The way read/select are implemented in unix.cc means that block reads are typically still pulling data a few bytes at a time. This can consume a lot of CPU for reads of 20+ consecutive bytes.

One of our customers has handled this by adding a lowlevel_read method, which gives the user a direct line to the underlying read() method: https://github.com/cedricpradalier/serial/blob/master/src/impl/unix.cc#L421

Another alternative would be to introduce a usleep between the select and read calls. If you know how many bytes are expected and the baud rate, you can compute how long of a pause will allow the entire message to arrive. I don't know how well aligned such functionality would be with the design goals of Serial, but it's something to consider.

As it stands, we have a sensor which mostly saturates a 115kbps serial line, and its Serial-based driver is still taking more CPU than we'd like.

A simpler, gentler take on what was previously discussed in #42 and #43.

• Everything is handled in ms, with a timespec_from_ms function providing last-minute conversion to timespecs for use in pselect.
• A MillisecondTimer class is provided which handles the total_timeout functionality in read and write. Its set/remaining is in millis, but internally it stores the expiry point as a timespec for maximal resolution.
• No funky math or operations on timespecs.

~~While the timespec PR is waiting on reviews, I want to get the ball rolling on this one, which is the actual change that one sets up for, based on the discussion in #37. Click here for a diff showing just the additional changes in this PR.~~

Haven't had a chance to test much at this point, but it will be going into battle on two different peripheral drivers in the next day or two, so there will be some practical miles on it long before an actual merge happens.

For the moment, it's also missing the stubs in the Windows implementation; compilation there will fail until those are added in.

Okay, the timespec PR is merged, and this one has been rebased accordingly.

This PR addresses #42. Summary:

• Introduce a new unix-timespec.h header, which includes timespec operators for +, -, * by int, min, and conversion from millis. Rationale for new file is logical separation and better testability (though no tests at this point).
• Cache the timeout values as timespecs in the SerialImpl, as that's the form in which they are exclusively used.
• Call setTimeout() with a default Timeout obj from the constructor, to initialize the timespec caches to 0.
• Simplify the timing logic in read and write methods to be based both around a fixed endtime, rather than timeout which is adjusted multiple times.
• Introduce an available() check in read, to avoid going to the first select() if data is already waiting. This is important for devices which deliver a larger amount of data (eg, 5000+ bytes/sec), and is independent of the select/pselect switchover.

Testing done has been against the um6 driver in Ubuntu Raring. The um6 driver uses both read and write.

A followup will introduce the sleepByteTimes() and waitAvailable() functions to the external API.

Hello,

I am having a problem setting the timeout. I am using ROS Hydro. I've catkin installed serial, and used rosbuild to build serial_utils. In my program I have the line

this->serial_port_->setTimeout(10);

When I try to run this I get the error

no matching function for call to 'serial::Serial::setTimeout(Int)'

When I tried changing the line to

this->serial_port_->setTimeout(10,10,10,10,10);

This gives me

  ../lib/libax2550.so: undefined reference to serial::Serial::setBytesize(serial::bytesize_t)'
  ../lib/libax2550.so: undefined reference to `serial::Serial::setBaudrate(unsigned int)'
  ../lib/libax2550.so: undefined reference to `serial::Serial::setTimeout(serial::Timeout&)'
  ../lib/libax2550.so: undefined reference to `serial::utils::SerialListener::SerialListener(unsigned long)'
  ../lib/libax2550.so: undefined reference to `serial::Serial::Serial(std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, unsigned int, serial::Timeout, serial::bytesize_t, serial::parity_t, serial::stopbits_t, serial::flowcontrol_t)'
  ../lib/libax2550.so: undefined reference to `serial::utils::SerialListener::startListening(serial::Serial&)'
  ../lib/libax2550.so: undefined reference to `serial::Serial::open()'
  ../lib/libax2550.so: undefined reference to `serial::Serial::write(std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)'
  ../lib/libax2550.so: undefined reference to `serial::utils::SerialListener::~SerialListener()'
  ../lib/libax2550.so: undefined reference to `serial::Serial::setPort(std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)'
  ../lib/libax2550.so: undefined reference to `serial::utils::SerialListener::createBufferedFilter(boost::function<bool (std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)>, unsigned long)'
  ../lib/libax2550.so: undefined reference to `serial::utils::SerialListener::stopListening()'
  ../lib/libax2550.so: undefined reference to `serial::Serial::setStopbits(serial::stopbits_t)'
  ../lib/libax2550.so: undefined reference to `serial::Serial::setParity(serial::parity_t)'
  ../lib/libax2550.so: undefined reference to `serial::utils::SerialListener::createFilter(boost::function<bool (std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)>, boost::function<void (std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)>)'

Hello,

What could be a valid value argument for the serial port in order to run the example? I tried COM1,COM2,etc and keep receiving the following error: "unhandled exception: I/O Exception:specified port, port_name, doesn't exist."

Any ideas?

Emulated the port enumeration functionality found in pyserial's list_ports.comports().

Available in linux and windows only at the moment. The linux implementation uses standard c++ and some posix calls to scrape sysfs. So it will compile on most "unixy" systems but obviously, will only return results on linux.

When I run "sudo make" for installing serial library I get following messages with error.

Can you please tell me what has gone wrong? I'm stuck here for two days now.

cd build && cmake -DCMAKE_INSTALL_PREFIX=/opt/ros/jade/ ..
-- The C compiler identification is GNU 4.8.4
-- The CXX compiler identification is GNU 4.8.4
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Using CATKIN_DEVEL_PREFIX: /home/odroid/Downloads/serial-master/build/devel
-- Using CMAKE_PREFIX_PATH: 
-- Found PythonInterp: /usr/bin/python (found version "2.7.6") 
-- Using PYTHON_EXECUTABLE: /usr/bin/python
-- Using Debian Python package layout
-- Using empy: /usr/bin/empy
-- Using CATKIN_ENABLE_TESTING: ON
-- Call enable_testing()
-- Using CATKIN_TEST_RESULTS_DIR: /home/odroid/Downloads/serial-master/build/test_results
-- Looking for include file pthread.h
-- Looking for include file pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - not found
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Found gtest sources under '/usr/src/gtest': gtests will be built
-- Using Python nosetests: /usr/bin/nosetests-2.7
-- catkin 0.6.14
Traceback (most recent call last):
  File "/home/odroid/Downloads/serial-master/build/catkin_generated/generate_cached_setup.py", line 20, in <module>
    from catkin.environment_cache import generate_environment_script
ImportError: No module named catkin.environment_cache
CMake Error at /opt/ros/jade/share/catkin/cmake/safe_execute_process.cmake:11 (message):
  execute_process(/usr/bin/python
  "/home/odroid/Downloads/serial-master/build/catkin_generated/generate_cached_setup.py")
  returned error code 1
Call Stack (most recent call first):
  /opt/ros/jade/share/catkin/cmake/all.cmake:186 (safe_execute_process)
  /opt/ros/jade/share/catkin/cmake/catkinConfig.cmake:20 (include)
  CMakeLists.txt:5 (find_package)


-- Configuring incomplete, errors occurred!
See also "/home/odroid/Downloads/serial-master/build/CMakeFiles/CMakeOutput.log".
See also "/home/odroid/Downloads/serial-master/build/CMakeFiles/CMakeError.log".
make: *** [serial] Error 1

Please forgive my ignorance. I am good on Arduino/Atmel AVR, but have never done C or C++ programming on a computer (vs microcontroller).

I want to run your serial example to start. I am using Windows 8 with CodeBlocks, GNU GCC Compiler. I downloaded your library zip file, extracted it, navigated there using command prompt, and typed "make" and I got the following errors:

I can't figure out how to install and use the library to run your test file....

I made catkin optional so this wonderful library can be used without it (now only with cmake) - as when you just want to install it and not use it with ROS.

The changes I made will only take effect if catkin is not found. If you have any comments or things I should fix, then lets discuss.

Tested on: Arch Linux, Ubuntu (both in and outside ROS)

I was able to successfully build and install serial but I can't include it in my CMake project.

This is the error I get:

CMake Warning at src/CMakeLists.txt:21 (find_package):
  By not providing "Findserial.cmake" in CMAKE_MODULE_PATH this project has
  asked CMake to find a package configuration file provided by "serial", but
  CMake did not find one.

  Could not find a package configuration file provided by "serial" with any
  of the following names:

    serialConfig.cmake
    serial-config.cmake

  Add the installation prefix of "serial" to CMAKE_PREFIX_PATH or set
  "serial_DIR" to a directory containing one of the above files.  If "serial"
  provides a separate development package or SDK, be sure it has been
  installed.

I believe it's because serial itself doesn't provide a package configuration file. Is there any solution to this problem?

Add full duplex communications support on Windows.

When I use a separate thread to read the serial data, I notice that the write operation is blocked.

Here is a simple program to illustrate this.

#include <iostream>
#include <chrono>
#include <thread>
#include "serial/serial.h"

using namespace std;
using namespace chrono;
using namespace std::chrono_literals;
using namespace serial;

int main(int argc, char **argv) {
    Serial serial_port("COM5", 9600, Timeout::simpleTimeout(200));

    thread read_serial([&](){
        while (true) {
            uint8_t ch;
            serial_port.read(&ch, 1);
        }
    });

    while (true) {
        // Test write latency.
        auto start = high_resolution_clock::now();
        serial_port.write("Hello world!\n");
        auto end = high_resolution_clock::now();
        double dt = duration<double>(end - start).count();

        cout << "Write using " << (dt * 1000) << " ms" << endl;

        this_thread::sleep_for(500ms);
    }
}

Output:

Write using 0.0576 ms
Write using 314.211 ms
Write using 325.428 ms
Write using 299.138 ms
Write using 524.347 ms
Write using 702.12 ms
Write using 513.909 ms
...

This does not make sense because the serial port is supposed to communicate asynchronously. To fix this problem, I modified win.h and win.cc to enable asynchronous communication by adding the OVERLAPPED option. This also helps to solve the issue #150.

This is the right result:

Write using 0.0093 ms
Write using 0.034 ms
Write using 0.0564 ms
Write using 0.012 ms
Write using 0.0343 ms
Write using 0.0494 ms
Write using 0.0124 ms
Write using 0.01 ms

Serial::SerialImpl::close () { if (is_open_ == true) { if (fd_ != INVALID_HANDLE_VALUE) { int ret; ret = CloseHandle(fd_); if (ret == 0) { stringstream ss; ss << "Error while closing serial port: " << GetLastError(); THROW (IOException, ss.str().c_str()); } else { fd_ = INVALID_HANDLE_VALUE; } } is_open_ = false; } }

Hey!

It seems there is a lot of PR/issues ongoing on this project, and no release has been made for a while.

Are you looking for help for maintaining this project? Being alone on an open source project can be tedious, I might help on small PR if you wish, or roadmap.

Thanks for this nice library BTW! Cheers!

on Mac M1, both PortInfo.description and PortInfo.hardware_id are always "n/a", because usb device type name was changed from "IOUSBDevice" to "IOUSBHostInterface".