Hi everyone, first of all let me thank anyone who's reading this. I really appreciate it. I know time is a valuable commodity so I'll try to summarize what I'd like to achieve in the second paragraph so if this topic is not interesting for you, you can skip it. Also, even though I could not find anything on the mailing list archive related to this topic, if you think this is not the right space to discuss this or has been discussed before somewhere else, I'll kindly ask for pointers or suggestions for a more suitable forum. Let's get to it. What I'm trying to achieve is to create ROS nodes (http://www.ros.org/) for Bluetooth low energy (from now on BLE) communication for robotics applications (like the one I'm working on). I'd like to base it on blueZ since it is the official Linux Bluetooth stack, so the code I would produce could be "easily" maintained, reviewed, and used. The vision I have for this in the long term is to have a family of Bluetooth related stable nodes based on blueZ that could become the de-facto standard for ROS. >From this point, I'll exposed what I tried so far on this quest before I got stuck. So far, based on the little knowledge I managed to gather. The sensible approach was to create a BLE GATT server ROS node, and a BLE GATT client ROS node as a starter. So far I only got to play with the server. Since this is my first Bluetooth related application done on Linux, googling around I found this (https://stackoverflow.com/questions/29128586/bluetooth-low-energy-in-c-using-bluez-to-create-a-gatt-server) where the steps to build and run tools/btgatt-server.c are described. Once I got it building a running, I connected to it through the Nordic "nRF Connect" Android app, and everything was visible including the hearth rate monitor. It worked like a charm, so it made sense to me as a first approach, to try and use the btgatt-server.c source from a ros node. So first I removed all options and defaulted them to the values I was using (no more argument parsing) and once that was working I rewrote main as a regular function (launch_server) and created a dummy main in btgatt-server.c that will call "launch_server". This also worked nicely, so next step was to remove the dummy main and call "launch_server" from the main located in the ROS node file. Surprisingly, that approach failed. Somehow the laptop will advertise himself, the phone connects to it, but it wont display any services (no more heart rate monitor for me). The application running detects the connection but the console output stops rather soon in comparison with the regular code being executed. If I insist re-connecting from the phone will display the primary services "Generic Access" and "Generic Attribute", but the console output remains "dead". This is, for the record, the output I get with my modifications. According to gdb the underlying "main_loop" is still running but nothing else will ever happen. Started listening on ATT channel. Waiting for connections Connect from 6F:50:8F:33:8F:9D Running GATT server [GATT server]# I saw that some functions are used to get the application "stuck" in the mentioned "main_loop" to serve the client petitions from there. Therefore, I thought that the problem could be the ROS integration (as ROS uses a different way of getting the application to serve several ROS events). So to make things more simple in a first step, I just wrote a dumb C project where I just provided a main with a "hello world" printout and I call my "launch_server". The behavior is exactly the same, so I could discard ROS influence for the moment. So at this point, since I managed to build the dumb C project in a different fashion from the ROS one, but behaves exactly as the ROS project, I'm starting to think that the way I build the code might be the key. For the dumb C project I just used a standard eclipse CDT project, so my first guess is that some things might be happening in blueZ makefile that makes the difference between the correct behavior and the bad one. I've been playing around with the makefiles myself for a couple of days but to be honest, I'm not really experienced with them. So I find myself in a position where I can't dig any further without help or a sudden flash of inspiration. If you made it to this point, thank you for you patience! The dumb project code with the modifications just for completion is located in https://gitlab.com/jagomo/dummy-bluez-test (note that btgatt-server.c has been renamed to btgatt-server-ros.c) Keep in mind that the blueZ source in there is not complete as eclipse was trying to build completely blueZ, even files that apparently weren't required to build btgatt-server.c. So I removed some files here and there to make my life easier and with less compilation errors. The ROS project used the full code base, as ROS build tools were smart enough to figure out what made sense to include. Up to this point, several questions arises: - What could I be missing that "makes the magic happen"? Any pointers and hints to keep digging would be greatly appreciated. - Is this approach sensible at all? Is there an easier (in sense of more maintainable in the long run) way to implement the GATT server and client in a custom application using blueZ? If so, which method would you recommend? which libraries would you recommend to master for such method? Do you know any good resources to do so? - Is there more blueZ documentation/resources beyond the source code and doc folder? Specifically, any good resources regarding BLE for programmers using blueZ. Resources that explain in a noob friendly the big picture about blueZ in a more general way are also welcome. Once again, thanks in advance for the help. I'll be grateful for any answer (big, small or partial) shared. Kind regards, Javier Gomez Moreno.
