[SCIP] debugsol mechanism using PyScipOpt

[Benjamin Hiller]

Dear João,

thank you very much for your reply and the explanations regarding debug 
builds. Essentially you confirmed my suspicions. I have been able to 
resolve my issues with SCIP debug mode, so there is no need from my side 
for a follow-up.

Thank you also for creating the PR for the debug solution. Using a small 
example that I created before I was able to use it successfully. I can 
try to create a test for PyScipOpt from that.

Best regards,

Benjamin Hiller

Am 27.06.25 um 17:37 schrieb João Dionísio:
> Dear Benjamin,
> 
> Thank you for your message. It turns out that more recent PIPs stopped 
> supporting the --install-option flag, and the workaround you used is 
> perfectly valid. We need to update the documentation.
> 
> Regarding using PySCIPOpt with SCIP in debug mode, the original 
> intention was to catch failing assertions, to make debugging the Python 
> code easier. To debug what is happening inside SCIP is a fair bit more 
> complex, from what I know. I've never done it, but I can try looking 
> into it further, if this is what you're after.
> 
> To use a debug solution in SCIP, I believe you first need to call 
> SCIPenableDebugSol, which hasn't been interfaced to PySCIPOpt yet. Here 
> you can find an open PR adding it: 
> https://github.com/scipopt/PySCIPOpt/pull/1013 
> <https://github.com/scipopt/PySCIPOpt/pull/1013>. You can try compiling 
> the branch locally and see if it solves your problem. It's not tested 
> yet, though.
> 
> Please let me know if this answers your questions. Given that you say 
> that you managed to use debugsol from the SCIP binary, I may be 
> misunderstanding things.
> 
> Best,
> João Dionísio

-- 
Dr. Benjamin Hiller

fon +49 (30) 609882246
fax +49 (30) 609882299
http://www.atesio.de

atesio GmbH
Bundesallee 89
D-12161 Berlin
Germany

Geschäftsführung: Dr. Andreas Eisenblätter, Dr. Roland Wessäly
Amtsgericht Charlottenburg, HRB 76689