Added OTA checksum verification

[ghost]

See #172

Here is a little test script for triggering OTA updates with a checksum:

#!/bin/bash

DEVICE=homie/awesome-relay
BROKER=localhost
FW_VER=1.0.0
FW_BIN=/path/to/firmware.bin
FW_MD5=$(md5sum -b $FW_BIN | cut -d" " -f1)

mosquitto_pub -h $BROKER -t "$DEVICE/\$ota" -m "$FW_VER"
sleep 1
mosquitto_pub -h $BROKER -t "$DEVICE/\$implementation/ota/checksum" -m $FW_MD5
sleep 1
mosquitto_pub -h $BROKER -t "$DEVICE/\$implementation/ota/firmware" -f $FW_BIN

[marvinroger]

I would not unsubscribe from firmware and checksum0, otherwise you need, as you've done in your script, to wait a little bit before sending next messages. What do you think? Maybe you have a good reason to do so?

[ghost]

Thanks for your feedback. Staying subscribed is actually what I was preferring when I was poking around in the code. I only left dynamic subscribe/unsubscribe in because that's what you had chosen for payload (now firmware) before. I was wondering why you had the if (!_flaggedForOta) check in the code. It should never have triggered because Homie only subscribed to the payload topic when _flaggedForOta was also true. Did you observe a race condition or something? Anyways. I added a commit to remove dynamic subscribe/unsubscribe.

BTW, the delay in my little test script is just for me to be able to observe what's going on looking at two terminals. I don't think the delay is needed. A real implementation would obviously evaluate Homie's status responses.

[marvinroger]

Well we chose this because the payload was retained, so to receive it we had to subscribe only when receiving the $ota message. But at first the payload was not retained, and it looks like I forgot the flagged for OTA check. So yes, it was actually useless.

Great! LGTM 👍