diff --git a/backend/drm/backend.c b/backend/drm/backend.c
index 2b54336f7..3782817ce 100644
--- a/backend/drm/backend.c
+++ b/backend/drm/backend.c
@@ -123,7 +123,7 @@ struct wlr_backend *wlr_drm_backend_create(struct wl_display *display,
 
 	char *name = drmGetDeviceNameFromFd2(gpu_fd);
 	drmVersion *version = drmGetVersion(gpu_fd);
-	wlr_log(L_INFO, "Initalizing DRM backend for %s (%s)", name, version->name);
+	wlr_log(L_INFO, "Initializing DRM backend for %s (%s)", name, version->name);
 	free(name);
 	drmFreeVersion(version);
 
diff --git a/backend/drm/drm.c b/backend/drm/drm.c
index 8d102d8af..7f51fe67d 100644
--- a/backend/drm/drm.c
+++ b/backend/drm/drm.c
@@ -218,6 +218,7 @@ static bool wlr_drm_connector_swap_buffers(struct wlr_output *output) {
 	}
 
 	conn->pageflip_pending = true;
+	wlr_output_update_enabled(output, true);
 	return true;
 }
 
@@ -253,6 +254,7 @@ void wlr_drm_connector_start_renderer(struct wlr_drm_connector *conn) {
 	struct wlr_drm_mode *mode = (struct wlr_drm_mode *)conn->output.current_mode;
 	if (drm->iface->crtc_pageflip(drm, conn, crtc, fb_id, &mode->drm_mode)) {
 		conn->pageflip_pending = true;
+		wlr_output_update_enabled(&conn->output, true);
 	} else {
 		wl_event_source_timer_update(conn->retry_pageflip,
 			1000000.0f / conn->output.current_mode->refresh);
@@ -477,7 +479,7 @@ static bool wlr_drm_connector_set_mode(struct wlr_output *output,
 	wlr_output_update_mode(&conn->output, mode);
 
 	// Since realloc_crtcs can deallocate planes on OTHER outputs,
-	// we actually need to reinitalise any than has changed
+	// we actually need to reinitialize any than has changed
 	ssize_t output_index = -1;
 	wl_list_for_each(conn, &drm->outputs, link) {
 		output_index += 1;
@@ -491,7 +493,7 @@ static bool wlr_drm_connector_set_mode(struct wlr_output *output,
 
 		if (!wlr_drm_plane_surfaces_init(crtc->primary, drm,
 				mode->width, mode->height, GBM_FORMAT_XRGB8888)) {
-			wlr_log(L_ERROR, "Failed to initalise renderer for plane");
+			wlr_log(L_ERROR, "Failed to initialize renderer for plane");
 			goto error_conn;
 		}
 
@@ -975,7 +977,7 @@ void wlr_drm_connector_cleanup(struct wlr_drm_connector *conn) {
 		conn->possible_crtc = 0;
 		/* Fallthrough */
 	case WLR_DRM_CONN_NEEDS_MODESET:
-		wlr_log(L_INFO, "Emmiting destruction signal for '%s'",
+		wlr_log(L_INFO, "Emitting destruction signal for '%s'",
 				conn->output.name);
 		wl_signal_emit(&drm->backend.events.output_remove, &conn->output);
 		break;
diff --git a/backend/drm/renderer.c b/backend/drm/renderer.c
index e28910573..0310a96a2 100644
--- a/backend/drm/renderer.c
+++ b/backend/drm/renderer.c
@@ -66,6 +66,21 @@ bool wlr_drm_surface_init(struct wlr_drm_surface *surf,
 	surf->width = width;
 	surf->height = height;
 
+	if (surf->gbm) {
+		if (surf->front) {
+			gbm_surface_release_buffer(surf->gbm, surf->front);
+			surf->front = NULL;
+		}
+		if (surf->back) {
+			gbm_surface_release_buffer(surf->gbm, surf->back);
+			surf->back = NULL;
+		}
+		gbm_surface_destroy(surf->gbm);
+	}
+	if (surf->egl) {
+		eglDestroySurface(surf->renderer->egl.display, surf->egl);
+	}
+
 	surf->gbm = gbm_surface_create(renderer->gbm, width, height,
 		format, GBM_BO_USE_RENDERING | flags);
 	if (!surf->gbm) {
diff --git a/examples/screenshot.c b/examples/screenshot.c
index ba2250233..7de2ab8ef 100644
--- a/examples/screenshot.c
+++ b/examples/screenshot.c
@@ -48,6 +48,7 @@ struct screenshooter_output {
 	struct wl_output *output;
 	struct wl_buffer *buffer;
 	int width, height, offset_x, offset_y;
+	enum wl_output_transform transform;
 	void *data;
 	struct wl_list link;
 };
@@ -60,6 +61,7 @@ static void output_handle_geometry(void *data, struct wl_output *wl_output,
 	if (wl_output == output->output) {
 		output->offset_x = x;
 		output->offset_y = y;
+		output->transform = transform;
 	}
 }
 
@@ -159,14 +161,75 @@ static void write_image(const char *filename, int width, int height) {
 	struct screenshooter_output *output, *next;
 	wl_list_for_each_safe(output, next, &output_list, link) {
 		int output_stride = output->width * 4;
-		void *s = output->data;
-		void *d = data + (output->offset_y - min_y) * buffer_stride +
-			(output->offset_x - min_x) * 4;
+		uint32_t *src = (uint32_t *)output->data;
+		uint32_t *dst = (uint32_t *)(data +
+			(output->offset_y - min_y) * buffer_stride +
+			(output->offset_x - min_x) * 4);
 
-		for (int i = 0; i < output->height; i++) {
-			memcpy(d, s, output_stride);
-			d += buffer_stride;
-			s += output_stride;
+		switch (output->transform) {
+		case WL_OUTPUT_TRANSFORM_NORMAL:
+			for (int i = 0; i < output->height; i++) {
+				memcpy(dst, src, output_stride);
+				dst += width;
+				src += output->width;
+			}
+			break;
+		case WL_OUTPUT_TRANSFORM_FLIPPED:
+			for (int i = 0; i < output->height; ++i) {
+				for (int j = 0; j < output->width; ++j) {
+					dst[i * width + j] =
+						src[i * output->width + output->width - 1 - j];
+				}
+			}
+			break;
+		case WL_OUTPUT_TRANSFORM_90:
+			for (int i = 0; i < output->width; ++i) {
+				for (int j = 0; j < output->height; ++j) {
+					dst[i * width + j] =
+						src[j * output->width + output->width - 1 - i];
+				}
+			}
+			break;
+		case WL_OUTPUT_TRANSFORM_FLIPPED_90:
+			for (int i = 0; i < output->width; ++i) {
+				for (int j = 0; j < output->height; ++j) {
+					dst[i * width + j] =
+						src[(output->height - 1 - j) * output->width + output->width - 1 - i];
+				}
+			}
+			break;
+		case WL_OUTPUT_TRANSFORM_180:
+			for (int i = 0; i < output->height; ++i) {
+				for (int j = 0; j < output->width; ++j) {
+					dst[i * width + j] =
+						src[(output->height - 1 - i) * output->width + output->width - 1 - j];
+				}
+			}
+			break;
+		case WL_OUTPUT_TRANSFORM_FLIPPED_180:
+			for (int i = 0; i < output->height; ++i) {
+				for (int j = 0; j < output->width; ++j) {
+					dst[i * width + j] =
+						src[(output->height - 1 - i) * output->width + j];
+				}
+			}
+			break;
+		case WL_OUTPUT_TRANSFORM_270:
+			for (int i = 0; i < output->width; ++i) {
+				for (int j = 0; j < output->height; ++j) {
+					dst[i * width + j] =
+						src[(output->height - 1 - j) * output->width + i];
+				}
+			}
+			break;
+		case WL_OUTPUT_TRANSFORM_FLIPPED_270:
+			for (int i = 0; i < output->width; ++i) {
+				for (int j = 0; j < output->height; ++j) {
+					dst[i * width + j] =
+						src[j * output->width + i];
+				}
+			}
+			break;
 		}
 
 		free(output);
@@ -211,15 +274,23 @@ static void write_image(const char *filename, int width, int height) {
 }
 
 static int set_buffer_size(int *width, int *height) {
+	int owidth, oheight;
 	min_x = min_y = INT_MAX;
 	max_x = max_y = INT_MIN;
 
 	struct screenshooter_output *output;
 	wl_list_for_each(output, &output_list, link) {
+		if (output->transform & 0x1) {
+			owidth = output->height;
+			oheight = output->width;
+		} else {
+			owidth = output->width;
+			oheight = output->height;
+		}
 		min_x = MIN(min_x, output->offset_x);
 		min_y = MIN(min_y, output->offset_y);
-		max_x = MAX(max_x, output->offset_x + output->width);
-		max_y = MAX(max_y, output->offset_y + output->height);
+		max_x = MAX(max_x, output->offset_x + owidth);
+		max_y = MAX(max_y, output->offset_y + oheight);
 	}
 
 	if (max_x <= min_x || max_y <= min_y) {
diff --git a/include/wlr/types/wlr_output.h b/include/wlr/types/wlr_output.h
index f10d82863..a1058a8cc 100644
--- a/include/wlr/types/wlr_output.h
+++ b/include/wlr/types/wlr_output.h
@@ -31,6 +31,10 @@ struct wlr_output_cursor {
 	struct wlr_surface *surface;
 	struct wl_listener surface_commit;
 	struct wl_listener surface_destroy;
+
+	struct {
+		struct wl_signal destroy;
+	} events;
 };
 
 struct wlr_output_impl;
diff --git a/include/wlr/types/wlr_xcursor_manager.h b/include/wlr/types/wlr_xcursor_manager.h
index 2a2c9035d..c7fa83be7 100644
--- a/include/wlr/types/wlr_xcursor_manager.h
+++ b/include/wlr/types/wlr_xcursor_manager.h
@@ -6,7 +6,7 @@
 #include <wlr/xcursor.h>
 
 /**
- * A scaled XCursor theme.
+ * An XCursor theme at a particular scale factor of the base size.
  */
 struct wlr_xcursor_manager_theme {
 	float scale;
@@ -15,11 +15,10 @@ struct wlr_xcursor_manager_theme {
 };
 
 /**
- * Manage multiple XCursor themes with different scales and set `wlr_cursor`
- * images.
- *
- * This manager can be used to display cursor images on multiple outputs having
- * different scale factors.
+ * wlr_xcursor_manager dynamically loads xcursor themes at sizes necessary for
+ * use on outputs at arbitrary scale factors. You should call
+ * wlr_xcursor_manager_load for each output you will show your cursor on, with
+ * the scale factor parameter set to that output's scale factor.
  */
 struct wlr_xcursor_manager {
 	char *name;
@@ -28,24 +27,33 @@ struct wlr_xcursor_manager {
 };
 
 /**
- * Create a new XCursor manager. After initialization, scaled themes need to be
- * loaded with `wlr_xcursor_manager_load`. `size` is the unscaled cursor theme
- * size.
+ * Creates a new XCursor manager with the given xcursor theme name and base size
+ * (for use when scale=1).
  */
 struct wlr_xcursor_manager *wlr_xcursor_manager_create(const char *name,
 	uint32_t size);
 
 void wlr_xcursor_manager_destroy(struct wlr_xcursor_manager *manager);
 
+/**
+ * Ensures an xcursor theme at the given scale factor is loaded in the manager.
+ */
 int wlr_xcursor_manager_load(struct wlr_xcursor_manager *manager,
 	float scale);
 
+/**
+ * Retrieves a wlr_xcursor reference for the given cursor name at the given
+ * scale factor, or NULL if this wlr_xcursor_manager has not loaded a cursor
+ * theme at the requested scale.
+ */
 struct wlr_xcursor *wlr_xcursor_manager_get_xcursor(
 	struct wlr_xcursor_manager *manager, const char *name, float scale);
 
 /**
- * Set a `wlr_cursor` image. The manager uses all currently loaded scaled
- * themes.
+ * Set a wlr_cursor's cursor image to the specified cursor name for all scale
+ * factors. wlr_cursor will take over from this point and ensure the correct
+ * cursor is used on each output, assuming a wlr_output_layout is attached to
+ * it.
  */
 void wlr_xcursor_manager_set_cursor_image(struct wlr_xcursor_manager *manager,
 	const char *name, struct wlr_cursor *cursor);
diff --git a/include/wlr/xcursor.h b/include/wlr/xcursor.h
index 42fcedb98..d97469f8a 100644
--- a/include/wlr/xcursor.h
+++ b/include/wlr/xcursor.h
@@ -50,6 +50,9 @@ struct wlr_xcursor {
 	uint32_t total_delay; /* length of the animation in ms */
 };
 
+/**
+ * Container for an Xcursor theme.
+ */
 struct wlr_xcursor_theme {
 	unsigned int cursor_count;
 	struct wlr_xcursor **cursors;
@@ -57,13 +60,27 @@ struct wlr_xcursor_theme {
 	int size;
 };
 
+/**
+ * Loads the named xcursor theme at the given cursor size (in pixels). This is
+ * useful if you need cursor images for your compositor to use when a
+ * client-side cursors is not available or you wish to override client-side
+ * cursors for a particular UI interaction (such as using a grab cursor when
+ * moving a window around).
+ */
 struct wlr_xcursor_theme *wlr_xcursor_theme_load(const char *name, int size);
 
 void wlr_xcursor_theme_destroy(struct wlr_xcursor_theme *theme);
 
+/**
+ * Obtains a wlr_xcursor image for the specified cursor name (e.g. "left_ptr").
+ */
 struct wlr_xcursor *wlr_xcursor_theme_get_cursor(
 	struct wlr_xcursor_theme *theme, const char *name);
 
+/**
+ * Returns the current frame number for an animated cursor give a monotonic time
+ * reference.
+ */
 int wlr_xcursor_frame(struct wlr_xcursor *cursor, uint32_t time);
 
 /**
diff --git a/meson.build b/meson.build
index 55b07efaa..236abd6b5 100644
--- a/meson.build
+++ b/meson.build
@@ -36,7 +36,7 @@ wlr_inc = include_directories('include')
 
 cc = meson.get_compiler('c')
 
-# Clang complains about some zeroed initialiser lists (= {0}), even though they
+# Clang complains about some zeroed initializer lists (= {0}), even though they
 # are valid
 if cc.get_id() == 'clang'
 	add_project_arguments('-Wno-missing-field-initializers', language: 'c')
diff --git a/types/wlr_output.c b/types/wlr_output.c
index 4aed1ba6c..aaee78a26 100644
--- a/types/wlr_output.c
+++ b/types/wlr_output.c
@@ -873,6 +873,7 @@ struct wlr_output_cursor *wlr_output_cursor_create(struct wlr_output *output) {
 		return NULL;
 	}
 	cursor->output = output;
+	wl_signal_init(&cursor->events.destroy);
 	wl_list_init(&cursor->surface_commit.link);
 	cursor->surface_commit.notify = output_cursor_handle_commit;
 	wl_list_init(&cursor->surface_destroy.link);
@@ -886,6 +887,7 @@ void wlr_output_cursor_destroy(struct wlr_output_cursor *cursor) {
 		return;
 	}
 	output_cursor_reset(cursor);
+	wl_signal_emit(&cursor->events.destroy, cursor);
 	if (cursor->output->hardware_cursor == cursor) {
 		// If this cursor was the hardware cursor, disable it
 		if (cursor->output->impl->set_cursor) {