build : remove LLAMA_HTTPLIB option (#19623)

This option was introduced as a workaround because cpp-httplib could not
build on visionOS. Since it has been fixed and now compiles on all platforms,
we can remove it and simplify many things.

Signed-off-by: Adrien Gallouët <angt@huggingface.co>

CMakeLists.txt

@@ -112,7 +112,6 @@ option(LLAMA_TOOLS_INSTALL  "llama: install tools"        ${LLAMA_TOOLS_INSTALL_
 option(LLAMA_TESTS_INSTALL  "llama: install tests"        ON)
 
 # 3rd party libs
-option(LLAMA_HTTPLIB    "llama: httplib for downloading functionality" ON)
 option(LLAMA_OPENSSL    "llama: use openssl to support HTTPS" ON)
 option(LLAMA_LLGUIDANCE "llama-common: include LLGuidance library for structured output in common utils" OFF)
 
@@ -197,9 +196,7 @@ add_subdirectory(src)
 
 if (LLAMA_BUILD_COMMON)
     add_subdirectory(common)
-    if (LLAMA_HTTPLIB)
+    add_subdirectory(vendor/cpp-httplib)
-        add_subdirectory(vendor/cpp-httplib)
-    endif()
 endif()
 
 if (LLAMA_BUILD_COMMON AND LLAMA_BUILD_TESTS AND NOT CMAKE_JS_VERSION)

build-xcframework.sh

@@ -449,10 +449,9 @@ cmake -B build-visionos -G Xcode \
     -DCMAKE_SYSTEM_NAME=visionOS \
     -DCMAKE_OSX_SYSROOT=xros \
     -DCMAKE_XCODE_ATTRIBUTE_SUPPORTED_PLATFORMS=xros \
-    -DCMAKE_C_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_C_FLAGS}" \
+    -DCMAKE_C_FLAGS="${COMMON_C_FLAGS}" \
-    -DCMAKE_CXX_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_CXX_FLAGS}" \
+    -DCMAKE_CXX_FLAGS="${COMMON_CXX_FLAGS}" \
     -DLLAMA_OPENSSL=OFF \
-    -DLLAMA_HTTPLIB=OFF \
     -DLLAMA_BUILD_SERVER=OFF \
     -S .
 cmake --build build-visionos --config Release -- -quiet
@@ -465,10 +464,9 @@ cmake -B build-visionos-sim -G Xcode \
     -DCMAKE_SYSTEM_NAME=visionOS \
     -DCMAKE_OSX_SYSROOT=xrsimulator \
     -DCMAKE_XCODE_ATTRIBUTE_SUPPORTED_PLATFORMS=xrsimulator \
-    -DCMAKE_C_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_C_FLAGS}" \
+    -DCMAKE_C_FLAGS="${COMMON_C_FLAGS}" \
-    -DCMAKE_CXX_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_CXX_FLAGS}" \
+    -DCMAKE_CXX_FLAGS="${COMMON_CXX_FLAGS}" \
     -DLLAMA_OPENSSL=OFF \
-    -DLLAMA_HTTPLIB=OFF \
     -DLLAMA_BUILD_SERVER=OFF \
     -S .
 cmake --build build-visionos-sim --config Release -- -quiet

common/CMakeLists.txt

@@ -112,11 +112,7 @@ endif()
 
 # TODO: use list(APPEND LLAMA_COMMON_EXTRA_LIBS ...)
 set(LLAMA_COMMON_EXTRA_LIBS build_info)
-
+set(LLAMA_COMMON_EXTRA_LIBS ${LLAMA_COMMON_EXTRA_LIBS} cpp-httplib)
-if (LLAMA_HTTPLIB)
-    target_compile_definitions(${TARGET} PUBLIC LLAMA_USE_HTTPLIB)
-    set(LLAMA_COMMON_EXTRA_LIBS ${LLAMA_COMMON_EXTRA_LIBS} cpp-httplib)
-endif()
 
 if (LLAMA_LLGUIDANCE)
     include(ExternalProject)

common/download.cpp

@@ -19,9 +19,7 @@
 #include <thread>
 #include <vector>
 
-#if defined(LLAMA_USE_HTTPLIB)
 #include "http.h"
-#endif
 
 #ifndef __EMSCRIPTEN__
 #ifdef __linux__
@@ -142,8 +140,6 @@ std::pair<std::string, std::string> common_download_split_repo_tag(const std::st
     return {hf_repo, tag};
 }
 
-#if defined(LLAMA_USE_HTTPLIB)
-
 class ProgressBar {
     static inline std::mutex mutex;
     static inline std::map<const ProgressBar *, int> lines;
@@ -768,30 +764,6 @@ std::string common_docker_resolve_model(const std::string & docker) {
     }
 }
 
-#else
-
-common_hf_file_res common_get_hf_file(const std::string &, const std::string &, bool, const common_header_list &) {
-    throw std::runtime_error("download functionality is not enabled in this build");
-}
-
-bool common_download_model(const common_params_model &, const std::string &, bool, const common_header_list &) {
-    throw std::runtime_error("download functionality is not enabled in this build");
-}
-
-std::string common_docker_resolve_model(const std::string &) {
-    throw std::runtime_error("download functionality is not enabled in this build");
-}
-
-int common_download_file_single(const std::string &,
-                                const std::string &,
-                                const std::string &,
-                                bool,
-                                const common_header_list &) {
-    throw std::runtime_error("download functionality is not enabled in this build");
-}
-
-#endif // defined(LLAMA_USE_HTTPLIB)
-
 std::vector<common_cached_model_info> common_list_cached_models() {
     std::vector<common_cached_model_info> models;
     const std::string cache_dir = fs_get_cache_directory();

scripts/sync_vendor.py

@@ -5,7 +5,7 @@ import os
 import sys
 import subprocess
 
-HTTPLIB_VERSION = "f80864ca031932351abef49b74097c67f14719c6"
+HTTPLIB_VERSION = "d4180e923f846b44a3d30acd938438d6e64fc9f6"
 
 vendor = {
     "https://github.com/nlohmann/json/releases/latest/download/json.hpp":     "vendor/nlohmann/json.hpp",

tools/server/CMakeLists.txt

@@ -28,10 +28,6 @@ target_link_libraries(${TARGET} PUBLIC common mtmd ${CMAKE_THREAD_LIBS_INIT})
 
 set(TARGET llama-server)
 
-if (NOT LLAMA_HTTPLIB)
-    message(FATAL_ERROR "LLAMA_HTTPLIB is OFF, cannot build llama-server. Hint: to skip building server, set -DLLAMA_BUILD_SERVER=OFF")
-endif()
-
 set(TARGET_SRCS
     server.cpp
     server-http.cpp

vendor/cpp-httplib/httplib.cpp

@@ -1264,78 +1264,32 @@ int poll_wrapper(struct pollfd *fds, nfds_t nfds, int timeout) {
 #endif
 }
 
-template <bool Read>
+ssize_t select_impl(socket_t sock, short events, time_t sec,
-ssize_t select_impl(socket_t sock, time_t sec, time_t usec) {
+                           time_t usec) {
-#ifdef __APPLE__
-  if (sock >= FD_SETSIZE) { return -1; }
-
-  fd_set fds, *rfds, *wfds;
-  FD_ZERO(&fds);
-  FD_SET(sock, &fds);
-  rfds = (Read ? &fds : nullptr);
-  wfds = (Read ? nullptr : &fds);
-
-  timeval tv;
-  tv.tv_sec = static_cast<long>(sec);
-  tv.tv_usec = static_cast<decltype(tv.tv_usec)>(usec);
-
-  return handle_EINTR([&]() {
-    return select(static_cast<int>(sock + 1), rfds, wfds, nullptr, &tv);
-  });
-#else
   struct pollfd pfd;
   pfd.fd = sock;
-  pfd.events = (Read ? POLLIN : POLLOUT);
+  pfd.events = events;
+  pfd.revents = 0;
 
   auto timeout = static_cast<int>(sec * 1000 + usec / 1000);
 
   return handle_EINTR([&]() { return poll_wrapper(&pfd, 1, timeout); });
-#endif
 }
 
 ssize_t select_read(socket_t sock, time_t sec, time_t usec) {
-  return select_impl<true>(sock, sec, usec);
+  return select_impl(sock, POLLIN, sec, usec);
 }
 
 ssize_t select_write(socket_t sock, time_t sec, time_t usec) {
-  return select_impl<false>(sock, sec, usec);
+  return select_impl(sock, POLLOUT, sec, usec);
 }
 
 Error wait_until_socket_is_ready(socket_t sock, time_t sec,
                                         time_t usec) {
-#ifdef __APPLE__
-  if (sock >= FD_SETSIZE) { return Error::Connection; }
-
-  fd_set fdsr, fdsw;
-  FD_ZERO(&fdsr);
-  FD_ZERO(&fdsw);
-  FD_SET(sock, &fdsr);
-  FD_SET(sock, &fdsw);
-
-  timeval tv;
-  tv.tv_sec = static_cast<long>(sec);
-  tv.tv_usec = static_cast<decltype(tv.tv_usec)>(usec);
-
-  auto ret = handle_EINTR([&]() {
-    return select(static_cast<int>(sock + 1), &fdsr, &fdsw, nullptr, &tv);
-  });
-
-  if (ret == 0) { return Error::ConnectionTimeout; }
-
-  if (ret > 0 && (FD_ISSET(sock, &fdsr) || FD_ISSET(sock, &fdsw))) {
-    auto error = 0;
-    socklen_t len = sizeof(error);
-    auto res = getsockopt(sock, SOL_SOCKET, SO_ERROR,
-                          reinterpret_cast<char *>(&error), &len);
-    auto successful = res >= 0 && !error;
-    return successful ? Error::Success : Error::Connection;
-  }
-
-  return Error::Connection;
-#else
   struct pollfd pfd_read;
   pfd_read.fd = sock;
   pfd_read.events = POLLIN | POLLOUT;
+  pfd_read.revents = 0;
 
   auto timeout = static_cast<int>(sec * 1000 + usec / 1000);
 
@@ -1354,7 +1308,6 @@ Error wait_until_socket_is_ready(socket_t sock, time_t sec,
   }
 
   return Error::Connection;
-#endif
 }
 
 bool is_socket_alive(socket_t sock) {
@@ -7138,17 +7091,6 @@ Server::process_request(Stream &strm, const std::string &remote_addr,
   res.version = "HTTP/1.1";
   res.headers = default_headers_;
 
-#ifdef __APPLE__
-  // Socket file descriptor exceeded FD_SETSIZE...
-  if (strm.socket() >= FD_SETSIZE) {
-    Headers dummy;
-    detail::read_headers(strm, dummy);
-    res.status = StatusCode::InternalServerError_500;
-    output_error_log(Error::ExceedMaxSocketDescriptorCount, &req);
-    return write_response(strm, close_connection, req, res);
-  }
-#endif
-
   // Request line and headers
   if (!parse_request_line(line_reader.ptr(), req)) {
     res.status = StatusCode::BadRequest_400;
@@ -12063,7 +12005,7 @@ bool get_cert_sans(cert_t cert, std::vector<SanEntry> &sans) {
   if (!names) return true; // No SANs is valid
 
   auto count = sk_GENERAL_NAME_num(names);
-  for (int i = 0; i < count; i++) {
+  for (decltype(count) i = 0; i < count; i++) {
     auto gen = sk_GENERAL_NAME_value(names, i);
     if (!gen) continue;
 

vendor/cpp-httplib/httplib.h

@@ -8,8 +8,8 @@
 #ifndef CPPHTTPLIB_HTTPLIB_H
 #define CPPHTTPLIB_HTTPLIB_H
 
-#define CPPHTTPLIB_VERSION "0.31.0"
+#define CPPHTTPLIB_VERSION "0.32.0"
-#define CPPHTTPLIB_VERSION_NUM "0x001F00"
+#define CPPHTTPLIB_VERSION_NUM "0x002000"
 
 /*
  * Platform compatibility check