Define 5 arg hprod! for QuasiNewtonModel

[mohamed82008]

This method is needed when using Percival with LBFGSModel.

[codecov]

Codecov Report

Merging #3 (d9ce4b5) into master (e1662b6) will decrease coverage by 0.34%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##           master       #3      +/-   ##
==========================================
- Coverage   98.25%   97.90%   -0.35%     
==========================================
  Files           6        6              
  Lines         572      574       +2     
==========================================
  Hits          562      562              
- Misses         10       12       +2     

Impacted Files	Coverage Δ
src/quasi-newton.jl	93.33% <0.00%> (-6.67%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e1662b6...d9ce4b5. Read the comment docs.

[abelsiqueira]

The breakage tests have been updated to work with forks now. Can you rebase?

[mohamed82008]

Sure

[github-actions]

Package name	latest	stable
ADNLPModels.jl
AmplNLReader.jl
CUTEst.jl
CaNNOLeS.jl
DCI.jl
JSOSolvers.jl
LLSModels.jl
NLPModelsIpopt.jl
NLPModelsJuMP.jl
NLPModelsTest.jl
Percival.jl
QuadraticModels.jl
SolverBenchmark.jl
SolverTools.jl

[abelsiqueira]

Thanks

[abelsiqueira]

I just realized that using LBFGSMOdel with Percival will need some work because you have to use push! to update the approximation.

[abelsiqueira]

Here's how we do it in tron: https://github.com/JuliaSmoothOptimizers/JSOSolvers.jl/blob/master/src/tron.jl#L138

[abelsiqueira]

But now I've been reminded that Percival calls tron, so maybe it's not needed?!
Well, it needs some thought on what Hessian is being approximated, and I don't remember it well enough.
But since you're testing it, let me know what you find.

[mohamed82008]

Ya Percival calls tron on the augmented Lagrangian. The Hessian being approximated is that of the augmented Lagrangian.