downgrade CI

[ArnoStrouwen]

This one is quite the Gordian knot to untie.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (222a515) 8.37% compared to head (bd624b4) 8.46%.

Additional details and impacted files

@@            Coverage Diff            @@
##           master    #650      +/-   ##
=========================================
+ Coverage    8.37%   8.46%   +0.08%     
=========================================
  Files          30      30              
  Lines        2496    2470      -26     
=========================================
  Hits          209     209              
+ Misses       2287    2261      -26     

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

[ArnoStrouwen]

The downgrade CI passes locally for me.
Not sure what the issue is.
https://github.com/SciML/Optimization.jl/actions/runs/7314646211/job/19927490678?pr=650#step:6:4177

[ChrisRackauckas]

Are you running the tests of the extension packages as well? It seems this is due to something missing in ReverseDiff when doing the Lux tests. Maybe @avik-pal knows.

[ArnoStrouwen]

The Lux error is a bad lower bound reversediff compat in Lux, but that extension should not be used here.
I'll try adding the tests of the libs.

[Vaibhavdixit02]

What's the goal here?

[ArnoStrouwen]

See https://julialang.slack.com/archives/CN04R7WKE/p1702296727671319

[ArnoStrouwen]

Seems like failures are stochastic.

[ArnoStrouwen]

This one looks good to go.

[ChrisRackauckas]

@Vaibhavdixit02 are you going to look into https://github.com/SciML/Optimization.jl/actions/runs/7315937069/job/19930076157?pr=650#step:7:856 ?