Move entirety of llvm namespace to wpi namespace.

During shared library loading, a different libLLVM can be pulled in, causing
llvm symbols from dependent libraries to resolve to that library instead of
this one. This has been seen in the wild with the Mesa OpenGL implementation
in JavaFX applications (see wpilibsuite/shuffleboard#361).

This is clearly a very breaking change. For some level of backwards
compatibility, a namespace alias from llvm to wpi is performed in the "llvm"
headers.  Unfortunately, forward declarations of llvm classes will still break,
but compilers seem to generate clear error messages in those cases
("namespace alias 'llvm' not allowed here, assuming 'wpi'").

This change also moves all the wpiutil headers to a single "wpi" subdirectory
from the previously split "llvm", "support", "tcpsockets", and "udpsockets".
Shim headers will be added for backwards compatibility in a later commit.
This commit is contained in:
Peter Johnson
2018-04-29 23:33:19 -07:00
parent 93859eb84f
commit f84018af5f
377 changed files with 2747 additions and 2742 deletions

View File

@@ -0,0 +1,129 @@
/*----------------------------------------------------------------------------*/
/* Copyright (c) 2017-2018 FIRST. All Rights Reserved. */
/* Open Source Software - may be modified and shared by FRC teams. The code */
/* must be accompanied by the FIRST BSD license file in the root directory of */
/* the project. */
/*----------------------------------------------------------------------------*/
#include "wpi/TCPConnector.h" // NOLINT(build/include_order)
#include <atomic>
#include <chrono>
#include <thread>
#include <tuple>
#include "wpi/SmallSet.h"
#include "wpi/condition_variable.h"
#include "wpi/mutex.h"
using namespace wpi;
// MSVC < 1900 doesn't have support for thread_local
#if !defined(_MSC_VER) || _MSC_VER >= 1900
// clang check for availability of thread_local
#if !defined(__has_feature) || __has_feature(cxx_thread_local)
#define HAVE_THREAD_LOCAL
#endif
#endif
std::unique_ptr<NetworkStream> TCPConnector::connect_parallel(
ArrayRef<std::pair<const char*, int>> servers, Logger& logger,
int timeout) {
if (servers.empty()) return nullptr;
// structure to make sure we don't start duplicate workers
struct GlobalState {
wpi::mutex mtx;
#ifdef HAVE_THREAD_LOCAL
SmallSet<std::pair<std::string, int>, 16> active;
#else
SmallSet<std::tuple<std::thread::id, std::string, int>, 16> active;
#endif
};
#ifdef HAVE_THREAD_LOCAL
thread_local auto global = std::make_shared<GlobalState>();
#else
static auto global = std::make_shared<GlobalState>();
auto this_id = std::this_thread::get_id();
#endif
auto local = global; // copy to an automatic variable for lambda capture
// structure shared between threads and this function
struct Result {
wpi::mutex mtx;
wpi::condition_variable cv;
std::unique_ptr<NetworkStream> stream;
std::atomic<unsigned int> count{0};
std::atomic<bool> done{false};
};
auto result = std::make_shared<Result>();
// start worker threads; this is I/O bound so we don't limit to # of procs
Logger* plogger = &logger;
unsigned int num_workers = 0;
for (const auto& server : servers) {
std::pair<std::string, int> server_copy{std::string{server.first},
server.second};
#ifdef HAVE_THREAD_LOCAL
const auto& active_tracker = server_copy;
#else
std::tuple<std::thread::id, std::string, int> active_tracker{
this_id, server_copy.first, server_copy.second};
#endif
// don't start a new worker if we had a previously still-active connection
// attempt to the same server
{
std::lock_guard<wpi::mutex> lock(local->mtx);
if (local->active.count(active_tracker) > 0) continue; // already in set
}
++num_workers;
// start the worker
std::thread([=]() {
if (!result->done) {
// add to global state
{
std::lock_guard<wpi::mutex> lock(local->mtx);
local->active.insert(active_tracker);
}
// try to connect
auto stream = connect(server_copy.first.c_str(), server_copy.second,
*plogger, timeout);
// remove from global state
{
std::lock_guard<wpi::mutex> lock(local->mtx);
local->active.erase(active_tracker);
}
// successful connection
if (stream) {
std::lock_guard<wpi::mutex> lock(result->mtx);
if (!result->done.exchange(true)) result->stream = std::move(stream);
}
}
++result->count;
result->cv.notify_all();
})
.detach();
}
// wait for a result, timeout, or all finished
std::unique_lock<wpi::mutex> lock(result->mtx);
if (timeout == 0) {
result->cv.wait(
lock, [&] { return result->stream || result->count >= num_workers; });
} else {
auto timeout_time =
std::chrono::steady_clock::now() + std::chrono::seconds(timeout);
result->cv.wait_until(lock, timeout_time, [&] {
return result->stream || result->count >= num_workers;
});
}
// no need to wait for remaining worker threads; shared_ptr will clean up
return std::move(result->stream);
}