chore: suppress warnings on Windows

engine/cli/main.cc

@@ -8,6 +8,7 @@
 #include "utils/file_manager_utils.h"
 #include "utils/logging_utils.h"
 #include "utils/system_info_utils.h"
+#include "utils/widechar_conv.h"
 
 #if defined(__APPLE__) && defined(__MACH__)
 #include <libgen.h>  // for dirname()
@@ -46,7 +47,7 @@ void SetupLogger(trantor::FileLogger& async_logger, bool verbose) {
 
     std::filesystem::create_directories(
 #if defined(_WIN32)
-        std::filesystem::u8path(config.logFolderPath) /
+        std::filesystem::path(cortex::wc::Utf8ToWstring(config.logFolderPath)) /
 #else
         std::filesystem::path(config.logFolderPath) /
 #endif

engine/common/message.h

@@ -107,7 +107,8 @@ struct Message : JsonSerializable {
           std::move(root.get("object", "thread.message").asString());
       message.created_at = root["created_at"].asUInt();
       if (message.created_at == 0 && root["created"].asUInt64() != 0) {
-        message.created_at = root["created"].asUInt64() / 1000;
+        message.created_at =
+            static_cast<uint32_t>(root["created"].asUInt64() / 1000);
       }
       message.thread_id = std::move(root["thread_id"].asString());
       message.status = StatusFromString(std::move(root["status"].asString()));

engine/config/model_config.h

@@ -35,8 +35,7 @@ struct RemoteModelConfig {
 
     // Load basic string fields
     model = json.get("model", model).asString();
-    header_template =
-        json.get("header_template", header_template).asString();
+    header_template = json.get("header_template", header_template).asString();
     engine = json.get("engine", engine).asString();
     version = json.get("version", version).asString();
     created =
@@ -405,7 +404,8 @@ struct ModelConfig {
     oss << format_utils::print_comment("END REQUIRED");
     oss << format_utils::print_comment("BEGIN OPTIONAL");
 
-    oss << format_utils::print_float("size", size);
+    oss << format_utils::print_kv("size", std::to_string(size),
+                                  format_utils::MAGENTA);
     oss << format_utils::print_bool("stream", stream);
     oss << format_utils::print_float("top_p", top_p);
     oss << format_utils::print_float("temperature", temperature);

engine/main.cc

@@ -105,7 +105,7 @@ void RunServer(std::optional<std::string> host, std::optional<int> port,
   // Create logs/ folder and setup log to file
   std::filesystem::create_directories(
 #if defined(_WIN32)
-      std::filesystem::u8path(config.logFolderPath) /
+      std::filesystem::path(cortex::wc::Utf8ToWstring(config.logFolderPath)) /
 #else
       std::filesystem::path(config.logFolderPath) /
 #endif

engine/services/hardware_service.cc

@@ -106,9 +106,19 @@ bool HardwareService::Restart(const std::string& host, int port) {
     return false;
   }
 
+#ifdef _MSC_VER
+  char* value = nullptr;
+  size_t len = 0;
+  _dupenv_s(&value, &len, "CUDA_VISIBLE_DEVICES");
+#else
   const char* value = std::getenv("CUDA_VISIBLE_DEVICES");
+#endif
+
   if (value) {
     LOG_INFO << "CUDA_VISIBLE_DEVICES is set to: " << value;
+#ifdef _MSC_VER
+    free(value);
+#endif
   } else {
     LOG_WARN << "CUDA_VISIBLE_DEVICES is not set.";
   }
@@ -128,9 +138,18 @@ bool HardwareService::Restart(const std::string& host, int port) {
     return false;
   }
 
+#ifdef _MSC_VER
+  char* vk_value = nullptr;
+  _dupenv_s(&vk_value, &len, "GGML_VK_VISIBLE_DEVICES");
+#else
   const char* vk_value = std::getenv("GGML_VK_VISIBLE_DEVICES");
+#endif
+
   if (vk_value) {
     LOG_INFO << "GGML_VK_VISIBLE_DEVICES is set to: " << vk_value;
+#ifdef _MSC_VER
+    free(vk_value);
+#endif
   } else {
     LOG_WARN << "GGML_VK_VISIBLE_DEVICES is not set.";
   }
@@ -240,7 +259,7 @@ bool HardwareService::SetActivateHardwareConfig(
   auto priority = [&ahc](int software_id) -> int {
     for (size_t i = 0; i < ahc.gpus.size(); i++) {
       if (ahc.gpus[i] == software_id)
-        return i;
+        return static_cast<int>(i);
       break;
     }
     return INT_MAX;
@@ -390,16 +409,33 @@ void HardwareService::UpdateHardwareInfos() {
 #if defined(_WIN32) || defined(_WIN64) || defined(__linux__)
   bool has_deactivated_gpu = a.value().size() != activated_gpu_af.size();
   if (!gpus.empty() && has_deactivated_gpu) {
+#ifdef _MSC_VER
+    char* value = nullptr;
+    size_t len = 0;
+    _dupenv_s(&value, &len, "CUDA_VISIBLE_DEVICES");
+#else
     const char* value = std::getenv("CUDA_VISIBLE_DEVICES");
+#endif
     if (value) {
       LOG_INFO << "CUDA_VISIBLE_DEVICES: " << value;
+#ifdef _MSC_VER
+      free(value);
+#endif
     } else {
       need_restart = true;
     }
 
+#ifdef _MSC_VER
+    char* vk_value = nullptr;
+    _dupenv_s(&vk_value, &len, "GGML_VK_VISIBLE_DEVICES");
+#else
     const char* vk_value = std::getenv("GGML_VK_VISIBLE_DEVICES");
+#endif
     if (vk_value) {
       LOG_INFO << "GGML_VK_VISIBLE_DEVICES: " << vk_value;
+#ifdef _MSC_VER
+      free(vk_value);
+#endif
     } else {
       need_restart = true;
     }

engine/services/model_service.cc

@@ -315,7 +315,7 @@ cpp::result<DownloadTask, std::string> ModelService::HandleDownloadUrlAsync(
 
   try {
     std::filesystem::create_directories(local_path.parent_path());
-  } catch (const std::filesystem::filesystem_error& e) {
+  } catch (const std::filesystem::filesystem_error&) {
     // if file exist, remove it
     std::filesystem::remove(local_path.parent_path());
     std::filesystem::create_directories(local_path.parent_path());
@@ -380,7 +380,7 @@ ModelService::EstimateModel(const std::string& model_handle,
     auto mc = yaml_handler.GetModelConfig();
     assert(hw_service_);
     auto hw_info = hw_service_->GetHardwareInfo();
-    auto free_vram_MiB = 0u;
+    int64_t free_vram_MiB = 0;
     for (const auto& gpu : hw_info.gpus) {
       free_vram_MiB += gpu.free_vram;
     }
@@ -444,7 +444,7 @@ cpp::result<std::string, std::string> ModelService::HandleUrl(
 
   try {
     std::filesystem::create_directories(local_path.parent_path());
-  } catch (const std::filesystem::filesystem_error& e) {
+  } catch (const std::filesystem::filesystem_error&) {
     // if file exist, remove it
     std::filesystem::remove(local_path.parent_path());
     std::filesystem::create_directories(local_path.parent_path());
@@ -1326,7 +1326,7 @@ ModelService::MayFallbackToCpu(const std::string& model_path, int ngl,
   }
   // If in GPU acceleration mode:
   // We use all visible GPUs, so only need to sum all free vram
-  auto free_vram_MiB = 0u;
+  int64_t free_vram_MiB = 0;
   for (const auto& gpu : hw_info.gpus) {
     free_vram_MiB += gpu.free_vram;
   }

engine/utils/command_executor.h

@@ -37,7 +37,8 @@ class CommandExecutor {
     std::array<char, 128> buffer;
     std::string result;
 
-    while (fgets(buffer.data(), buffer.size(), m_pipe.get()) != nullptr) {
+    while (fgets(buffer.data(), static_cast<int>(buffer.size()),
+                 m_pipe.get()) != nullptr) {
       result += buffer.data();
     }
 

engine/utils/cortex_utils.h

@@ -1,4 +1,5 @@
 #pragma once
+
 #include <drogon/HttpClient.h>
 #include <drogon/HttpResponse.h>
 #include <sys/stat.h>
@@ -29,9 +30,16 @@ inline std::string logs_cli_base_name = "./logs/cortex-cli.log";
 // example: Mon, 25 Nov 2024 09:57:03 GMT
 inline std::string GetDateRFC1123() {
   std::time_t now = std::time(nullptr);
-  std::tm* gmt_time = std::gmtime(&now);
+  std::tm gmt_time = {};
+#ifdef _MSC_VER
+  gmtime_s(&gmt_time, &now);
+  std::ostringstream oss;
+  oss << std::put_time(&gmt_time, "%a, %d %b %Y %H:%M:%S GMT");
+#else
+  std::tm* gmt_time_ptr = std::gmtime(&now);
   std::ostringstream oss;
-  oss << std::put_time(gmt_time, "%a, %d %b %Y %H:%M:%S GMT");
+  oss << std::put_time(gmt_time_ptr, "%a, %d %b %Y %H:%M:%S GMT");
+#endif
   return oss.str();
 }
 

engine/utils/engine_matcher_utils.h

@@ -51,52 +51,56 @@ inline std::string GetSuitableCudaVariant(
   std::regex cuda_reg("cuda-(\\d+)-(\\d+)");
   std::smatch match;
 
-  int requestedMajor = 0;
-  int requestedMinor = 0;
+  int requested_major = 0;
+  int requested_minor = 0;
 
   if (!cuda_version.empty()) {
-    // Split the provided CUDA version into major and minor parts
-    sscanf(cuda_version.c_str(), "%d.%d", &requestedMajor, &requestedMinor);
+// Split the provided CUDA version into major and minor parts
+#if defined(_MSC_VER)
+    sscanf_s(cuda_version.c_str(), "%d.%d", &requested_major, &requested_minor);
+#else
+    sscanf(cuda_version.c_str(), "%d.%d", &requested_major, &requested_minor);
+#endif
   }
 
-  std::string selectedVariant;
-  int bestMatchMajor = -1;
-  int bestMatchMinor = -1;
+  std::string selected_variant;
+  int best_match_major = -1;
+  int best_match_minor = -1;
 
   for (const auto& variant : variants) {
     if (std::regex_search(variant, match, cuda_reg)) {
       // Found a CUDA version in the variant
-      int variantMajor = std::stoi(match[1]);
-      int variantMinor = std::stoi(match[2]);
+      int variant_major = std::stoi(match[1]);
+      int variant_minor = std::stoi(match[2]);
 
-      if (requestedMajor == variantMajor) {
+      if (requested_major == variant_major) {
         // If the major versions match, prefer the closest minor version
-        if (requestedMinor >= variantMinor &&
-            (variantMajor > bestMatchMajor ||
-             (variantMajor == bestMatchMajor &&
-              variantMinor > bestMatchMinor))) {
-          selectedVariant = variant;
-          bestMatchMajor = variantMajor;
-          bestMatchMinor = variantMinor;
+        if (requested_minor >= variant_minor &&
+            (variant_major > best_match_major ||
+             (variant_major == best_match_major &&
+              variant_minor > best_match_minor))) {
+          selected_variant = variant;
+          best_match_major = variant_major;
+          best_match_minor = variant_minor;
         }
       }
     }
   }
 
   // If no CUDA version is provided, select the variant without any CUDA in the name
-  if (selectedVariant.empty()) {
+  if (selected_variant.empty()) {
     LOG_WARN
         << "No suitable CUDA variant found, selecting a variant without CUDA";
     for (const auto& variant : variants) {
       if (variant.find("cuda") == std::string::npos) {
-        selectedVariant = variant;
-        LOG_INFO << "Found variant without CUDA: " << selectedVariant << "\n";
+        selected_variant = variant;
+        LOG_INFO << "Found variant without CUDA: " << selected_variant << "\n";
         break;
       }
     }
   }
 
-  return selectedVariant;
+  return selected_variant;
 }
 
 inline std::string ValidateTensorrtLlm(const std::vector<std::string>& variants,

engine/utils/format_utils.h

@@ -67,7 +67,7 @@ inline std::string WriteKeyValue(const std::string& key,
         strValue.pop_back();
       }
       out_file << strValue;
-    } catch (const std::exception& e) {
+    } catch (const std::exception&) {
       out_file << value;  // If not a float, write as is
     }
   } else {

engine/utils/hardware/cpu_info.h

@@ -187,7 +187,7 @@ struct CpuInfo {
       return CPU{};
     auto cpu = res[0];
     cortex::cpuid::CpuInfo inst;
-    float usage = GetCPUUsage();
+    auto usage = static_cast<float>(GetCPUUsage());
     return CPU{.cores = cpu.numPhysicalCores(),
                .arch = std::string(GetArch()),
                .model = cpu.modelName(),

engine/utils/hardware/gguf/ggml.h

@@ -49,21 +49,21 @@ inline float GetQuantBit(GGMLType gt) {
   switch (gt) {
     case GGML_TYPE_I32:
     case GGML_TYPE_F32:
-      return 32.0;
+      return 32.0f;
     case GGML_TYPE_I16:
     case GGML_TYPE_BF16:
     case GGML_TYPE_F16:
-      return 16.0;
+      return 16.0f;
     case GGML_TYPE_IQ2_S:
     case GGML_TYPE_IQ2_XXS:
     case GGML_TYPE_IQ2_XS:
-      return 2.31;
+      return 2.31f;
     case GGML_TYPE_Q2_K:
-      return 2.5625;
+      return 2.5625f;
     case GGML_TYPE_IQ3_XXS:
     case GGML_TYPE_IQ3_S:
     case GGML_TYPE_Q3_K:
-      return 3.4375;
+      return 3.4375f;
     case GGML_TYPE_Q4_0_4_4:
     case GGML_TYPE_Q4_0_4_8:
     case GGML_TYPE_Q4_0_8_8:
@@ -72,25 +72,25 @@ inline float GetQuantBit(GGMLType gt) {
     case GGML_TYPE_Q4_0:
     case GGML_TYPE_Q4_1:
     case GGML_TYPE_Q4_K:
-      return 4.5;
+      return 4.5f;
     case GGML_TYPE_Q5_0:
     case GGML_TYPE_Q5_1:
     case GGML_TYPE_Q5_K:
-      return 5.5;
+      return 5.5f;
     case GGML_TYPE_Q6_K:
-      return 6.5625;
+      return 6.5625f;
     case GGML_TYPE_I8:
     case GGML_TYPE_Q8_0:
     case GGML_TYPE_Q8_1:
     case GGML_TYPE_Q8_K:
-      return 8.0;
+      return 8.0f;
 
     case GGML_TYPE_I64:
     case GGML_TYPE_F64:
-      return 64.0;
+      return 64.0f;
 
     default:
-      return 8.0;
+      return 8.0f;
   }
 }