Our product uses an STM32F4xx with the 4343W chip. And our firmware uses ThreadX and NetX.
We have the firmware ported and running on every Wiced SDK version from 4.0 up to the most recent 6.2.
BUT we recently discovered that as of v6.1 (where NetX changed from 5.5 to 5.10), we are now getting spontaneous TCP disconnects from the NetX library.
More specifically, we have code that sits in a loop and does a simple get against our server every 3 seconds. Our code built with v6.0 works very reliably and will run that loop all day if we let it. Same code updated for v6.1 (no significant API changes) makes it through the first minute or two of such gets, but eventually we get the internal_wiced_tcp_disconnect_socket_callback() callback from NetX. And the socket stays down for between 20 seconds and 2 minutes, and then reconnects and starts working again.
We have checked the server is sending keep-alive, and we have checked our server logs and see no difference between the working 6.0 case and the failing 6.1 case. We have also reviewed all the applicable (visible) diffs for libraries/protocols/HTTP_client and WICED/network (particularly NetX) and have not found any changes that seem related.
Has anyone else encountered (and hopefully debugged) this issue?
I should add that we've checked and the issue is still there in SDK v6.2.x.
We have escalated this issue internally for support.