Add IOControl library

[mikee47]

The IO Control library is an asynchronous Input/Output device stack for embedded devices such as the ESP8266.
The main reason for building this was to provide a mechanism for serialising modbus requests along with commands to other types of device such as DMX512 dimmers and 433MHz light switches.

This PR is related to #1991.

todo:

• Thorough test with real hardware

[mikee47]

Preview documentation https://smingdev.readthedocs.io/en/feature-io-control/_inc/Sming/Libraries/IOControl/index.html

[slaff]

I've tried running the sample under Host but it segfaults with assert. Is that expected under Host or I am doing something wrong?

2745898 MB: Sent 8 bytes...
2746493 devmgrCallback: 00000000 / 00000000
2748053 Request 4d93bc0 created
2750213 setCommand(0x00000009: adjust)
2757379 Request 4d93bc0 (adjust#7) complete - success
app: /tmp/Sming/Sming/Libraries/IOControl/src/Controller.cpp:213: virtual void IO::Controller::handleEvent(IO::Request*, IO::Event): Assertion `m_queue.head() == request' failed.

signal_handler: SIGABRT - usually caused by an abort() or assert()

[mikee47]

The DMX driver needs looking at. If you want to continue testing without it just remove the corresponding call to registerDeviceClass. I'm not actively using DMX (yet) and the library has been through some major revisions, so clearly I broke something!

[slaff]

Mike, can you run codespell to fix small spelling mistakes?

[slaff]

@mikee47 Do you want to make the best of your IDE and replace the occurrences of, if I am not mistaken, Hungarian Notation for variables? I guess these are left-overs from previous versions.

[slaff]

https://github.com/mikee47/IOControl/blob/13976f4b98b8bc5b80ce258d874444264eb77710/src/Modbus/R421A/Device.cpp#L43

@mikee47 replace:

	debug_d("Device %s has %u channels", id().c_str(), channelCount);

with

	debug_d("Device %s has %u channels", getId().c_str(), channelCount);

[slaff]

There might be still concurrency issue(s) and the application is segfaulting. Steps to reproduce it is to compile the sample.

cd samples/Basic_RS485
make SMING_ARCH=Host DEBUG_VERBOSE_LEVEL=3 ENABLE_GDB=1
make run

And wait. Soon the application segfaults during release of a request.

28126301 Executing request 57120070, : update
28126310 [DMX512] updateSlaves()
>: 00000000 00 00 00 27 00 00 00 00 1a 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ...'............................
...
  ................................
   00000200 00 00 00                                                                                         ...
28128732 Request 57120070 () complete - success
28128746 Request 57120070 () destroyed

signal_handler: SIGSEGV - segmentation fault

My wild guess is that this request is still used somewhere after it was freed. This seems to be a timing issue and, unfortunately, I cannot reproduce it under valgrind or gdb.

Can you reproduce it on your system?

[mikee47]

I managed to get it once, using --cpulimit=2 and running other CPU intensive stuff at the same time. I'm doing the same in debugger and so far behaving itself... It could be the host serial emulation (or threading) again...

Different error though:

11612979 Executing request 56ad2f40, : update
11613002 [DMX512] updateSlaves()
>: 00000000 00 00 00 0e ...
11613818 Request 56ad2f40 () complete - success
Device.cpp:200: virtual void IO::DMX512::Device::handleEvent(IO::Request*, IO::Event): Assertion `updating' failed.

signal_handler: SIGABRT - usually caused by an abort() or assert()

[mikee47]

OK, found and fixed problem.

[slaff]

@mikee47 I am very glad that the long-awaited IOControl library made it into the official develop branch. Thanks a lot for your work ! :)