Potential fix for Arm Linux testing.

[EricEngle-NOAA]

Potential fix for #560

[edwardhartnett]

We don't want this define.

Is there a way to detect whatever is happening on ARM and handle it without having platform specific defines?

[EricEngle-NOAA]

@edwardhartnett After thinking about this, and looking at the output, it is not writing in accordance with the GRIB2 standard for negative values. So 63116bb despite passing the checks is not correct. Continuing to work on this...

[EricEngle-NOAA]

If this is an acceptable PR, can we issue g2c v2.1.1?

[edwardhartnett]

If we can agree on a fix, we can merge it and do a release. But I don't want to reintroduce machine-specific macros to g2c.

So what is the condition on ARM that is being detected here?

[EricEngle-NOAA]

@edwardhartnett the most recent change, f51770c, removes architecture/platform specific defines.

In digging deeper, on ARM Linux the default char is unsigned char where on every other platform, including Apple silicon, default char is signed char.

[edwardhartnett]

OK, that's a lot better.

Can we have some testing for this code? Perhaps a test which fails with the old code but succeeds with the new code?

[EricEngle-NOAA]

I am bit confused with this. What do you mean by old code? It would not exist. Do you mean forcing a failure in the tests?

[edwardhartnett]

Yes, I mean a test which passes with your changes, but which would fail without your changes.

[EricEngle-NOAA]

Reset my develop branch. Will open a new PR.