[OTA] Device no longer found by Arduino IDE 2.5.0 (Beta 1)

[Swiftnesses]

Basic Infos

• This issue complies with the issue POLICY doc.
• I have read the documentation at readthedocs and the issue is not addressed there.
• I have tested that the issue is present in current master branch (aka latest git).
• I have searched the issue tracker for a similar issue.
• If there is a stack dump, I have decoded it.
• I have filled out all fields below.

Platform

• Hardware: [NodeMCU V1.0]
• Core Version: [Latest Arduino IDE build]
• Development Env: [Arduino IDE]
• Operating System: [MacOS]

Settings in IDE

• Module: [NodeMCU V1.0]
• Flash Mode: [qio|dio|other]
• Flash Size: [4MB (No SPIFFS)]
• lwip Variant: [v2 Higher Bandwidth]
• CPU Frequency: [80Mhz|160MHz]
• Upload Using: [OTA]

Problem Description

Following the update to 2.5.0 (Beta 1), devices are no longer found by Arduino IDE (1.8.9).

Restarted IDE, no change. Reverting back to 2.4.2 (full flash erase via OTA), devices show within 10 seconds. Note: the valve is remote to the house, so I have not tried a serial flash (I will, once the weather improves!).

Sketch contains the following ESP parameters:

  // Prevent wifi sleeping
  WiFi.setSleepMode(WIFI_NONE_SLEEP);
  // 802.11n
  WiFi.setPhyMode(WIFI_PHY_MODE_11N);
  // Station mode
  WiFi.mode(WIFI_STA);

[pfeerick]

I can reproduce this with a Wemos D1 mini on Windows using Arduino IDE 1.8.7 and the latest 2.5.0 Beta1 ESP8266 core. OTA on 2.4.2 was working perfectly, but the device is no longer visible with 2.5.0 Beta1. This is with lwIP 1.4 Higher Bandwidth. Re-installing 2.4.2 and re-compiling/re-uploading the code (erasing only the sketch) resulted in the device being visible for OTA within a few seconds.

[devyte]

@Swiftnesses @pfeerick
ArduinoOTA relies on MDNS internally. In order for MDNS to work, you have to call MDNS.update() in the loop. This was a requirement in the old implementation as well, but in that case it was commonly omitted, and things got weird in some cases depending on network traffic. Or not.
Please let me know if calling MDNS.update() in the loop fixes this.

[pfeerick]

I was going to say that this didn't make any difference for me - but some more testing has shown that adding the MDNS.update() call does make the device start advertising again. Core v2.4.2 OTA was always visible and working regardless of the presence of a MDNS.update() call in the `loop(). When Core 2.5.0 Beta1 didn't advertise,as long as I didn't close the Arduino IDE, it seemed to be sticky and still responded to OTA updates.

Two observations - a quick grep of the ESP8266 libriries doesn't show a single call to MDNS.update in even the MDNS examples except for the newLEAmDNS examples that have been added post 2.4.2. Also, MDNS isn't highlighted as a keyword anymore with the 2.5.0 Beta1 core??? If going forward MDNS.update() is needed, all the OTA examples etc need updating, as they don't use it. And if it was technically needed before, that's a bit... unsettling... as the built-in examples weren't indicating/documenting that!

Edit: For consistency, here is the code I used to test this - it's a modified version of the BasicOTA sketch with a few extra serial outputs, etc. I just commented/uncommented line 79 to test 2.4.2/2.5.0Beta1 OTA functionality. https://gist.github.com/pfeerick/bef13f437b16e11b427bbb27416e4194

[devyte]

@pfeerick MDNS.update() has always been needed. It's just that, given certain conditions that have nothing to do with the ESP, the legacy implementation would still work without it.
Now, MDNS.update() is mandatory, or it won't work. No in-between half-sometimes-maybe working state.
About the examples, you're correct, and art of the problem was that nobody knew the need of calling update() and that the examples weren't quite correct.
I'll keep this open to track fixing them.

[Swiftnesses]

@devyte Thanks for helping us understand this.

What I don't understand is, none of my OTA sketches used to include (even though they're part of the examples):

#include <ESP8266mDNS.h>
#include <WiFiUdp.h>

They still worked without issue (very well in fact), how is this possible? Trying to get my head around why it worked before and why I need to add 2 new libraries...?