[fwesselm]

Description

1. Add CentringStatus.
2. Only fix variables (from analytic centre) if solution is actually centred.
3. This fixes Feasible model reported infeasible with presolve enabled #3451 and Wrong optimal objective reported with presolve enabled #3452.

Checklist

• I have read the contributing guidelines
• This PR targets the latest branch
• Tests are passing
• Documentation was updated where relevant
• This PR is not primarily AI-generated (per the AI contributions policy in CONTRIBUTING.md)

[codecov]

Codecov Report

❌ Patch coverage is 90.00000% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 75.13%. Comparing base (1f304ce) to head (28117fb).
⚠️ Report is 46 commits behind head on latest.

Files with missing lines	Patch %	Lines
highs/lp_data/HighsInfo.cpp	50.00%	1 Missing ⚠️
highs/lp_data/HighsInfo.h	66.66%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           latest    #3456      +/-   ##
==========================================
+ Coverage   75.12%   75.13%   +0.01%     
==========================================
  Files         460      460              
  Lines      112155   112175      +20     
  Branches    17779    17787       +8     
==========================================
+ Hits        84254    84287      +33     
+ Misses      27619    27606      -13     
  Partials      282      282              

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[filikat]

Looks good.
Centring is only available for IPX (at least for now). Maybe in the reasons why the status can be kCentringStatusNone we should mention that it can also happen if HiPO is selected.

[Opt-Mucca]

Looks good! Two small questions:

• The instance takes a bit to solve (8s on my machine). Do we want this overhead to our tests? I'm not strictly against, but thought I should raise it.
• The comment from @filikat about whether this changes the HiPO execution path. Do we now skip fixings if HiPO is used? Can we even reliably use HiPO for these fixings if it doesn't currently have centering?

Also: TIL one can actually say centring. It looks funny.

[filikat]

Interior point methods usually return well centred solutions (before crossover).
The extra centring steps are there just to improve the centrality, but they are not always needed.
You could have a function isWellCentred (similar to ipm/ipx/ipm.cc:AssessCentrality) that is agnostic of which ipm is used and of whether centring was run and/or successful.

With the current changes, centring_status would always be none if HiPO is used for analytic centre computation.

[fwesselm]

Interior point methods usually return well centred solutions (before crossover). The extra centring steps are there just to improve the centrality, but they are not always needed. You could have a function isWellCentred (similar to ipm/ipx/ipm.cc:AssessCentrality) that is agnostic of which ipm is used and of whether centring was run and/or successful.

With the current changes, centring_status would always be none if HiPO is used for analytic centre computation.

Thanks, @filikat. I am fine with updating this PR to use a method for querying whether a solution is well centred.

[filikat]

That would probably be better.
I can write the function later or tomorrow, unless you want to give it a try.

[jajhall]

This addresses the issue well, without compromising the general good behaviour of this valuable technique