Reduce tolerance for change in Julia v1.9 #861
Conversation
Codecov Report
@@ Coverage Diff @@
## master #861 +/- ##
=======================================
Coverage 93.79% 93.79%
=======================================
Files 43 43
Lines 9893 9893
=======================================
Hits 9279 9279
Misses 614 614 Continue to review full report in Codecov by Sentry.
|
|
I and fine with this change. Any idea why it is still failing on windows Julia v1 for this one power flow case? |
Nope. Iteration limit, so I guess some numerical issue? It seems like lots of nearby tests are commented out with numerical difficulties. |
| end | ||
| @testset "5-bus ext case" begin | ||
| data = PowerModels.parse_file("../test/data/matpower/case5_ext.m") | ||
| sm = calc_susceptance_matrix(data) | ||
|
|
||
| @test isa(sm, AdmittanceMatrix{Float64}) | ||
| @test SparseArrays.nnz(sm.matrix) == 17 | ||
| @test isapprox(LinearAlgebra.det(sm.matrix), 0.0, atol=1e-6) | ||
| @test isapprox(LinearAlgebra.det(sm.matrix), 0.0, atol=1e-5) |
There was a problem hiding this comment.
TODO(odow): take another look at this
|
@odow we can add this to the list of mysterious changes in results with v1.9, https://github.com/lanl-ansi/PowerModelsAnnex.jl/actions/runs/5063454342/jobs/9090086852 These differences do seem to be outside the bounds of machine precision issues. I was not able to reproduce the issue in my intel based MacBook pro. |
PowerModels is failing tests on Julia 1.9 due to a tolerance issue: