ARDUINO=100 always added as define to g++

[jgfoster]

cpp_library.rb:327 hard-codes -DARDUINO=100 on command line

• OS: macOS
• ruby 2.6.3p62 (2019-04-16 revision 67580) [universal.x86_64-darwin19]
• Bundler version 1.17.2
• arduino_ci (0.3.0)
• Apple clang version 12.0.0 (clang-1200.0.32.2)
• Arduino IDE version: 1.8.13
• java version "11.0.7" 2020-04-14 LTS

.arduino-ci.yaml:

platforms:
  mega2560:
    board: arduino:avr:mega:cpu=atmega2560
    package: arduino:avr
    gcc:
      features:
      defines:
        - __AVR_ATmega2560__
        # added this line
        - ARDUINO_CI
        - ARDUINO=10813
      warnings:
      flags:

unittest:
  platforms:
      - mega2560

compile:
  platforms:
    - mega2560

Warning

In file included from <built-in>:399:
<command line>:4:9: warning: 'ARDUINO' macro redefined [-Wmacro-redefined]
#define ARDUINO 10813
        ^
<command line>:1:9: note: previous definition is here
#define ARDUINO 100
        ^
1 warning generated.

Arduino Architecture(s) Affected: Mega2560

[ianfixes]

I'm not clear on why ARDUINO would need to be defined in mega2650, which is a statement of genuine ignorance; I set it to 100 based on this forum.arduion.cc post, which says:

Why is (ARDUINO) && ARDUINO >= 100 needed?

It's needed because early versions (pre 1.0) had a Serial method "print" which allowed binary when used with the qualifier BYTE, but this has now been superseded by the "write" method.

In other words, I was under the impression that this was used by the IDE and not by a particular board. Can someone enlighten me on the proper use of this #define?

[jgfoster]

I'll look for the specifics, but there was some change to Arduino core library at 1.0.6 and an external library had #if ARDUINO < 106 to take different code paths. When it runs in the Arduino IDE it uses 10813 and uses the correct code. When it runs in the mocks library is uses the wrong code and fails.

[per1234]

I was under the impression that this was used by the IDE and not by a particular board. Can someone enlighten me on the proper use of this #define?

It's not used by the IDE, it's defined by the IDE. The ARDUINO macro serves two purposes:

---

Indicate that the code is being compiled by the Arduino build system. So you can do things like this:

#if defined(ARDUINO)
#include <Arduino.h>
#elif defined(PARTICLE)
#include <application.h>
#else
#error unsupported build system
#endif

This particular code is used to deal with the fact that Particle uses the Arduino core API, meaning that most Arduino code can be used with Particle, and vice versa, but Particle changed the core library's primary header filename from "Arduino.h" to "application.h".

---

Indicate which version of the Arduino IDE is being used for the compilation, so you can do things like this:

#if defined(ARDUINO) && ARDUINO >= 100
#include <Arduino.h>
#else
#include <WProgram.h>
#endif

This code is used to deal with Arduino changing the core library's primary header filename from "WProgram.h" (an artifact of Arduino's origin as Wiring) to "Arduino.h" in the Arduino IDE 1.0.0 release.

More information:
https://arduino.github.io/arduino-cli/latest/platform-specification/#global-predefined-properties:~:text=%7Bruntime.ide.version%7D%3A

[jgfoster]

This is where I ran into the problem (line 90):

#elif ARDUINO > 106 || TEENSYDUINO > 121

And, just to clarify, it isn't needed by a particular board. This was just the showing how I came to the warning (and I happen to be using the Mega2560). It could have been any board that supports Ethernet. It was just a demo that if, for whatever reason, you wanted to provide your own value for ARDUINO, then you would get a warning. Once I got into it it seemed that #179 is a reasonable fix.

[ianfixes]

[ARDUINO=100 is] not used by the IDE, it's defined by the IDE

Sorry for my poor wording there, I meant to say that the definition is associated with the IDE as a whole and not individual boards. i.e. the IDE "uses" the ARDUINO variable (setting it to 100) as a signal to the code it's compiling, and the boards do not do this.

Based on what I'm hearing here, it sounds like it's perfectly appropriate for me to be defining ARDUINO=100 in CppLIbrary, since I am indeed attempting to "Indicate that the code is being compiled by the Arduino build system".

Worrying about the ARDUINO > 106 case is going to be its own can of worms, and we should probably open a separate issue to track what that feature might involve.

(note to future self: on one hand, it could involve iterating the entire CI process over a set of possible values like 100, 106, etc. On the other hand, the companion IDE performing the CI tests will be only one version, and we could talk about whether proper Arduino library development means keeping current with one's IDE version.)

I've been hesitant to merge #179 because it breaks my fundamental assumption that there is only one Arduino IDE version for me to install and imitate (thus, whether the right value is 100 or 106 I can hard-code it because I'm hard-coding the IDE.) Am I thinking about this the wrong way?

[jgfoster]

Actually, I think you are right to be concerned. I brought in the current version of Ethernet and found that it didn't work. My work-around might be wrong. It would probably have been better if I brought in an old version of Ethernet to match ARDUINO=100. But, my preference would be for Arduino CI to be more current (1.6 if not 1.8.13). If you were going to move forward, then I'd be helped; if you don't, then I should go look for an old version of Ethernet.

My preference for a current version of ARDUINO is that I'm writing production code and will be deploying on 1.8.13 (or later).

[ianfixes]

I think the right course of action for me is to resolve #185 before anything else

[jgfoster]

I agree that #185 is better.