cthulhu/internal/atspi/ocr_native_test.go

// Copyright (c) 2026 Storm Dragon
// Authors: Storm Dragon, Angus Kola
// SPDX-License-Identifier: GPL-3.0-or-later

package atspi

import (
	"bufio"
	"context"
	"encoding/json"
	"image"
	"image/color"
	"os"
	"os/exec"
	"testing"
	"time"

	"git.stormux.org/storm/cthulhu/internal/ocr"
)

func TestNativeOCRProviderGeometry(t *testing.T) {
	if os.Getenv("CTHULHU_NATIVE_OCR_GEOMETRY") != "1" {
		t.Skip("requires private Xorg/AT-SPI session")
	}
	executable := os.Getenv("CTHULHU_OCR_CAPTURE_BINARY")
	if executable == "" {
		t.Fatal("set freshly built CTHULHU_OCR_CAPTURE_BINARY")
	}
	ctx, cancel := context.WithTimeout(context.Background(), 40*time.Second)
	defer cancel()
	adapter := New()
	if err := adapter.Start(ctx); err != nil {
		t.Fatal(err)
	}
	defer adapter.Stop(context.Background())
	for _, scale := range []string{"1", "2"} {
		t.Run("scale"+scale, func(t *testing.T) {
			ctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
			defer cancel()
			fixture := exec.CommandContext(ctx, "/usr/bin/python3", "-u", "../../tests/ocr_geometry_fixture.py")
			fixture.Env = append(os.Environ(), "GDK_SCALE="+scale, "NO_AT_BRIDGE=0", "GTK_MODULES=gail:atk-bridge")
			stdin, err := fixture.StdinPipe()
			if err != nil {
				t.Fatal(err)
			}
			stdout, err := fixture.StdoutPipe()
			if err != nil {
				t.Fatal(err)
			}
			fixture.Stderr = os.Stderr
			if err = fixture.Start(); err != nil {
				t.Fatal(err)
			}
			defer func() { stdin.Close(); fixture.Wait() }()
			scanner := bufio.NewScanner(stdout)
			type state struct {
				Window uint64
				Target [4]int
				Focus  uint64
			}
			read := func() state {
				t.Helper()
				if !scanner.Scan() {
					t.Fatal("fixture did not respond")
				}
				var s state
				if err = json.Unmarshal(scanner.Bytes(), &s); err != nil {
					t.Fatal(err)
				}
				return s
			}
			initial := read()
			action := func(name string) state {
				t.Helper()
				data, _ := json.Marshal(map[string]string{"action": name})
				if _, err = stdin.Write(append(data, '\n')); err != nil {
					t.Fatal(err)
				}
				return read()
			}
			var tree *AccessibleTree
			var ref ObjectReference
			for !ref.Valid() {
				tree, err = adapter.CaptureDesktop(ctx)
				if err != nil {
					t.Fatal(err)
				}
				ref = findNamedReference(t, tree, "Painted target")
				if !ref.Valid() {
					select {
					case <-ctx.Done():
						t.Fatal(ctx.Err())
					case <-time.After(10 * time.Millisecond):
					}
				}
			}
			geometry, err := adapter.CaptureOCRGeometry(ctx, ref.ID(), nil)
			expected := image.Rect(initial.Target[0], initial.Target[1], initial.Target[2], initial.Target[3])
			if err != nil || geometry.ScreenBounds != expected || geometry.Foreground.Client.Window != initial.Window {
				provider, _ := tree.Geometry(ref)
				t.Fatalf("geometry=%+v err=%v want=%v provider=%+v foreground=%+v", geometry, err, expected, provider, adapter.Foreground())
			}
			result, err := ocr.NewCapturer(executable).Capture(ctx, ocr.CaptureRequest{Kind: ocr.ObjectSource, Bounds: geometry.ScreenBounds, Foreground: initial.Window, Generation: 1})
			if err != nil {
				t.Fatal(err)
			}
			for _, point := range []image.Point{image.Pt(0, 0), image.Pt(result.Pixels.Bounds().Dx()-1, result.Pixels.Bounds().Dy()-1)} {
				if got := color.RGBAModel.Convert(result.Pixels.At(point.X, point.Y)); got != (color.RGBA{51, 102, 153, 255}) {
					t.Fatalf("mapped pixel %v = %v", point, got)
				}
			}
			zone := geometry.ProviderBounds
			zone.Width /= 2
			zone.Height /= 2
			part, err := adapter.CaptureOCRGeometry(ctx, ref.ID(), &zone)
			if err != nil || part.ScreenBounds.Dx() != expected.Dx()/2 || part.ScreenBounds.Dy() != expected.Dy()/2 {
				t.Fatalf("zone: %+v %v", part, err)
			}
			invalid := zone
			invalid.X -= 10
			if _, err = adapter.CaptureOCRGeometry(ctx, ref.ID(), &invalid); err == nil {
				t.Fatal("outside zone accepted")
			}
			after := action("state")
			if after.Focus != initial.Focus {
				t.Fatal("geometry/capture stole focus")
			}
			protected := findNamedReference(t, tree, "Protected target")
			if !protected.Valid() {
				t.Fatal("protected entry missing")
			}
			if _, err = adapter.CaptureOCRGeometry(ctx, protected.ID(), nil); err != nil {
				t.Fatalf("masked password capture geometry unavailable: %v", err)
			}
			action("hide")
			if _, err = adapter.CaptureOCRGeometry(ctx, ref.ID(), nil); err == nil {
				t.Fatal("hidden source accepted")
			}
		})
	}
}