh / 1000000;
|
|
}
|
|
}
|
|
|
|
int delay = msec_until_refresh - output->max_render_time;
|
|
|
|
// If the delay is less than 1 millisecond (which is the least we can wait)
|
|
// then just render right away.
|
|
if (delay < 1) {
|
|
output_repaint_timer_handler(output);
|
|
} else {
|
|
output->wlr_output->frame_pending = true;
|
|
wl_event_source_timer_update(output->repaint_timer, delay);
|
|
}
|
|
|
|
// Send frame done to all visible surfaces
|
|
struct send_frame_done_data data = {0};
|
|
clock_gettime(CLOCK_MONOTONIC, &data.when);
|
|
data.msec_until_refresh = msec_until_refresh;
|
|
send_frame_done(output, &data);
|
|
}
|
|
|
|
void output_damage_whole(struct sway_output *output) {
|
|
// The output can exist with no wlr_output if it's just been disconnected
|
|
// and the transaction to evacuate it has't completed yet.
|
|
if (output && output->wlr_output && output->damage) {
|
|
wlr_output_damage_add_whole(output->damage);
|
|
}
|
|
}
|
|
|
|
static void damage_surface_iterator(struct sway_output *output,
|
|
struct sway_view *view, struct wlr_surface *surface,
|
|
struct wlr_box *_box, void *_data) {
|
|
bool *data = _data;
|
|
bool whole = *data;
|
|
|
|
struct wlr_box box = *_box;
|
|
scale_box(&box, output->wlr_output->scale);
|
|
|
|
pixman_region32_t damage;
|
|
pixman_region32_init(&damage);
|
|
wlr_surface_get_effective_damage(surface, &damage);
|
|
wlr_region_scale(&damage, &damage, output->wlr_output->scale);
|
|
if (ceil(output->wlr_output->scale) > surface->current.scale) {
|
|
// When scaling up a surface, it'll become blurry so we need to
|
|
// expand the damage region
|
|
wlr_region_expand(&damage, &damage,
|
|
ceil(output->wlr_output->scale) - surface->current.scale);
|
|
}
|
|
pixman_region32_translate(&damage, box.x, box.y);
|
|
wlr_output_damage_add(output->damage, &damage);
|
|
pixman_region32_fini(&damage);
|
|
|
|
if (whole) {
|
|
wlr_output_damage_add_box(output->damage, &box);
|
|
}
|
|
|
|
if (!wl_list_empty(&surface->current.frame_callback_list)) {
|
|
wlr_output_schedule_frame(output->wlr_output);
|
|
}
|
|
}
|
|
|
|
void output_damage_surface(struct sway_output *output, double ox, double oy,
|
|
struct wlr_surface *surface, bool whole) {
|
|
output_surface_for_each_surface(output, surface, ox, oy,
|
|
damage_surface_iterator, &whole);
|
|
}
|
|
|
|
void output_damage_from_view(struct sway_output *output,
|
|
struct sway_view *view) {
|
|
if (!view_is_visible(view)) {
|
|
return;
|
|
}
|
|
bool whole = false;
|
|
output_view_for_each_surface(output, view, damage_surface_iterator, &whole);
|
|
}
|
|
|
|
// Expecting an unscaled box in layout coordinates
|
|
void output_damage_box(struct sway_output *output, struct wlr_box *_box) {
|
|
struct wlr_box box;
|
|
memcpy(&box, _box, sizeof(struct wlr_box));
|
|
box.x -= output->lx;
|
|
box.y -= output->ly;
|
|
scale_box(&box, output->wlr_output->scale);
|
|
wlr_output_damage_add_box(output->damage, &box);
|
|
}
|
|
|
|
static void damage_child_views_iterator(struct sway_container *con,
|
|
void *data) {
|
|
if (!con->view || !view_is_visible(con->view)) {
|
|
return;
|
|
}
|
|
struct sway_output *output = data;
|
|
bool whole = true;
|
|
output_view_for_each_surface(output, con->view, damage_surface_iterator,
|
|
&whole);
|
|
}
|
|
|
|
void output_damage_whole_container(struct sway_output *output,
|
|
struct sway_container *con) {
|
|
// Pad the box by 1px, because the width is a double and might be a fraction
|
|
struct wlr_box box = {
|
|
.x = con->current.x - output->lx - 1,
|
|
.y = con->current.y - output->ly - 1,
|
|
.width = con->current.width + 2,
|
|
.height = con->current.height + 2,
|
|
};
|
|
scale_box(&box, output->wlr_output->scale);
|
|
wlr_output_damage_add_box(output->damage, &box);
|
|
// Damage subsurfaces as well, which may extend outside the box
|
|
if (con->view) {
|
|
damage_child_views_iterator(con, output);
|
|
} else {
|
|
container_for_each_child(con, damage_child_views_iterator, output);
|
|
}
|
|
}
|
|
|
|
static void damage_handle_destroy(struct wl_listener *listener, void *data) {
|
|
struct sway_output *output =
|
|
wl_container_of(listener, output, damage_destroy);
|
|
if (!output->enabled) {
|
|
return;
|
|
}
|
|
output_disable(output);
|
|
|
|
wl_list_remove(&output->damage_destroy.link);
|
|
wl_list_remove(&output->damage_frame.link);
|
|
|
|
transaction_commit_dirty();
|
|
}
|
|
|
|
static void update_output_manager_config(struct sway_server *server) {
|
|
struct wlr_output_configuration_v1 *config =
|
|
wlr_output_configuration_v1_create();
|
|
|
|
struct sway_output *output;
|
|
wl_list_for_each(output, &root->all_outputs, link) {
|
|
if (output == root->fallback_output) {
|
|
continue;
|
|
}
|
|
struct wlr_output_configuration_head_v1 *config_head =
|
|
wlr_output_configuration_head_v1_create(config, output->wlr_output);
|
|
struct wlr_box *output_box = wlr_output_layout_get_box(
|
|
root->output_layout, output->wlr_output);
|
|
// We mark the output enabled even if it is switched off by DPMS
|
|
config_head->state.enabled = output->current_mode != NULL && output->enabled;
|
|
config_head->state.mode = output->current_mode;
|
|
if (output_box) {
|
|
config_head->state.x = output_box->x;
|
|
config_head->state.y = output_box->y;
|
|
}
|
|
}
|
|
|
|
wlr_output_manager_v1_set_configuration(server->output_manager_v1, config);
|
|
}
|
|
|
|
static void handle_destroy(struct wl_listener *listener, void *data) {
|
|
struct sway_output *output = wl_container_of(listener, output, destroy);
|
|
struct sway_server *server = output->server;
|
|
wl_signal_emit(&output->events.destroy, output);
|
|
|
|
if (output->enabled) {
|
|
output_disable(output);
|
|
}
|
|
output_begin_destroy(output);
|
|
|
|
wl_list_remove(&output->destroy.link);
|
|
wl_list_remove(&output->commit.link);
|
|
wl_list_remove(&output->mode.link);
|
|
wl_list_remove(&output->present.link);
|
|
|
|
transaction_commit_dirty();
|
|
|
|
update_output_manager_config(server);
|
|
}
|
|
|
|
static void handle_mode(struct wl_listener *listener, void *data) {
|
|
struct sway_output *output = wl_container_of(listener, output, mode);
|
|
if (!output->enabled && !output->enabling) {
|
|
struct output_config *oc = find_output_config(output);
|
|
if (output->wlr_output->current_mode != NULL &&
|
|
(!oc || oc->enabled)) {
|
|
// We want to enable this output, but it didn't work last time,
|
|
// possibly because we hadn't enough CRTCs. Try again now that the
|
|
// output has a mode.
|
|
sway_log(SWAY_DEBUG, "Output %s has gained a CRTC, "
|
|
"trying to enable it", output->wlr_output->name);
|
|
apply_output_config(oc, output);
|
|
}
|
|
return;
|
|
}
|
|
if (!output->enabled) {
|
|
return;
|
|
}
|
|
arrange_layers(output);
|
|
arrange_output(output);
|
|
transaction_commit_dirty();
|
|
|
|
update_output_manager_config(output->server);
|
|
}
|
|
|
|
static void update_textures(struct sway_container *con, void *data) {
|
|
container_update_title_textures(con);
|
|
container_update_marks_textures(con);
|
|
}
|
|
|
|
static void handle_commit(struct wl_listener *listener, void *data) {
|
|
struct sway_output *output = wl_container_of(listener, output, commit);
|
|
struct wlr_output_event_commit *event = data;
|
|
|
|
if (!output->enabled) {
|
|
return;
|
|
}
|
|
|
|
if (event->committed & WLR_OUTPUT_STATE_SCALE) {
|
|
output_for_each_container(output, update_textures, NULL);
|
|
}
|
|
|
|
if (event->committed & (WLR_OUTPUT_STATE_TRANSFORM | WLR_OUTPUT_STATE_SCALE)) {
|
|
arrange_layers(output);
|
|
arrange_output(output);
|
|
transaction_commit_dirty();
|
|
|
|
update_output_manager_config(output->server);
|
|
}
|
|
}
|
|
|
|
static void handle_present(struct wl_listener *listener, void *data) {
|
|
struct sway_output *output = wl_container_of(listener, output, present);
|
|
struct wlr_output_event_present *output_event = data;
|
|
|
|
if (!output->enabled || !output_event->presented) {
|
|
return;
|
|
}
|
|
|
|
output->last_presentation = *output_event->when;
|
|
output->refresh_nsec = output_event->refresh;
|
|
}
|
|
|
|
static unsigned int last_headless_num = 0;
|
|
|
|
void handle_new_output(struct wl_listener *listener, void *data) {
|
|
struct sway_server *server = wl_container_of(listener, server, new_output);
|
|
struct wlr_output *wlr_output = data;
|
|
|
|
if (wlr_output == root->fallback_output->wlr_output) {
|
|
return;
|
|
}
|
|
|
|
if (wlr_output_is_headless(wlr_output)) {
|
|
char name[64];
|
|
snprintf(name, sizeof(name), "HEADLESS-%u", ++last_headless_num);
|
|
wlr_output_set_name(wlr_output, name);
|
|
}
|
|
|
|
sway_log(SWAY_DEBUG, "New output %p: %s (non-desktop: %d)",
|
|
wlr_output, wlr_output->name, wlr_output->non_desktop);
|
|
|
|
if (wlr_output->non_desktop) {
|
|
sway_log(SWAY_DEBUG, "Not configuring non-desktop output");
|
|
if (server->drm_lease_manager) {
|
|
wlr_drm_lease_v1_manager_offer_output(server->drm_lease_manager,
|
|
wlr_output);
|
|
}
|
|
return;
|
|
}
|
|
|
|
if (!wlr_output_init_render(wlr_output, server->allocator,
|
|
server->renderer)) {
|
|
sway_log(SWAY_ERROR, "Failed to init output render");
|
|
return;
|
|
}
|
|
|
|
struct sway_output *output = output_create(wlr_output);
|
|
if (!output) {
|
|
return;
|
|
}
|
|
output->server = server;
|
|
output->damage = wlr_output_damage_create(wlr_output);
|
|
|
|
wl_signal_add(&wlr_output->events.destroy, &output->destroy);
|
|
output->destroy.notify = handle_destroy;
|
|
wl_signal_add(&wlr_output->events.commit, &output->commit);
|
|
output->commit.notify = handle_commit;
|
|
wl_signal_add(&wlr_output->events.mode, &output->mode);
|
|
output->mode.notify = handle_mode;
|
|
wl_signal_add(&wlr_output->events.present, &output->present);
|
|
output->present.notify = handle_present;
|
|
wl_signal_add(&output->damage->events.frame, &output->damage_frame);
|
|
output->damage_frame.notify = damage_handle_frame;
|
|
wl_signal_add(&output->damage->events.destroy, &output->damage_destroy);
|
|
output->damage_destroy.notify = damage_handle_destroy;
|
|
|
|
output->repaint_timer = wl_event_loop_add_timer(server->wl_event_loop,
|
|
output_repaint_timer_handler, output);
|
|
|
|
struct output_config *oc = find_output_config(output);
|
|
apply_output_config(oc, output);
|
|
free_output_config(oc);
|
|
|
|
transaction_commit_dirty();
|
|
|
|
update_output_manager_config(server);
|
|
}
|
|
|
|
void handle_output_layout_change(struct wl_listener *listener,
|
|
void *data) {
|
|
struct sway_server *server =
|
|
wl_container_of(listener, server, output_layout_change);
|
|
update_output_manager_config(server);
|
|
}
|
|
|
|
static void output_manager_apply(struct sway_server *server,
|
|
struct wlr_output_configuration_v1 *config, bool test_only) {
|
|
// TODO: perform atomic tests on the whole backend atomically
|
|
|
|
struct wlr_output_configuration_head_v1 *config_head;
|
|
// First disable outputs we need to disable
|
|
bool ok = true;
|
|
wl_list_for_each(config_head, &config->heads, link) {
|
|
struct wlr_output *wlr_output = config_head->state.output;
|
|
struct sway_output *output = wlr_output->data;
|
|
if (!output->enabled || config_head->state.enabled) {
|
|
continue;
|
|
}
|
|
struct output_config *oc = new_output_config(output->wlr_output->name);
|
|
oc->enabled = false;
|
|
|
|
if (test_only) {
|
|
ok &= test_output_config(oc, output);
|
|
} else {
|
|
oc = store_output_config(oc);
|
|
ok &= apply_output_config(oc, output);
|
|
}
|
|
}
|
|
|
|
// Then enable outputs that need to
|
|
wl_list_for_each(config_head, &config->heads, link) {
|
|
struct wlr_output *wlr_output = config_head->state.output;
|
|
struct sway_output *output = wlr_output->data;
|
|
if (!config_head->state.enabled) {
|
|
continue;
|
|
}
|
|
struct output_config *oc = new_output_config(output->wlr_output->name);
|
|
oc->enabled = true;
|
|
if (config_head->state.mode != NULL) {
|
|
struct wlr_output_mode *mode = config_head->state.mode;
|
|
oc->width = mode->width;
|
|
oc->height = mode->height;
|
|
oc->refresh_rate = mode->refresh / 1000.f;
|
|
} else {
|
|
oc->width = config_head->state.custom_mode.width;
|
|
oc->height = config_head->state.custom_mode.height;
|
|
oc->refresh_rate =
|
|
config_head->state.custom_mode.refresh / 1000.f;
|
|
}
|
|
oc->x = config_head->state.x;
|
|
oc->y = config_head->state.y;
|
|
oc->transform = config_head->state.transform;
|
|
oc->scale = config_head->state.scale;
|
|
|
|
if (test_only) {
|
|
ok &= test_output_config(oc, output);
|
|
} else {
|
|
oc = store_output_config(oc);
|
|
ok &= apply_output_config(oc, output);
|
|
}
|
|
}
|
|
|
|
if (ok) {
|
|
wlr_output_configuration_v1_send_succeeded(config);
|
|
} else {
|
|
wlr_output_configuration_v1_send_failed(config);
|
|
}
|
|
wlr_output_configuration_v1_destroy(config);
|
|
|
|
if (!test_only) {
|
|
update_output_manager_config(server);
|
|
}
|
|
}
|
|
|
|
void handle_output_manager_apply(struct wl_listener *listener, void *data) {
|
|
struct sway_server *server =
|
|
wl_container_of(listener, server, output_manager_apply);
|
|
struct wlr_output_configuration_v1 *config = data;
|
|
|
|
output_manager_apply(server, config, false);
|
|
}
|
|
|
|
void handle_output_manager_test(struct wl_listener *listener, void *data) {
|
|
struct sway_server *server =
|
|
wl_container_of(listener, server, output_manager_test);
|
|
struct wlr_output_configuration_v1 *config = data;
|
|
|
|
output_manager_apply(server, config, true);
|
|
}
|
|
|
|
void handle_output_power_manager_set_mode(struct wl_listener *listener,
|
|
void *data) {
|
|
struct wlr_output_power_v1_set_mode_event *event = data;
|
|
struct sway_output *output = event->output->data;
|
|
|
|
struct output_config *oc = new_output_config(output->wlr_output->name);
|
|
switch (event->mode) {
|
|
case ZWLR_OUTPUT_POWER_V1_MODE_OFF:
|
|
oc->dpms_state = DPMS_OFF;
|
|
break;
|
|
case ZWLR_OUTPUT_POWER_V1_MODE_ON:
|
|
oc->dpms_state = DPMS_ON;
|
|
break;
|
|
}
|
|
oc = store_output_config(oc);
|
|
apply_output_config(oc, output);
|
|
}
|