use oneunit(T) in construct_seeds

[longemen3000]

should fix #651 ?

[longemen3000]

the invalidations tests has an unrelated error

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (16417ec) 89.65% compared to head (59c9cd1) 89.65%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #652   +/-   ##
=======================================
  Coverage   89.65%   89.65%           
=======================================
  Files          11       11           
  Lines         967      967           
=======================================
  Hits          867      867           
  Misses        100      100           

Files Changed	Coverage Δ
src/partials.jl	84.21% <100.00%> (ø)

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

[mcabbott]

This seems fine. Is it easy to add a test, e.g. loading the smallest / slowest-changing package which defines a number for which one != oneunit?

[longemen3000]

Measurements.jl? and an unary function.
Binary functions require defining some interactions between ForwardDiff.Dual and Measurements.Measurement (AMBIGUOUS_TYPES), (i had to define my own in https://github.com/longemen3000/ForwardDiffOverMeasurements.jl ). In the julia 1.9 world, that integration is a prime candidate for extensions.

Another possible candidate for this is Unitful.jl. if we see #328 , the error posted there is caused by the different types of the tuple (one(T),zero(T)

julia> g(x)
ERROR: MethodError: no method matching ForwardDiff.Partials(::Tuple{Float64,Quantity{Float64,Unitful.Dimensions{()},Unitful.FreeUnits{(),Unitful.Dimensions{()}}}})
Closest candidates are:
  ForwardDiff.Partials(::Tuple{Vararg{V,N}}) where {N, V} at /home/cst-jean/.julia/packages/ForwardDiff/hnKaN/src/partials.jl:2
Stacktrace:

and Unitful.jl defines one and oneunit in the following lines:
https://github.com/PainterQubits/Unitful.jl/blob/3798720d77e09a8c0c994db1e142848545a27ba3/src/quantities.jl#L425-L428

[longemen3000]

I added Measurements.jl as an example package that has oneunit(T) != one(T).

The invalidation tests fails with the stat_rep(::Nothing) method error, and there are errors on nightly, that i don't know if are unrelated.

Edit: it seems that the errors in nightly and in the invalidations tests are unrelated

[longemen3000]

replaced by #659