Fix enum redeclaration issue when using WiFi and Ethernet about arduino-pico

[rpineau]

Thanks (I did find this after digging in the code).
I'll wire my spare pin to the W5500 int_n and see if this helps with the speed (time to add that bodge wire ;) , it's not a proper prototype if there is not at least one bodge wire on it :) ).

[earlephilhower]

Looks like you found a bug and a solution. Want to make a PR?

[rpineau]

Yes.
I guess I need to fork this first, make my change and make a PR from it ? (as I don't have commit rights).

[JAndrassy]

sorry, but what error do you get?
there is no problem to include LwipIntfDev.h multiple times. there is #pragma once to guard this

and #ifndef only works for preprocessor macros, not for a regular identifier

cc: @earlephilhower

[earlephilhower]

I just grepped for that identifier and it's only defined in 1 spot in the entire tree. In addition to @JAndrassy 's comments that the suggested diff doesn't do what you think it does, it also would break the enum typename.

earle@amd:~/Arduino/hardware/pico/rp2040$ grep -r EthernetLinkStatus
libraries/lwIP_Ethernet/src/LwipIntfDev.h:enum EthernetLinkStatus {
libraries/lwIP_Ethernet/src/LwipIntfDev.h:    EthernetLinkStatus linkStatus();
libraries/lwIP_Ethernet/src/LwipIntfDev.h:EthernetLinkStatus LwipIntfDev<RawDev>::linkStatus() {

Closing for now, but if you have an MCVE showing how it fails post it here and we'll look.

[rpineau]

There is a #pragma once in it but if I mix it with the standard Ethernet library this doesn't prevent the redefinition.
Why do I use the standard Ethernet library with my W5500, because the code was already written before (by me) to use this for my project that I'm now porting to the pico W.

The diff above allows me to use the standard ethernet library, define __EthernetLinkStatus__ (Github mangled the codfe and removed the double underscore around EthernetLinkStatus) and then include WiFi.h without having EthernetLinkStatus redefine a 2nd time

[rpineau]

I spent the last few hours moving all my code to Wiznet5500lwIPc for the W5500 I use (the WiFi code is disable for now).
Wiznet5500lwIPc Ethernet works fine for standard TCP connectivity, but UDP begin return 1 (Ok):

discoveryServer = new WiFiUDP();
nErr = discoveryServer->begin(m_UDPPort);

but the discovery response are not working (exact same code).

When using the standard Ethernet library with EternetUDP, the discovery works.
When changing to WiFiUDP , it gets the broadcast packet and replies, but the discovery service doesn't recognise the response.
tcpdump doesn't really show any difference in the payload so it's probably related to the failure on begin
Looking at the code for begin in WiFiUdp.cpp it uses a IPAddress() call to get the ip to listen on, but there is no way to set this IP (or I haven't found it) and I don't want to listen on 0.0.0.0.
The constructor for WiFiServer does take an IP and port as parameters, the constructor for WiFiUDP takes no parameters.
I'm for now only testing with the Ethernet side of my project as I can't move to Wiznet5500lwIPc + WiFi if I'm not able to get parity with what I have working with the standard Ethernet library.
May be the real problem I'm facing is how do I tie a server to an interface (on Linux or macOS I would bind to the respective interfaces IP to run services on different IP or 0.0.0.0 for all interfaces)

[rpineau]

So after a lot more time spend on this, turn out the code works, but the lwIP_w5500 library is more than 20 times slower than the standard Arduino Ethernet library with a W5500.
The code implements a simple REST server and the exact same code returns a reponse in ~ 40ms with the standard Ethernet library versus slightly over a second with the lwIP_w5500 library. During the discovery after replying to the UDP broadcast, the service then query 3 of the API endpoint successively. As the lwIP_w5500 is slow to respond it fails the discovery. So the question is why is lwIP_w5500 so much slower ?

[JAndrassy]

did you wire the INT pin?

[rpineau]

No I didn't, but I can modify the PCB for this.
Which pin should be used as the interrupt pin ?
The only pin I have left available on the PICO is GP27.
And is this configurable in the lwIP_w5500 library (is there a code example that shows how to set the pin).
If I need to use a specific pin I can probably change the board design (this is all open source in case you wonder, it's for an astronomy dome controller).

[JAndrassy]

Wiznet5500lwIP Ethernet(CS_PIN, SPI, INT_PIN);

[rpineau]

So after wiring the INT_N pin of the W5500 chip (nothing like a bit of SMD soldering on a Sunday :) ) and declaring the ethernet object like this Wiznet5500lwIP domeEthernet(ETHERNET_CS, SPI, 27); , the query response time is now around 50ms and the discovery system now sees my board properly.
Thanks for all the help !

[earlephilhower]

Awesome work, thanks for the update!

FWIW, if you don't want to run a new PCB and have CPU cycles to spare you can run the default polled mode at a higher frequency. Add the function call before you start the Ethernet device and you can go from polling every 50ms all the way down to every 1-2ms. It'll eat more CPU cycles than IRQ mode, of course, but if you have existing devices and you're not fond of bodging it may give you enough performance.

[rpineau]

I'm ok modifying the PCB as I only made 2 prototypes for this round :

https://sbforum.s3.amazonaws.com/DomeEthernetRP2040.jpg

So for these 2, I just added the missing signal with a small wire and I have modified the PCB files to add it for the next prototype round.
I'd rather fix it the right way as using interrupt is so much more efficient than polling.
As this is for an open source / open hardware project , it's fine. It's not like I have to go fix a large number of existing board.

Thanks again for all the help.