RFC: Use invokelatest in jl_Function_call

[ihnorton]

Fixes JuliaPy/pyjulia#105

I thought it would also fix #343, but the example in that issue description seems to work fine for me on 0.6 release without this PR (but the issue is still open).

Directly calls Core._apply_latest with a closure over kw args as in JuliaLang/julia#22646 for 0.6.

cc @TsurHerman

[Keno]

So @vtjnash should correct me here, but I think this is the wrong place to do it. Instead, pyjulia should update the worldcounter in the ptls whenever it has imported some code. I'm just not sure if there's currently a C Api for that.

[Keno]

typo

[Keno]

On the other hand, I also don't really see a downside to doing this here.

[Keno]

This package still works on 0.5, so you need a check here e.g. isdefined(Base, :invokelatest).

[vtjnash]

jl_Function_call is already wrapped in a trampoline (via cfunction) which does the equivalent of invokelatest, so this shouldn't be necessarily (unless JuliaLang/julia#20167 isn't doing its job correctly)

[ihnorton]

Updated.

@vtjnash I don't know whether that PR is doing its job, but there's a simple repro of the issue this solves, here: JuliaPy/pyjulia#105 (comment)

[codecov-io]

Codecov Report

Merging #416 into master will increase coverage by 0.17%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #416      +/-   ##
==========================================
+ Coverage   66.04%   66.22%   +0.17%     
==========================================
  Files          17       17              
  Lines        1511     1516       +5     
==========================================
+ Hits          998     1004       +6     
+ Misses        513      512       -1

Impacted Files	Coverage Δ
src/callback.jl	95.23% <100%> (+3.34%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update bde073b...df1d16c. Read the comment docs.

[Keno]

Ok, so the underlying problem here will be addressed by JuliaLang/julia#22987. However, we should probably still merge this here to fix this on current versions of 0.6. Once base is merged, we can put in an appropriate version check.

[stevengj]

Thanks!