Handle case in EnzymeTestUtils of reduction over empties

[kshyatt]

This keeps screwing up my life!!!!

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic main) to apply these changes.

Click here to view the suggested changes.

diff --git a/lib/EnzymeTestUtils/src/finite_difference_calls.jl b/lib/EnzymeTestUtils/src/finite_difference_calls.jl
index 87d6f707..11f89564 100644
--- a/lib/EnzymeTestUtils/src/finite_difference_calls.jl
+++ b/lib/EnzymeTestUtils/src/finite_difference_calls.jl
@@ -71,7 +71,7 @@ function j′vp(fdm, f_vec, ȳ, x)
         end
     end
   end
-  mat = isempty(ẏs) ? similar(x, 0) : transpose(reduce(hcat, ẏs))
+    mat = isempty(ẏs) ? similar(x, 0) : transpose(reduce(hcat, ẏs))
   result = zero(x)
   for i in 1:length(ȳ)
     tp = @inbounds ȳ[i] 

[github-actions]

Benchmark Results

	main	b283e0c...	main / b283e0c...
basics/make_zero/namedtuple	0.0519 ± 0.0016 μs	0.0521 ± 0.0012 μs	0.995 ± 0.038
basics/make_zero/struct	0.264 ± 0.0054 μs	0.265 ± 0.0058 μs	0.996 ± 0.03
basics/overhead	5.87 ± 0.92 ns	4.33 ± 0.01 ns	1.36 ± 0.21
basics/remake_zero!/namedtuple	0.23 ± 0.0091 μs	0.231 ± 0.0078 μs	0.996 ± 0.052
basics/remake_zero!/struct	0.236 ± 0.0071 μs	0.225 ± 0.007 μs	1.05 ± 0.045
fold_broadcast/multidim_sum_bcast/1D	10.2 ± 0.22 μs	10.2 ± 0.26 μs	1 ± 0.034
fold_broadcast/multidim_sum_bcast/2D	12.1 ± 0.22 μs	12.1 ± 0.23 μs	1 ± 0.026
time_to_load	1.02 ± 0.0042 s	1.03 ± 0.0064 s	0.991 ± 0.0074

Benchmark Plots

A plot of the benchmark results has been uploaded as an artifact at https://github.com/EnzymeAD/Enzyme.jl/actions/runs/26869987673/artifacts/7378332384.

[wsmoses]

add a test?

[kshyatt]

add a test?

I'd love to but it's kind of annoying because this function is only tested implicitly by other stuff... I'll see if I can think of a good way to do it. Or do you mind if I just call it directly in a test?

[wsmoses]

basic gist is that if you hit this in your real code, if we don't test it -- there's a chance the functionality is reverted accidentally. so anything that exercises that

[kshyatt]

Yeah I'm aware of how tests work lol, I'm just trying to find a good case because clearly this is a weird edge scenario

[wsmoses]

ah yeah fair

[kshyatt]

Need to engage my "goofball test case generator" powers again...