ESP32DevelopmentBoardSecure example has LWMQTT_BUFFER_TOO_SHORT error

[BrianSidebotham]

The ESP32DevelopmentBoardSecure example sketch for the ESP32 fails to receive the first message after connect and sits in an endless re-connecting loop.

Inserting the getLastError() call before the re-connect shows an error of -1 (LWMQTT_BUFFER_TOO_SHORT) - in order to get the example to run I have to initialise the client with a much larger buffer: MQTTClient client(1024);

If the buffer is smaller than this, the example doesn't work for me.

The first message I receive is always bonkers accordng to the serial output:

checking wifi....
connecting...
connected!
incoming: /hello - 01234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789
incoming: /hello - world
incoming: /hello - world

[256dpi]

The first message you see is from someone that published a retained message in the "try" namespace on shiftr.io. Sure, not an ideal experience, but theres nothing we can do about it... 🤷‍♂️

[BrianSidebotham]

Perhaps we can immediately publish a blank retained message (The topic can only have a single retained message) and that should prevent us from receiving the "bad" retained message that requires a large buffer.

I think a simple client.publish("/hello", "", true); should make the example code more reliable for people.

I can do a PR if you think this is acceptable?

[256dpi]

Yes, let's do that. The client.publish("/hello", "", 1, true) should be placed right before the subscribe to ensure the retained message has been reset before subscribing. Also a comment should explain why it is needed and that the program may still crash if the "adversary" is online and sending. Would you be able to update all examples?

[BrianSidebotham]

Sure, no problem.

[256dpi]

Closing in favour of #199.