Linear BIG_CONST macros and try lcov-excl

[ckormanyos]

The purpose of this PR is to streamline/linearize macros sich as BOOST_MATH_BIG_CONSTANT and subsequently check if line-by-line lcov-exclude makes sense.

[codecov]

Codecov Report

Attention: 8 lines in your changes are missing coverage. Please review.

Comparison is base (61a03e0) 88.61% compared to head (12ee031) 89.82%.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1086      +/-   ##
===========================================
+ Coverage    88.61%   89.82%   +1.21%     
===========================================
  Files          768      768              
  Lines        64669    63722     -947     
===========================================
- Hits         57306    57241      -65     
+ Misses        7363     6481     -882     

Files	Coverage Δ
include/boost/math/quadrature/gauss.hpp	93.04% <ø> (ø)
include/boost/math/quadrature/gauss_kronrod.hpp	95.45% <ø> (ø)
...de/boost/math/special_functions/detail/erf_inv.hpp	95.04% <ø> (ø)
...ost/math/special_functions/detail/igamma_large.hpp	97.77% <ø> (+64.66%)	⬆️
...ost/math/special_functions/detail/lgamma_small.hpp	91.72% <ø> (+32.36%)	⬆️
include/boost/math/special_functions/expm1.hpp	76.92% <100.00%> (+11.32%)	⬆️
include/boost/math/special_functions/lanczos.hpp	87.86% <ø> (+50.92%)	⬆️
include/boost/math/tools/big_constant.hpp	100.00% <ø> (ø)
test/test_1F1.hpp	100.00% <ø> (ø)
test/test_gamma_edge.cpp	100.00% <100.00%> (ø)
... and 1 more

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 61a03e0...12ee031. Read the comment docs.

[ckormanyos]

Hi @jzmaddock and @mborland and @NAThompson this work might seem ugly/controversial/hurried. But I looked into simplifying the big-constant macros such that they can be neatly expressed as single-line macros. This allows for simple line-by-line LCOV_EXCL_LINE on lines that (should be) rather well-known to be covered but which lcov seems to be missing (maybe due to optimization settings).

Anyway, it is a method, but might be controversial. Comments? Thoughts?

[ckormanyos]

I've tried (but not yet succeeded) to directly put the code annotation into the macros themselves.

[ckormanyos]

One thing I did notice is that when the couple hundred annotations are in place, it really draws focus to the several hundred remaining specfun missed lines that are actually much more, let's say, interesting and content-rich. So feedback would be appreciated, ...

Cc: @jzmaddock and @mborland and @NAThompson

[mborland]

Looks good to me. 200+ lines removed does help.

[ckormanyos]

Looks good to me. 200+ lines removed does help.

Thanks Matt. It ended up being closer to 1,000 lines. I could not find any other large, missed tables in specfun. But some were, in fact, covered. I annotated the ones I found must have been covered but were seemingly mistakenly marked by lcov as uncovered.

[ckormanyos]

Here are the remaining trouble spots:

• specfun
• distributions
• octonion.hpp

This is a really great achievement so far in Math!

[mborland]

Here are the remaining trouble spots:

• octonion.hpp

The one test file doesn't have real testing in it: https://github.com/boostorg/math/blob/develop/test/octonion_test.cpp.

[ckormanyos]

The one test file doesn't have real testing in it:

I suppose we might as well just exclude it from the lcov analysis at the command_line-level?

[ckormanyos]

The one test file doesn't have real testing in it:

I suppose we might as well just exclude it from the lcov analysis at the command_line-level?

On second thought, that is too hasty. This is our one-time chance to actually get some octernion testing. Or notice that they fail. I think we should try for tests. I can give this one a crack in another branch later, ...

[jzmaddock]

Forgive me, late night last night... but BOOST_MATH_BIG_CONSTANT is now being called with 2 arguments, but still defined with 3? I know I'm missing something... !

BTW the basic issue is that multi-line statements are badly handled by lcov, so there are quite a few function calls that look like:

foo(
   a, 
   b,
   c);

especially when the arguments are long error messages... and these are marked as uncovered except for the last line. The issue with the arrays are the same, it's not the macros as such that cause the issue, just that we have dynamic initialization spread over several lines. For the functions I've been collapsing down to one line where possible, and using LCOV_EXCL_LINE when not. For the arrays, I've topped and tailed the list of initializers with LCOV_EXCL_START/LCOV_EXCL_STOP.

BTW I have just started on coverage for beta, hoping to get to erf or so in the next tranche, I am going to be super busy this week though unfortunately :(

[ckormanyos]

Hi John, thanks for those explanations.

but BOOST_MATH_BIG_CONSTANT is now being called with 2 arguments, but still defined with 3?

Not exaclty, but there are such changes in this PR. That's me, messing with stuff.

I left the number of parameters of BOOST_MATH_BIG_CONSTANTto be 3. And it is still called in the PR with 3.

The other mega-macro, BOOST_MATH_HUGE_CONSTANT originally had 3 parameters, but the digits parameter was used nowhere. So I removed it.

For the arrays, I've topped and tailed the list of initializers with LCOV_EXCL_START/LCOV_EXCL_STOP

I can do that and eliminate the thousand or so single-line annotations in this PR. I did, however, diligently look if I felt that they were covered prior to x-ing them out. So I'm good with that.

Also, I kind of like the way I refactored BOOST_MATH_BIG_CONSTANT and BOOST_MATH_HUGE_CONSTANT, as I think the intermediary inline function add clarity. Sorry about silently going from 3 to 2 params on _HUGE, but the unused param was in my mind simply confusing.

[ckormanyos]

oops. I did mess up some of the calls to BOOST_MATH_BIG_CONSTANT. My bad. I will repair all these.

[ckormanyos]

oops. I did mess up some of the calls to BOOST_MATH_BIG_CONSTANT. My bad. I will repair all these.

I'll finish this up (whatever I started) but just let it sit there, as it seems like it was a bit hurried.

@jzmaddock if you want some, none, or all of it, pls let me know. It was/is a good exercise.

[ckormanyos]

I'll finish this up (whatever I started) but just let it sit there, as it seems like it was a bit hurried.

I do, in fact, like the single-line expression(s) of BOOST_MATH_BIG_CONSTANT/BOOST_MATH_HUGE_CONSTANT. This allows these to be lcov-excluded on single-line-annotations.

The three-parameters of _HUGE are controversial. But if any part of this is cool, it's that work. Or any part of it.

Anyway, I'll let this one bake for a while.

[ckormanyos]

I converted this PR to a draft. I'm not going to really dive into this any time soon, but I synchronized with Develop and it now shows some of the tabble/macro work that subjectively seemed to make sense to me.