Impl OCR. Close #46

WeChatFerry/rpc/pb_types.h

@@ -56,3 +56,8 @@ typedef struct {
     string mobile;
     string home;
 } UserInfo_t;
+
+typedef struct {
+    int32_t status;
+    string result;
+} OcrResult_t;

WeChatFerry/rpc/proto/wcf.proto

@@ -30,6 +30,7 @@ enum Functions {
     FUNC_GET_CONTACT_INFO = 0x55;
     FUNC_REVOKE_MSG       = 0x56;
     FUNC_DECRYPT_IMAGE    = 0x60;
+    FUNC_EXEC_OCR         = 0x61;
     FUNC_ADD_ROOM_MEMBERS = 0x70;
     FUNC_DEL_ROOM_MEMBERS = 0x71;
     FUNC_INV_ROOM_MEMBERS = 0x72;
@@ -73,6 +74,7 @@ message Response
         DbTables tables      = 8;  // 表列表
         DbRows rows          = 9;  // 行列表
         UserInfo ui          = 10; // 个人信息
+        OcrMsg ocr           = 11; // OCR 结果
     };
 }
 
@@ -217,3 +219,9 @@ message PatMsg
     string roomid = 1; // 群 id
     string wxid   = 2; // wxid
 }
+
+message OcrMsg
+{
+    int32 status  = 1; // 状态
+    string result = 2; // 结果
+}

WeChatFerry/spy/funcs.cpp

@@ -50,6 +50,10 @@ static string get_key(uint8_t header1, uint8_t header2, uint8_t *key)
 
 string DecryptImage(string src, string dir)
 {
+    if (!fs::exists(src)) {
+        return "";
+    }
+
     ifstream in(src.c_str(), ios::binary);
     if (!in.is_open()) {
         LOG_ERROR("Failed to read file {}", src);
@@ -335,3 +339,72 @@ string GetAudio(uint64_t id, string dir)
 
     return mp3path;
 }
+
+OcrResult_t GetOcrResult(string path)
+{
+    OcrResult_t ret = { -1, "" };
+
+    if (!fs::exists(path)) {
+        LOG_ERROR("Can not find: {}", path);
+        return ret;
+    }
+
+    // 路径分隔符有要求，必须为 `\`
+    wstring wsPath = String2Wstring(fs::path(path).make_preferred().string());
+
+    WxString wxPath(wsPath);
+    WxString nullObj;
+    WxString ocrBuffer;
+
+    DWORD ocrCall1 = g_WeChatWinDllAddr + g_WxCalls.ocr.call1;
+    DWORD ocrCall2 = g_WeChatWinDllAddr + g_WxCalls.ocr.call2;
+    DWORD ocrCall3 = g_WeChatWinDllAddr + g_WxCalls.ocr.call3;
+
+    DWORD tmp  = 0;
+    int status = -1;
+    __asm {
+        pushad;
+        pushfd;
+        lea   ecx, ocrBuffer;
+        call  ocrCall1;
+        call  ocrCall2;
+        lea   ecx, nullObj;
+        push  ecx;
+        lea   ecx, tmp;
+        push  ecx;
+        lea   ecx, ocrBuffer;
+        push  ecx;
+        push  0x0;
+        lea   ecx, wxPath;
+        push  ecx;
+        mov   ecx, eax;
+        call  ocrCall3;
+        mov   status, eax;
+        popfd;
+        popad;
+    }
+
+    if (status != 0)
+    {
+        LOG_ERROR("OCR status: {}", to_string(status));
+        return ret; // 识别出错
+    }
+
+    ret.status = status;
+
+    DWORD addr   = (DWORD)&ocrBuffer;
+    DWORD header = GET_DWORD(addr);
+    DWORD num    = GET_DWORD(addr + 0x4);
+    if (num <= 0) {
+        return ret; // 识别内容为空
+    }
+
+    for (uint32_t i = 0; i < num; i++) {
+        DWORD content = GET_DWORD(header);
+        ret.result += Wstring2String(GET_WSTRING(content + 0x14));
+        ret.result += "\n";
+        header = content;
+    }
+
+    return ret;
+}

WeChatFerry/spy/funcs.h

@@ -8,3 +8,4 @@ std::string DecryptImage(std::string src, std::string dst);
 int RefreshPyq(uint64_t id);
 int DownloadAttach(uint64_t id, std::string thumb, std::string extra);
 int RevokeMsg(uint64_t id);
+OcrResult_t GetOcrResult(std::string path);

WeChatFerry/spy/load_calls.cpp

@@ -37,7 +37,9 @@ WxCalls_t wxCalls = {
     /*  call1,    call2,    call3 */
     {0x931730, 0x1D58751, 0x1421940},
     /*  call1,    call2,    call3,    call4,    call5,    call6,    call7,    call8*/
-    {0x78CB40, 0x7F99D0, 0x78CF20, 0x78CEF0, 0xF59E40, 0xBD1A00, 0x7FA980, 0x755060}
+    {0x78CB40, 0x7F99D0, 0x78CF20, 0x78CEF0, 0xF59E40, 0xBD1A00, 0x7FA980, 0x755060},
+    /*  call1,    call2,    call3 */
+    {0x80A800, 0x80F270, 0x13DA3E0}
 };
 
 int LoadCalls(const wchar_t *version, WxCalls_t *calls)

WeChatFerry/spy/rpc_server.cpp

@@ -688,6 +688,31 @@ bool func_decrypt_image(DecPath dec, uint8_t *out, size_t *len)
     return true;
 }
 
+bool func_exec_ocr(char *path, uint8_t *out, size_t *len)
+{
+    Response rsp    = Response_init_default;
+    rsp.func        = Functions_FUNC_EXEC_OCR;
+    rsp.which_msg   = Response_ocr_tag;
+    OcrResult_t ret = { -1, "" };
+
+    if (path == NULL) {
+        LOG_ERROR("Empty path.");
+    } else {
+        ret = GetOcrResult(path);
+    }
+
+    rsp.msg.ocr.status = ret.status;
+    rsp.msg.ocr.result = (char *)ret.result.c_str();
+
+    pb_ostream_t stream = pb_ostream_from_buffer(out, *len);
+    if (!pb_encode(&stream, Response_fields, &rsp)) {
+        LOG_ERROR("Encoding failed: {}", PB_GET_ERROR(&stream));
+        return false;
+    }
+    *len = stream.bytes_written;
+    return true;
+}
+
 bool func_add_room_members(char *roomid, char *wxids, uint8_t *out, size_t *len)
 {
     Response rsp   = Response_init_default;
@@ -901,6 +926,11 @@ static bool dispatcher(uint8_t *in, size_t in_len, uint8_t *out, size_t *out_len
             ret = func_decrypt_image(req.msg.dec, out, out_len);
             break;
         }
+        case Functions_FUNC_EXEC_OCR: {
+            LOG_DEBUG("[Functions_FUNC_EXEC_OCR]");
+            ret = func_exec_ocr(req.msg.str, out, out_len);
+            break;
+        }
         case Functions_FUNC_ADD_ROOM_MEMBERS: {
             LOG_DEBUG("[Functions_FUNC_ADD_ROOM_MEMBERS]");
             ret = func_add_room_members(req.msg.m.roomid, req.msg.m.wxids, out, out_len);

WeChatFerry/spy/spy_types.h

@@ -145,6 +145,12 @@ typedef struct CallInviteCM {
     DWORD call8;
 } CallInviteCM_t;
 
+typedef struct CallOcr {
+    DWORD call1;
+    DWORD call2;
+    DWORD call3;
+} CallOcr_t;
+
 typedef struct WxCalls {
     DWORD login;         // 登录状态
     UserInfoCall_t ui;   // 用户信息
@@ -166,6 +172,7 @@ typedef struct WxCalls {
     CallRichText_t rt;   // 发送消息卡片
     CallPatMsg_t pm;     // 发送拍一拍消息
     CallInviteCM_t irm;  // 邀请群成员
+    CallOcr_t ocr;       // OCR
 } WxCalls_t;
 
 struct WxString {

clients/python/wcferry/client.py

@@ -755,6 +755,40 @@ class Wcf():
         rsp = self._send_request(req)
         return rsp.str
 
+    def get_ocr_result(self, extra: str, timeout: int = 2) -> str:
+        """获取 OCR 结果。鸡肋，需要图片能自动下载；通过下载接口下载的图片无法识别。
+
+        Args:
+            extra (str): 待识别的图片路径，消息里的 extra
+
+        Returns:
+            str: OCR 结果
+        """
+        def _inner(extra):
+            req = wcf_pb2.Request()
+            req.func = wcf_pb2.FUNC_EXEC_OCR  # FUNC_EXEC_OCR
+            req.str = extra
+            rsp = self._send_request(req)
+            ocr = json_format.MessageToDict(rsp.ocr)
+            return ocr.get("status", 0), ocr.get("result", "")
+
+        cnt = 0
+        while True:
+            status, result = _inner(extra)
+            if status == 0:
+                break
+
+            cnt += 1
+            if cnt > timeout:
+                break
+
+            sleep(1)
+
+        if status != 0:
+            self.LOG.error(f"OCR failed, status: {status}")
+
+        return result
+
     def download_image(self, id: int, extra: str, dir: str, timeout: int = 30) -> str:
         """下载图片