Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 68.96%. Comparing base (4b83beb) to head (4644b62).
Report is 33 commits behind head on master.

@@             Coverage Diff             @@
##           master     #298       +/-   ##
===========================================
+ Coverage   38.24%   68.96%   +30.72%     
===========================================
  Files         233      233               
  Lines        6139     6142        +3     
  Branches     1607     1607               
===========================================
+ Hits         2348     4236     +1888     
+ Misses       3394     1596     -1798     
+ Partials      397      310       -87     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Just thinking out aloud, if the musl libm passes all these tests - would it be best to pull out the libm from musl into a standalone library and update openlibm to that?

pull out the libm from musl into a standalone library
and update openlibm to that?

I like this idea.
Moving a step further, we could take the libm related tests from libc-testsuite (for musl)
and put them in a standalone project, maybe call it openlibm-test.
Instead of adding all these things to openlibm.

Then, we have:

• openlibm == musl fork
• openlibm-test == libc-testsuite fork

@ViralBShah
Another thing I am thinking about is after julia removes its dependency on openlibm (and then links to the system libm by default), would it be more useful to have a libm written by julia than to continue to improve the precision of openlibm?

It's a good idea to have a separate repo for the tests. Julia almost has no dependency on openlibm, except in a few cases where LLVM on win32 needs it for a couple of things.

If the tests are in a separate repo, we can recreate a libm from Julia and then use the testsuite to test the Julia implementation.