Uninitialised Variable in EAGLE Cooling

This fixes an uninitialised variable in the EAGLE cooling and fixes the SIMD expf functions for __APPLE__ platforms with the clang compiler. Perhaps there's a better way of doing this?

Also in this MR there is a missed include in the EAGLE star formation.

SIMD? The exp10 we use here has nothing to do with vectors.

It's a GNU extension to C however. So if a compiler declares __GNUC__ it ought to implement it.

Sorry, the implementations that we have are not SIMD ones, no, just that the ones that apple clang provides also have SIMD variants.

Technically they do implement the functions. I think the reason that they are hidden is because when using the apple headers for building iOS and macOS applications they have their own implementation, hence they're hidden behind __.

added 1 commit

• a4e08c13 - Update exp10.h

Compare with previous version

That does compile on our regular systems so I guess you are the only judge of whether this helps your unorthodox fruit-based system.

However, I am confused by the comments. If clang does define __GNUC__ (as the comment states) then shouldn't the OSX-specific block be inside the main GNUC block?

The __GNUC__ block is !defined(__GNUC__) so that is not reached when we use the clang provided by Apple, hence why we need the lower block.

This is so ugly I think we need to look at moving the logic into the place where all the ugliness resides, namely configure. We could test for exp10 and __exp10 and set some appropriate defines.

Can you try the apple-yuk branch and check that works for you.

BTW, note that in the macros I use the function-like form with an argument, that is the best way to define replacements like this, otherwise you can match other tokens.

If the branch is good recover the other fixes in this branch for another MR, or fix this one up.

Thanks Peter :). I'll implement the changes in this branch as there are a couple of other exp10 related fixes here.

Are there supposed to be double closing braces at the end of these lines (e.g. #if !defined(HAVE_EXP10) && !defined(HAVE___EXP10))) or is that a mistake?

added 3 commits

• 6bc056d5 - Check for exp10 and exp10f math library functions
• 7f5cdd5b - Fixed merge
• 04011bda - Fixed a botched merge

Compare with previous version

This branch now compiles fine on my machine and includes the changes from apple-yuk.

Thanks for fixing the parenthesis mismatch, strange it all seemed to work for me. Hah, sorry I didn't check for when this was used, need to enable cooling or star-formation...

Accepted, let's see what Jenkins makes of it next.

Was this merged in?

I meant to, but was interrupted at the critical moment. It refused to squash the commits, the history must be quite confusing.

merged

mentioned in commit 42801392

Done now, without a squash.

Uh-oh, the formatting script is somehow unhappy...

Didn't like your comment in the cooling change. Fixed now.