TCP Client [wiring] ERROR: recv error = 128

[GJMorrison]

Using the TCP Client to communicate with another device and it will work for a while and then I get a [wiring] ERROR: recv error = 128 after sending a block of data(1300 - 1400 bytes in size).
What does this error actually mean and how can I prevent it?

I assume it is coming from int TCPClient::available() in spark_wiring_tcpclient_posix.cpp and this code then closes the socket.

I have tried SINGLE_THREADED_BLOCK() and nointerrupts() around the send and receive routines.

Device is connected to Particle at the same time.

Currently trying it on DeviceOS 5.9.0 on an Argon

[rickkas7]

It's not clear what that is. It's an error code from sock_recv() but 128 is not a normal errno. Farther in it comes from lwip_recv() but that doesn't seem like a valid errno code for LWIP either.

I'd guess either a buffer is full or the other side closed the connection.

In any case, don't attempt to make networking calls with interrupts or threading disabled. There's a good chance that the system will deadlock if a buffer is full.

[rickkas7]

The error code is coming from newlib, the standard C/C++ library.

#define ENOTCONN 128		/* Socket is not connected */

The socket disconnected so you can no longer read from it.

[GJMorrison]

The ERROR: recv error = 128 error was actually when using DeviceOS 5.6.0 I had reverted to it as we know it was stable.

Tried DeviceOS 5.9.0 and now getting [wiring] ERROR: recv error = 113

[rickkas7]

The POSIX error codes are now in the docs.

113 is connection aborted, which is what you get if the connection is abruptly closed. I think it can also happen if a router between the two sides loses the route to the other side.

[GJMorrison]

Finally tracked the problem down to a pointer to data storage that no longer existed giving nonsense data that the far end didn't like and it was closing the connection.