Cthulhu is no longer a fork. Now it is a screen reader written in Golang.

tests/ocr_session_fixture.py

@@ -0,0 +1,125 @@
+"""Painted OCR text with accessible focus targets on a private Xorg display."""
+import json
+import os
+from pathlib import Path
+import sys
+import cairo
+
+import gi
+gi.require_version("Gtk", "3.0")
+gi.require_version("Gdk", "3.0")
+gi.require_version("GdkX11", "3.0")
+gi.require_version("GdkPixbuf", "2.0")
+from gi.repository import Gdk, GdkPixbuf, GdkX11, GLib, Gtk
+from Xlib import X, Xatom, display
+
+
+def main() -> None:
+    window = Gtk.Window(title="OCR session fixture")
+    window.set_decorated(False)
+    window.move(20, 20)
+    fixed = Gtk.Fixed()
+    window.add(fixed)
+    painted = Gtk.DrawingArea()
+    painted.set_size_request(640, 220)
+    painted.set_can_focus(True)
+    painted.get_accessible().set_name("Painted OCR target")
+    imagePath = Path(__file__).resolve().parents[1] / "internal/ocr/testdata/painted-text.png"
+    pixels = GdkPixbuf.Pixbuf.new_from_file(str(imagePath))
+
+    def draw(_widget: Gtk.Widget, context: object) -> bool:
+        Gdk.cairo_set_source_pixbuf(context, pixels, 0, 0)
+        context.paint()
+        return False
+
+    painted.connect("draw", draw)
+    fixed.put(painted, 0, 0)
+    password = Gtk.Entry()
+    password.set_visibility(False)
+    password.get_accessible().set_name("Protected OCR focus")
+    fixed.put(password, 0, 230)
+    other = Gtk.Button(label="Other OCR focus")
+    fixed.put(other, 280, 230)
+    window.show_all()
+    painted.grab_focus()
+    clipboard = Gtk.Clipboard.get(Gdk.SELECTION_CLIPBOARD)
+    clipboard.set_image(pixels)
+    connection = display.Display()
+    root = connection.screen().root
+    clientId = GdkX11.X11Window.get_xid(window.get_window())
+    client = connection.create_resource_object("window", clientId)
+    root.change_property(connection.intern_atom("_NET_ACTIVE_WINDOW"), Xatom.WINDOW, 32, [clientId])
+    client.change_property(connection.intern_atom("_NET_WM_PID"), Xatom.CARDINAL, 32, [os.getpid()])
+    client.set_input_focus(X.RevertToParent, X.CurrentTime)
+    connection.sync()
+    activeClient = client
+    rawClient = None
+
+    def show_inaccessible_window() -> None:
+        nonlocal activeClient, rawClient
+        # A raw X11 client deliberately has no AT-SPI object or focused widget.
+        surface = cairo.ImageSurface.create_from_png(str(imagePath))
+        width, height = surface.get_width(), surface.get_height()
+        rawClient = root.create_window(20, 20, width, height, 0,
+                                      connection.screen().root_depth,
+                                      X.InputOutput, X.CopyFromParent,
+                                      background_pixel=0xFFFFFF)
+        rawClient.set_wm_name("Inaccessible OCR fixture")
+        rawClient.set_wm_class("example", "example")
+        rawClient.change_property(connection.intern_atom("_NET_WM_PID"),
+                                  Xatom.CARDINAL, 32, [os.getpid()])
+        rawClient.map()
+        rawClient.configure(stack_mode=X.Above)
+        graphics = rawClient.create_gc()
+        data = bytes(surface.get_data())
+        stride = surface.get_stride()
+        for row in range(0, height, 16):
+            rows = min(16, height - row)
+            rawClient.put_image(graphics, 0, row, width, rows, X.ZPixmap,
+                                connection.screen().root_depth, 0,
+                                data[row * stride:(row + rows) * stride])
+        graphics.free()
+        root.change_property(connection.intern_atom("_NET_ACTIVE_WINDOW"),
+                             Xatom.WINDOW, 32, [rawClient.id])
+        rawClient.set_input_focus(X.RevertToParent, X.CurrentTime)
+        activeClient = rawClient
+
+    def state() -> bool:
+        connection.sync()
+        print(json.dumps({"window": activeClient.id, "focus": connection.get_input_focus().focus.id}), flush=True)
+        return False
+
+    def command_ready(_source: object, condition: int) -> bool:
+        if condition & GLib.IO_HUP:
+            Gtk.main_quit()
+            return False
+        line = sys.stdin.readline()
+        if not line:
+            Gtk.main_quit()
+            return False
+        action = json.loads(line)["action"]
+        if action == "painted":
+            painted.grab_focus()
+        elif action == "password":
+            password.grab_focus()
+        elif action == "other":
+            other.grab_focus()
+        elif action == "clipboard-text":
+            clipboard.set_text("Generic replacement text", -1)
+        elif action == "inaccessible":
+            show_inaccessible_window()
+        elif action != "state":
+            raise ValueError("Unknown fixture action")
+        GLib.idle_add(state)
+        return True
+
+    GLib.idle_add(state)
+    GLib.io_add_watch(sys.stdin, GLib.IO_IN | GLib.IO_HUP, command_ready)
+    Gtk.main()
+    if rawClient is not None:
+        rawClient.destroy()
+    connection.close()
+
+
+if __name__ == "__main__":
+    main()