improve type stability with StaticArrays

[mapclyps]

Replace an anonymous function with Base.Fix1 for type stability in some cases, see #639

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (d300209) 89.65% compared to head (51d3d65) 87.28%.

❗ Current head 51d3d65 differs from pull request most recent head 02beae6. Consider uploading reports for the commit 02beae6 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #640      +/-   ##
==========================================
- Coverage   89.65%   87.28%   -2.38%     
==========================================
  Files          11       10       -1     
  Lines         967      912      -55     
==========================================
- Hits          867      796      -71     
- Misses        100      116      +16     

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

[devmotion]

Thank you! Could you add a test for #639?

[mapclyps]

I tried to replicate the type instability in the tests and put a test at the end of JacobianTest.jl like

ff(x) = sin.(x)
ssx = SA[1.0, 2.0]
ssy = ff(ssx)
result_store = DiffResults.JacobianResult(ssx)
@test begin
    result = (@inferred ForwardDiff.jacobian!(result_store, ff, ssx)) #the actual test
    jac = ForwardDiff.jacobian(ff, ssx)
    expected = DiffResults.ImmutableDiffResult(ssy, (jac,)) 
    result == expected
end

But to my surprise, there was no type instability with this test also on master.

However, just spinning up a REPL and executing this throws the error as expected and reported in #639.
Then I just moved this test to the top of the JacobianTest.jl file, and then it also threw the error on master (so it was working correctly as a test for the fix for #639). But it made me wonder what is actually going on there... and where this "hidden state" comes from.

Or is there a better, more stable way to test it?