fix: support path with special characters on windows (#1730)

* fix: utf8 * fix: uft8 for cli * fix: codecvt_utf8_utf16 is deprecated * fix: more * feat: support wstring string conversion * fix: build * fix: engine path env * fix: wstring * fix: cli start server * fix: utf8 file * fix: get env * fix: db * fix: e2e * fix: e2e * fix: cli delete * fix: comment * fix: e2e windows * fix: e2e windows continue * fix: e2e windows skip because of progress bar log issue * fix: add sleep in case of cuda for e2e * fix: import --------- Co-authored-by: vansangpfiev <[email protected]>
janhq · Nov 28, 2024 · 719de33 · 719de33
1 parent 43dab3b
commit 719de33
Show file tree

Hide file tree

Showing 24 changed files with 340 additions and 140 deletions.
diff --git a/engine/CMakeLists.txt b/engine/CMakeLists.txt
@@ -28,6 +28,10 @@ if(MSVC)
       $<$<CONFIG:Debug>:/MTd> #---|-- Statically link the runtime libraries
       $<$<CONFIG:Release>:/MT> #--|
   )
+
+  add_compile_options(/utf-8)
+  add_definitions(-DUNICODE -D_UNICODE)
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /DUNICODE /D_UNICODE")
 endif()
 
 if(NOT DEFINED CORTEX_VARIANT)

diff --git a/engine/cli/CMakeLists.txt b/engine/cli/CMakeLists.txt
@@ -26,6 +26,10 @@ if(MSVC)
       $<$<CONFIG:Debug>:/MTd> #---|-- Statically link the runtime libraries
       $<$<CONFIG:Release>:/MT> #--|
   )
+
+  add_compile_options(/utf-8)
+  add_definitions(-DUNICODE -D_UNICODE)
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /DUNICODE /D_UNICODE")
 endif()
 
 if(NOT DEFINED CORTEX_VARIANT)

diff --git a/engine/cli/commands/server_start_cmd.cc b/engine/cli/commands/server_start_cmd.cc
@@ -2,6 +2,7 @@
 #include "commands/cortex_upd_cmd.h"
 #include "utils/cortex_utils.h"
 #include "utils/file_manager_utils.h"
+#include "utils/widechar_conv.h"
 
 namespace commands {
 
@@ -57,24 +58,32 @@ bool ServerStartCmd::Exec(const std::string& host, int port,
   ZeroMemory(&si, sizeof(si));
   si.cb = sizeof(si);
   ZeroMemory(&pi, sizeof(pi));
-  std::string params = "--start-server";
-  params += " --config_file_path " + get_config_file_path();
-  params += " --data_folder_path " + get_data_folder_path();
-  params += " --loglevel " + log_level_;
-  std::string cmds = cortex_utils::GetCurrentPath() + "/" + exe + " " + params;
+  std::wstring params = L"--start-server";
+  params += L" --config_file_path " +
+            file_manager_utils::GetConfigurationPath().wstring();
+  params += L" --data_folder_path " +
+            file_manager_utils::GetCortexDataPath().wstring();
+  params += L" --loglevel " + cortex::wc::Utf8ToWstring(log_level_);
+  std::wstring exe_w = cortex::wc::Utf8ToWstring(exe);
+  std::wstring current_path_w =
+      file_manager_utils::GetExecutableFolderContainerPath().wstring();
+  std::wstring wcmds = current_path_w + L"/" + exe_w + L" " + params;
+  CTL_DBG("wcmds: " << wcmds);
+  std::vector<wchar_t> mutable_cmds(wcmds.begin(), wcmds.end());
+  mutable_cmds.push_back(L'\0');
   // Create child process
   if (!CreateProcess(
           NULL,  // No module name (use command line)
-          const_cast<char*>(
-              cmds.c_str()),  // Command line (replace with your actual executable)
-          NULL,               // Process handle not inheritable
-          NULL,               // Thread handle not inheritable
-          FALSE,              // Set handle inheritance to FALSE
-          0,                  // No creation flags
-          NULL,               // Use parent's environment block
-          NULL,               // Use parent's starting directory
-          &si,                // Pointer to STARTUPINFO structure
-          &pi))               // Pointer to PROCESS_INFORMATION structure
+          mutable_cmds
+              .data(),  // Command line (replace with your actual executable)
+          NULL,         // Process handle not inheritable
+          NULL,         // Thread handle not inheritable
+          FALSE,        // Set handle inheritance
+          0,            // No creation flags
+          NULL,         // Use parent's environment block
+          NULL,         // Use parent's starting directory
+          &si,          // Pointer to STARTUPINFO structure
+          &pi))         // Pointer to PROCESS_INFORMATION structure
   {
     std::cout << "Could not start server: " << GetLastError() << std::endl;
     return false;
@@ -115,7 +124,8 @@ bool ServerStartCmd::Exec(const std::string& host, int port,
     std::string p = cortex_utils::GetCurrentPath() + "/" + exe;
     execl(p.c_str(), exe.c_str(), "--start-server", "--config_file_path",
           get_config_file_path().c_str(), "--data_folder_path",
-          get_data_folder_path().c_str(), "--loglevel", log_level_.c_str(), (char*)0);
+          get_data_folder_path().c_str(), "--loglevel", log_level_.c_str(),
+          (char*)0);
   } else {
     // Parent process
     if (!TryConnectToServer(host, port)) {

diff --git a/engine/cli/main.cc b/engine/cli/main.cc
@@ -25,6 +25,9 @@
 #error "Unsupported platform!"
 #endif
 
+#include <codecvt>
+#include <locale>
+
 void RemoveBinaryTempFileIfExists() {
   auto temp =
       file_manager_utils::GetExecutableFolderContainerPath() / "cortex_temp";
@@ -40,11 +43,20 @@ void RemoveBinaryTempFileIfExists() {
 void SetupLogger(trantor::FileLogger& async_logger, bool verbose) {
   if (!verbose) {
     auto config = file_manager_utils::GetCortexConfig();
+
     std::filesystem::create_directories(
+#if defined(_WIN32)
+        std::filesystem::u8path(config.logFolderPath) /
+#else
         std::filesystem::path(config.logFolderPath) /
+#endif
         std::filesystem::path(cortex_utils::logs_folder));
-    async_logger.setFileName(config.logFolderPath + "/" +
-                             cortex_utils::logs_cli_base_name);
+
+    // Do not need to use u8path here because trantor handles itself
+    async_logger.setFileName(
+        (std::filesystem::path(config.logFolderPath) /
+         std::filesystem::path(cortex_utils::logs_cli_base_name))
+            .string());
     async_logger.setMaxLines(config.maxLogLines);  // Keep last 100000 lines
     async_logger.startLogging();
     trantor::Logger::setOutputFunction(
@@ -192,8 +204,7 @@ int main(int argc, char* argv[]) {
   // Check if server exists, if not notify to user to install server
   auto exe = commands::GetCortexServerBinary();
   auto server_binary_path =
-      std::filesystem::path(cortex_utils::GetCurrentPath()) /
-      std::filesystem::path(exe);
+      file_manager_utils::GetExecutableFolderContainerPath() / exe;
   if (!std::filesystem::exists(server_binary_path)) {
     std::cout << CORTEX_CPP_VERSION
               << " requires server binary, to install server, run: "

diff --git a/engine/database/database.h b/engine/database/database.h
@@ -20,7 +20,7 @@ class Database {
 
  private:
   Database()
-      : db_(file_manager_utils::GetCortexDataPath().string() + "/cortex.db",
+      : db_(file_manager_utils::GetCortexDataPath() / "cortex.db",
             SQLite::OPEN_READWRITE | SQLite::OPEN_CREATE) {}
   SQLite::Database db_;
 };

diff --git a/engine/e2e-test/test_api_engine_uninstall.py b/engine/e2e-test/test_api_engine_uninstall.py
@@ -1,4 +1,5 @@
 import pytest
+import time
 import requests
 from test_runner import (
     run,
@@ -21,26 +22,26 @@ def setup_and_teardown(self):
 
         # Teardown
         stop_server()
-
-    def test_engines_uninstall_llamacpp_should_be_successful(self):
-        # install first, using cli for synchronously
-        run(
-            "Install Engine",
-            ["engines", "install", "llama-cpp"],
-            timeout=120,
-            capture=False,
-        )
+
+    @pytest.mark.asyncio
+    async def test_engines_uninstall_llamacpp_should_be_successful(self):
+        response = requests.post("http://localhost:3928/v1/engines/llama-cpp/install")
+        assert response.status_code == 200
+        await wait_for_websocket_download_success_event(timeout=None)
+        time.sleep(30)
+
         response = requests.delete("http://localhost:3928/v1/engines/llama-cpp/install")
         assert response.status_code == 200
 
-    def test_engines_uninstall_llamacpp_with_only_version_should_be_failed(self):
+    @pytest.mark.asyncio
+    async def test_engines_uninstall_llamacpp_with_only_version_should_be_failed(self):
         # install first
-        run(
-            "Install Engine",
-            ["engines", "install", "llama-cpp", "-v", "v0.1.35"],
-            timeout=None,
-            capture=False,
+        data = {"variant": "mac-arm64"}
+        install_response = requests.post(
+            "http://127.0.0.1:3928/v1/engines/llama-cpp/install", json=data
         )
+        await wait_for_websocket_download_success_event(timeout=120)
+        assert install_response.status_code == 200
 
         data = {"version": "v0.1.35"}
         response = requests.delete(

diff --git a/engine/e2e-test/test_api_model_start.py b/engine/e2e-test/test_api_model_start.py
@@ -1,8 +1,10 @@
 import pytest
+import time
 import requests
 from test_runner import run, start_server, stop_server
-
-
+from test_runner import (
+    wait_for_websocket_download_success_event
+)
 class TestApiModelStart:
 
     @pytest.fixture(autouse=True)
@@ -12,20 +14,28 @@ def setup_and_teardown(self):
         success = start_server()
         if not success:
             raise Exception("Failed to start server")
-        run("Install engine", ["engines", "install", "llama-cpp"], 5 * 60)
         run("Delete model", ["models", "delete", "tinyllama:gguf"])
-        run(
-            "Pull model",
-            ["pull", "tinyllama:gguf"],
-            timeout=None,
-        )
 
         yield
 
         # Teardown
         stop_server()
-
-    def test_models_start_should_be_successful(self):
+
+    @pytest.mark.asyncio
+    async def test_models_start_should_be_successful(self):
+        response = requests.post("http://localhost:3928/v1/engines/llama-cpp/install")
+        assert response.status_code == 200
+        await wait_for_websocket_download_success_event(timeout=None)
+        # TODO(sang) need to fix for cuda download
+        time.sleep(30)
+
+        json_body = {
+            "model": "tinyllama:gguf"
+        }
+        response = requests.post("http://localhost:3928/v1/models/pull", json=json_body)
+        assert response.status_code == 200, f"Failed to pull model: tinyllama:gguf"
+        await wait_for_websocket_download_success_event(timeout=None)
+
         json_body = {"model": "tinyllama:gguf"}
         response = requests.post(
             "http://localhost:3928/v1/models/start", json=json_body

diff --git a/engine/e2e-test/test_api_model_stop.py b/engine/e2e-test/test_api_model_stop.py
@@ -1,7 +1,10 @@
 import pytest
+import time
 import requests
 from test_runner import run, start_server, stop_server
-
+from test_runner import (
+    wait_for_websocket_download_success_event
+)
 
 class TestApiModelStop:
 
@@ -13,14 +16,19 @@ def setup_and_teardown(self):
         if not success:
             raise Exception("Failed to start server")
 
-        run("Install engine", ["engines", "install", "llama-cpp"], 5 * 60)
         yield
 
         run("Uninstall engine", ["engines", "uninstall", "llama-cpp"])
         # Teardown
         stop_server()
 
-    def test_models_stop_should_be_successful(self):
+    @pytest.mark.asyncio
+    async def test_models_stop_should_be_successful(self):
+        response = requests.post("http://localhost:3928/v1/engines/llama-cpp/install")
+        assert response.status_code == 200
+        await wait_for_websocket_download_success_event(timeout=None)
+        time.sleep(30)
+
         json_body = {"model": "tinyllama:gguf"}
         response = requests.post(
             "http://localhost:3928/v1/models/start", json=json_body

diff --git a/engine/e2e-test/test_cli_engine_install.py b/engine/e2e-test/test_cli_engine_install.py
@@ -19,6 +19,7 @@ def setup_and_teardown(self):
         # Teardown
         stop_server()
 
+    @pytest.mark.skipif(platform.system() == "Windows", reason="Progress bar log issue on Windows")
     def test_engines_install_llamacpp_should_be_successfully(self):
         exit_code, output, error = run(
             "Install Engine",
@@ -46,6 +47,7 @@ def test_engines_install_onnx_on_tensorrt_should_be_failed(self):
         assert "is not supported on" in output, "Should display error message"
         assert exit_code == 0, f"Install engine failed with error: {error}"
 
+    @pytest.mark.skipif(platform.system() == "Windows", reason="Progress bar log issue on Windows")
     def test_engines_install_pre_release_llamacpp(self):
         engine_version = "v0.1.29"
         exit_code, output, error = run(
@@ -67,6 +69,7 @@ def test_engines_install_pre_release_llamacpp(self):
         assert is_engine_version_exist, f"Engine version {engine_version} is not found"
         assert exit_code == 0, f"Install engine failed with error: {error}"
 
+    @pytest.mark.skipif(platform.system() == "Windows", reason="Progress bar log issue on Windows")
     def test_engines_should_fallback_to_download_llamacpp_engine_if_not_exists(self):
         exit_code, output, error = run(
             "Install Engine",

diff --git a/engine/e2e-test/test_cli_engine_uninstall.py b/engine/e2e-test/test_cli_engine_uninstall.py
@@ -25,7 +25,7 @@ def setup_and_teardown(self):
     @pytest.mark.asyncio
     async def test_engines_uninstall_llamacpp_should_be_successfully(self):
         requests.post("http://127.0.0.1:3928/v1/engines/llama-cpp/install")
-        await wait_for_websocket_download_success_event(timeout=120)
+        await wait_for_websocket_download_success_event(timeout=None)
         exit_code, output, error = run(
             "Uninstall engine", ["engines", "uninstall", "llama-cpp"]
         )

diff --git a/engine/e2e-test/test_cli_model_delete.py b/engine/e2e-test/test_cli_model_delete.py
@@ -1,6 +1,10 @@
 import pytest
+import requests
 from test_runner import popen, run
 from test_runner import start_server, stop_server
+from test_runner import (
+    wait_for_websocket_download_success_event
+)
 
 class TestCliModelDelete:
 
@@ -11,18 +15,22 @@ def setup_and_teardown(self):
         if not success:
             raise Exception("Failed to start server")
 
-        # Pull model
-
-        run("Pull model", ["pull", "tinyllama:gguf"], timeout=None,)
-
         yield
 
         # Teardown
         # Clean up
         run("Delete model", ["models", "delete", "tinyllama:gguf"])
         stop_server()
 
-    def test_models_delete_should_be_successful(self):
+    @pytest.mark.asyncio    
+    async def test_models_delete_should_be_successful(self):
+        json_body = {
+            "model": "tinyllama:gguf"
+        }
+        response = requests.post("http://localhost:3928/v1/models/pull", json=json_body)
+        assert response.status_code == 200, f"Failed to pull model: tinyllama:gguf"
+        await wait_for_websocket_download_success_event(timeout=None)
+
         exit_code, output, error = run(
             "Delete model", ["models", "delete", "tinyllama:gguf"]
         )