I am trying to debug why my IOT devices don't always recover from a DNS/DHCP failure – it is quite hard as they simply get stuck, not doing anything until rebooted.
I'll go over what I am currently doing to try and shed some light on the possible causes.

Preparing for disaster

Only one of the devices I have currently installed is accessible from the serial debug port – the rest are carefully hidden inside walls, and while I was confident that my changes to the firmware would work (as they work on a 'blank' device) I wanted to be doubly sure and avoid the pain of re-wiring everything.

The most accessible ESP8266 is connected to my Odroid N2 via USB, but this Odroid is running CoreElec, a very stripped down linux distribution for media which has no package manager.

Usually the tool I choose for serial debugging is minicom so I set on to cross-compile it from my x86_64 box so that it'd run on the aarch64 Odroid.

Compiling minicom

minicom $ ./configure --host=aarch64 --build=x86_64 CC='arm-linux-gnueabihf-gcc-8'
minicom $ make -j8
ld: minicom/src/window.c:196: undefined reference to `tputs'
ld: minicom/src/window.c:1954: undefined reference to `tgetstr'
ld: minicom/src/window.c:1967: undefined reference to `tgetnum'

The complaints about tputs point to missing ncurses references, so off to build ncurses.

ncurses $ ./configure --host=aarch64 --build=x86_64 CC='arm-linux-gnueabihf-gcc-8'
ncurses $ make -j8

Linking the newly-built ncurses

minicom $ ./configure --host=aarch64 --build=x86_64 CC='arm-linux-gnueabihf-gcc-8 -L~/git/ncurses/lib'
minicom $ make -j8
minicom $ file src/minicom
src/minicom: ELF 32-bit LSB shared object, ARM, EABI5 version 1 (SYSV), dynamically linked, interpreter /lib/ld-linux-armhf.so.3, BuildID[sha1]=4a85f2b5e2e7d0a48214743265505cc23b461a7e, for GNU/Linux 3.2.0, with debug_info, not stripped

Copy the binary over aaand:

root@odroid $  ./minicom -D /dev/ttyUSB0
No termcap database present!

Checking /etc/terminfo/README in my debian machine has all the info I need

This directory is for system-local terminfo descriptions. By default,
ncurses will search ${HOME}/.terminfo first, then /etc/terminfo (this
directory), then /lib/terminfo, and last not least /usr/share/terminfo.

so let's retry after copying over /usr/share/terminfo to ~/.terminfo

root@odroid $ ./minicom -D /dev/ttyUSB0
No termcap entry for xterm

Let's just override that with something present in ~/.terminfo:

root@odroid $ ./minicom -D /dev/ttyUSB0 -t rxvt-256color
Welcome to minicom 2.6.2

OPTIONS: I18n
Compiled on Jul  6 2019, 10:27:35.
Port /dev/ttyUSB0, 15:39:48

Press CTRL-A Z for help on special keys
^D
{r�#5 ets_task(40100130, 3, 3fff83ec, 4)
LOCAL-Connecting to Wi-Fi... SSID
LOCAL-Waiting wifi...
Connected to SSID!
Subscribing to b'RFPOWER/set/#'
Subscribing to b'HDMI/set/#'

We have an emergency avenue into this device!

Logging

By the very nature of these devices, it is hard to get information out of them. They have no screens, and although they do have a single LED on which they could show some pattern, they are usually inside the walls.

Adding to these charateristics, there's hardly any guarantee on when we'll see a failure – and given there's no local storage, the best that can be done is to try and output something useful on the network.

I implemented this with a very basic UDP server and a pipe-delimited message format NAME|CONTENTS.
After updating the common.py and making some small tweaks to the firmware on all devices, the data received on the logging server looks something like:

During normal operation:

$ ./logserver.py
2019-07-06 12:03:29,101 - INFO - [RFPOWER]: Topic: HDMI/set
2019-07-06 12:03:29,102 - INFO - [RFPOWER]: Msg: 1
2019-07-06 12:03:58,672 - INFO - [RFPOWER]: Topic: RFPOWER/set/2
2019-07-06 12:03:58,676 - INFO - [RFPOWER]: Msg: 0
2019-07-06 12:03:59,109 - INFO - [RFPOWER]: Publish 0 to b'RFPOWER/state/2'

For OTA:

$ ./logserver.py
2019-07-06 12:00:02,062 - [RFPOWER]: Receiving OTA update..
2019-07-06 12:00:02,074 - [RFPOWER]: Target IP: 192.168.2.189, Target Port: 1233, Local filename: common.py, hash: 3471ff7786fb04d064e555e6588ca3d74f6b111a
2019-07-06 12:00:02,518 - [RFPOWER]: renaming tmp to common.py
2019-07-06 12:00:02,719 - [RFPOWER]: restarting
2019-07-06 12:00:07,491 - [RFPOWER]: Connected to SSID!
2019-07-06 12:00:07,844 - [RFPOWER]: Subscribing to b'RFPOWER/set/#'
2019-07-06 12:00:07,970 - [RFPOWER]: Subscribing to b'HDMI/set/#'

Now we are ready to start developing more and more robust devices!