Obtain more code coverage

[ckormanyos]

This PR obtains more code coverage. But it is for later in 1.91 (not for today's 1.90).

[codecov]

Codecov Report

Attention: Patch coverage is 99.89166% with 1 line in your changes missing coverage. Please review.

Project coverage is 96.2%. Comparing base (3d32b38) to head (ccb113d).

Files with missing lines	Patch %	Lines
include/boost/multiprecision/cpp_dec_float.hpp	99.7%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff            @@
##           develop    #711     +/-   ##
=========================================
+ Coverage     94.7%   96.2%   +1.5%     
=========================================
  Files          296     297      +1     
  Lines        30881   31178    +297     
=========================================
+ Hits         29233   29967    +734     
+ Misses        1648    1211    -437     

Files with missing lines	Coverage Δ
example/exercise_threading_log_agm.cpp	100.0% <100.0%> (ø)
include/boost/multiprecision/complex_adaptor.hpp	92.3% <100.0%> (+2.8%)	⬆️
include/boost/multiprecision/cpp_double_fp.hpp	100.0% <100.0%> (ø)
...nclude/boost/multiprecision/detail/default_ops.hpp	93.5% <100.0%> (+2.5%)	⬆️
...oost/multiprecision/detail/functions/constants.hpp	99.2% <100.0%> (ø)
...lude/boost/multiprecision/detail/functions/pow.hpp	99.4% <100.0%> (+4.7%)	⬆️
...ude/boost/multiprecision/detail/functions/trig.hpp	97.5% <ø> (+0.8%)	⬆️
test/git_issue_167.cpp	100.0% <100.0%> (+50.0%)	⬆️
test/git_issue_464.cpp	100.0% <100.0%> (ø)
test/test_arithmetic.hpp	99.7% <ø> (+0.4%)	⬆️
... and 16 more

... and 5 files 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 3d32b38...ccb113d. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ckormanyos]

Ok so I worked on coverage, mostly with focus on cpp_dec_float. That file is pretty well covered now, but missing 8 interesting lines. I'll let this one sit for a while.

I would like to improve the line coverage of the tests and then we can discuss which headers need more dedicated edge tests.

Let's just leave this as a draft until the action on 1.90 wraps up.

[ckormanyos]

In cpp_dec_float going to nearly $99\%$ coverage was straightforward. Sure it needed a few hundred lines of dedicated test code.

I only found 1 bug, but harmless, in the area of negative infinity ($-\infty$). There were several areas having redundant or unclear lines. I cleared up (or subjectively think I celared up) a bunch of these sequences. Nothing really critical was found, and this shows how well this thing was tested to begin with. I'm really impressed with the test-suite.

[ckormanyos]

Hi @jzmaddock and @mborland

Check it out, WE WILL NOT merge this until AFTER 1.89.

I have now marked this as ready for review. With this, I mean AFTER 1.89 gets out the door. But the scope of change is non-trivial.

So I saw a window of opportunity (timewise and energy-wise) to go for coverage. I got a solid $96.1\%$, which is good.

OK so there were some highlights.

• I found a small handful of bugs in cpp_dec_float and in some special functions edge-cases.
• I think that some infinity/NaN edge-cases in default_ops-pow were incorrect (also erfand erfc, maybe tgamma).
• In general I think these things are more right now than earlier, but there is work to be done.
• With this, I got the special-functions C99 thing to actually use signed-zero on cpp_bin_float to get lots of coverage lines. I think it is right-er than before, but not yet fully done.
• I leaned heavily into cpp_dec_float and got under 10 lines missing in coverage. I hit this one, changed quite a bit actually. But my confidence is high here.
• I did some trivial changes to complex_adaptor and its tests to gain lots of lines.
• Along those lines, I got cpp_dec_float working with complex_adaptor, which was previously not the case since some eval_whatever functions were actually missing from this backend.
• I got the test cases issue464 and issue167 actually working and contributing to coverage. These were simply exception-ing out previously.
• I found redundant code in a few places like powfunctions and streamlined them.

Well, that's about it.

[ckormanyos]

There is a lot of material within the scope of this PR. It might take a while to sift through it all and negotiate on the best (or better) fixes, as I might not have subjectively always chosen the right path.

But in general, I think this PR is intact. And we can start taking a serious look at it.