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

internal/ocrcapture/capture.go

@@ -0,0 +1,89 @@
+// Copyright (c) 2026 Storm Dragon
+// Authors: Storm Dragon, Angus Kola
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+// Package ocrcapture is used only by the short-lived, headless native helper.
+// The reader's worker uses ocr.Capturer and never calls GTK/X11 capture in-process.
+package ocrcapture
+
+/*
+#cgo pkg-config: gtk+-3.0 x11
+#include "native.h"
+*/
+import "C"
+
+import (
+	"bytes"
+	"context"
+	"image"
+	"image/png"
+	"io"
+	"os"
+	"runtime"
+	"unsafe"
+
+	"git.stormux.org/storm/cthulhu/internal/ocr"
+)
+
+func Run(ctx context.Context, input io.Reader, output io.Writer) error {
+	request, err := ocr.ReadCaptureRequest(input)
+	if err != nil {
+		return err
+	}
+	if err := ctx.Err(); err != nil {
+		return err
+	}
+	header := ocr.CaptureHeader{Request: request, Status: "unavailable"}
+	runtime.LockOSThread()
+	defer runtime.UnlockOSThread()
+	// This backend captures physical X11 pixels, including Xwayland when it
+	// provides them. Native display/source checks decide availability.
+	_ = os.Setenv("GDK_BACKEND", "x11")
+	_ = os.Setenv("NO_AT_BRIDGE", "1")
+	kind := map[ocr.SourceKind]int{ocr.ObjectSource: 1, ocr.WindowSource: 2, ocr.DesktopSource: 3, ocr.ClipboardSource: 4}[request.Kind]
+	native := C.cthulhu_ocr_capture(C.int(kind), C.ulong(request.Foreground), C.ulong(request.Window),
+		C.int(request.Bounds.Min.X), C.int(request.Bounds.Min.Y), C.int(request.Bounds.Max.X), C.int(request.Bounds.Max.Y))
+	defer C.cthulhu_ocr_capture_free(&native)
+	if err := ctx.Err(); err != nil {
+		return err
+	}
+	if native.status != 0 {
+		header.Status = map[int]string{1: "unavailable", 2: "source", 3: "changed", 4: "limit"}[int(native.status)]
+		return ocr.WriteCaptureResponse(output, header, nil)
+	}
+	header.Status = ""
+	var pixels image.Image
+	if request.Kind == ocr.ClipboardSource {
+		data := C.GoBytes(unsafe.Pointer(native.clipboard), native.clipboardLength)
+		config, decodeErr := png.DecodeConfig(bytes.NewReader(data))
+		if decodeErr != nil {
+			header.Status = "source"
+		} else if config.Width <= 0 || config.Height <= 0 || config.Height > ocr.MaxPixels || config.Width > ocr.MaxPixels/config.Height {
+			header.Status = "limit"
+		} else {
+			pixels, decodeErr = png.Decode(bytes.NewReader(data))
+			if decodeErr != nil {
+				header.Status = "source"
+			}
+		}
+	} else {
+		header.SourceWindow = uint64(native.sourceWindow)
+		header.RootBounds = image.Rect(0, 0, int(native.rootWidth), int(native.rootHeight))
+		header.SourceBounds = image.Rect(int(native.sourceX), int(native.sourceY), int(native.sourceX)+int(native.sourceWidth), int(native.sourceY)+int(native.sourceHeight))
+		header.ScreenBounds = image.Rect(int(native.x), int(native.y), int(native.x)+int(native.width), int(native.y)+int(native.height))
+		width, height := int(native.width), int(native.height)
+		stride := int(C.gdk_pixbuf_get_rowstride(native.pixels))
+		data := C.GoBytes(unsafe.Pointer(C.gdk_pixbuf_get_pixels(native.pixels)), C.int((height-1)*stride+width*3))
+		owned := image.NewRGBA(image.Rect(0, 0, width, height))
+		for row := range height {
+			for column := range width {
+				source := row*stride + column*3
+				target := row*owned.Stride + column*4
+				copy(owned.Pix[target:target+3], data[source:source+3])
+				owned.Pix[target+3] = 255
+			}
+		}
+		pixels = owned
+	}
+	return ocr.WriteCaptureResponse(output, header, pixels)
+}