Closed benlorenz closed 6 days ago
It did indeed fail:
julia> [representative(elliptic_parameter(Y2, f)) for f in fibers_in_Y2[4:6]]
- 3-element Vector{AbstractAlgebra.Generic.FracFieldElem{QQMPolyRingElem}}:
- (y//z)//((x//z)*t)
- ((y//z) + t^3)//((x//z)*t - t^4)
- ((y//z) + t^3)//((x//z) - t^3)
+ ERROR: AssertionError: found
+ Stacktrace:
+ [1] horizontal_decomposition(X::EllipticSurface{QQField, AbstractAlgebra.Generic.FracFieldElem{…}}, F::Vector{QQFieldElem})
+ @ Oscar ~/work/Oscar.jl/Oscar.jl/experimental/Schemes/src/elliptic_surface.jl:1349
+ [2] elliptic_parameter(X::EllipticSurface{QQField, AbstractAlgebra.Generic.FracFieldElem{QQPolyRingElem}}, F::Vector{QQFieldElem})
+ @ Oscar ~/work/Oscar.jl/Oscar.jl/experimental/Schemes/src/elliptic_surface.jl:1419
+ [3] (::var\"#23#24\")(f::Vector{QQFieldElem})
+ @ Main.__255 ./none:0
+ [4] iterate
+ @ ./generator.jl:47 [inlined]
+ [5] collect(itr::Base.Generator{Vector{Vector{QQFieldElem}}, var\"#23#24\"})
+ @ Base ./array.jl:834
+ [6] top-level scope
+ @ REPL[48]:1
+ Some type information was truncated. Use `show(err)` to see complete types.
And it added about 13 minutes to the booktest job. I guess we could afford this since it still finished before a bunch of other jobs in the same run but it is still quite a lot.
Thanks for the effort! I agree that 13 minutes is (too?) long for a regular test on every PR. Is there a way to run it once for every release?
In the sense that we have, tests, long tests, very long tests (but only very few ones)
Seems like we can speed up things by a factor of 50 ... so we will not need very long tests. Best wait before taking any further steps.
Since there will be no fix in the near future, I close it. Thanks @benlorenz for bringing this up.
Lets see how long it takes there and whether it succeeds, ref #3881, cc @simonbrandhorst.