[SCIP] isModelCons in Orbisack -- Crash
Christopher Hojny
c.hojny at tue.nl
Tue Feb 9 21:05:03 CET 2021
Hi Gleb,
sorry, somehow I focused on the cip file rather than your original
question. Coming back to this question, I think you have found a bug in
SCIP:
Although I could not reproduce exactly the same stack trace, I get a
similar error message and I think the lines
if ( !sourcedata->ismodelcons )
return SCIP_OKAY;
in CONSTRANS of cons_orbisack.c are responsible for the crash. If the
orbisack constraint is not a model constraint, these lines cause that
the transformation of orbisack constraints is skipped although SCIP
expects transformation to happen. If I remove these two lines, the code
runs fine on your cip example.
Could you please try whether removing these two lines of code fixes the
issue also for you?
Best,
Christopher
On 09-02-2021 08:49, Gleb Belov wrote:
> Hi,
>
> the original problem is that when creating that model with C API and
> isModelCons==FALSE, there is a crash... Do you need a small C API source
> or can you follow my steps from previous email?
>
> Cheers,
>
> *Gesendet:* Dienstag, 09. Februar 2021 um 18:14 Uhr
> *Von:* "Christopher Hojny" <c.hojny at tue.nl>
> *An:* scip at zib.de
> *Betreff:* Re: [SCIP] isModelCons in Orbisack -- Crash
> Hi Gleb,
>
> thanks for the example. It seems that we have missed to implement
> parsing routines for orbisacks (and symresacks). My apologies for the
> inconveniences. I will put it on my TODO list.
>
> Until the missing feature has been added, you can still add orbisacks
> indirectly via orbitopes (which are generalizations of orbisacks and for
> which parsing routines are available). I have modified your cip file
> slightly, see below, and it seems to work.
>
> Best,
> Christopher
>
> ========================================================
> STATISTICS
> Problem name : mzn_scip
> Variables : 11 (10 binary, 1 integer, 0 implicit integer, 0
> continuous)
> Constraints : 0 initial, 1 maximal
> OBJECTIVE
> Sense : minimize
> VARIABLES
> [binary] <X_INTRODUCED_0_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_1_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_2_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_3_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_4_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_5_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_6_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_7_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_8_>: obj=0, original bounds=[0,1]
> [binary] <X_INTRODUCED_9_>: obj=0, original bounds=[0,1]
> [integer] <objective>: obj=0, original bounds=[2,2]
> CONSTRAINTS
> [orbitope] <p_lex_lesseq__orbisack_0>:
> fullOrbitope(<X_INTRODUCED_0_>,<X_INTRODUCED_1_>.<X_INTRODUCED_2_>,<X_INTRODUCED_3_>.<X_INTRODUCED_4_>,<X_INTRODUCED_5_>.<X_INTRODUCED_6_>,<X_INTRODUCED_7_>.<X_INTRODUCED_8_>,<X_INTRODUCED_9_>);
> END
> ============================================================
>
>
>
> On 08-02-2021 14:19, Gleb Belov wrote:
>> Hi Christopher,
>>
>> here is a small example:
>>
>> =============================
>> STATISTICS
>> Problem name : mzn_scip
>> Variables : 11 (10 binary, 1 integer, 0 implicit integer, 0
>> continuous)
>> Constraints : 0 initial, 1 maximal
>> OBJECTIVE
>> Sense : minimize
>> VARIABLES
>> [binary] <X_INTRODUCED_0_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_1_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_2_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_3_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_4_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_5_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_6_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_7_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_8_>: obj=0, original bounds=[0,1]
>> [binary] <X_INTRODUCED_9_>: obj=0, original bounds=[0,1]
>> [integer] <objective>: obj=0, original bounds=[2,2]
>> CONSTRAINTS
>> [orbisack] <p_lex_lesseq__orbisack_0>:
>>
> orbisack(X_INTRODUCED_0_,X_INTRODUCED_1_.X_INTRODUCED_2_,X_INTRODUCED_3_.X_INTRODUCED_4_,X_INTRODUCED_5_.X_INTRODUCED_6_,X_INTRODUCED_7_.X_INTRODUCED_8_,X_INTRODUCED_9_;
>> END
>> =============================
>>
>> However when I tried to read it in SCIP I got message:
>> WARNING: constraint handler <orbisack> does not support parsing
> constraints
>>
>> The .cip model was exported using C API, and the model created too, from
>> MiniZinc.
>>
>> Stack trace (gcc 9.3.0, Ubuntu 20.04):
>> =============================
>> Program received signal SIGSEGV, Segmentation fault.
>> 0x00007ffff16367ea in SCIPconsTransform (origcons=0x555555eab350,
>> blkmem=0x555555ddf070, set=0x555555dd47a0, transcons=0x7fffffffce98)
>> at
>>
> /home/bg25/Downloads/Software/Solvers/compiled_packages/SCIP/scipoptsuite-7.0.2/scip/src/scip/cons.c:6439
>> 6439 (*transcons)->transorigcons = origcons;
>> (gdb) where
>> #0 0x00007ffff16367ea in SCIPconsTransform (origcons=0x555555eab350,
>> blkmem=0x555555ddf070, set=0x555555dd47a0, transcons=0x7fffffffce98)
>> at
>>
> /home/bg25/Downloads/Software/Solvers/compiled_packages/SCIP/scipoptsuite-7.0.2/scip/src/scip/cons.c:6439
>> #1 0x00007ffff18646cf in SCIPprobTransform (source=0x555555c7a290,
>> blkmem=0x555555ddf070, set=0x555555dd47a0, stat=0x555555f793c0,
>> primal=0x555555c8d9e0,
>> tree=0x555555e51630, reopt=0x0, lp=0x555555cefd60,
>> branchcand=0x555555e2a0c0, eventfilter=0x555555ef10c0,
>> eventqueue=0x555555e24bc0, conflictstore=0x555555f831b0,
>> target=0x555555dbb518) at
>>
> /home/bg25/Downloads/Software/Solvers/compiled_packages/SCIP/scipoptsuite-7.0.2/scip/src/scip/prob.c:584
>> #2 0x00007ffff19043ea in SCIPtransformProb (scip=0x555555dbb470)
>> at
>>
> /home/bg25/Downloads/Software/Solvers/compiled_packages/SCIP/scipoptsuite-7.0.2/scip/src/scip/scip_solve.c:407
>> #3 0x00007ffff190e4ff in SCIPpresolve (scip=0x555555dbb470)
>> at
>>
> /home/bg25/Downloads/Software/Solvers/compiled_packages/SCIP/scipoptsuite-7.0.2/scip/src/scip/scip_solve.c:2390
>> #4 0x00007ffff190f1f6 in SCIPsolve (scip=0x555555dbb470)
>> at
>>
> /home/bg25/Downloads/Software/Solvers/compiled_packages/SCIP/scipoptsuite-7.0.2/scip/src/scip/scip_solve.c:2620
>> #5 0x000055555567120c in MIPScipWrapper::solveSCIP() ()
>> #6 0x0000555555660dca in
>> MiniZinc::MIPSolverinstance<MIPScipWrapper>::solve() ()
>> #7 0x00005555555bfd2e in MiniZinc::MznSolver::solve() ()
>> #8 0x00005555555c874e in
>> MiniZinc::MznSolver::run(std::vector<std::__cxx11::basic_string<char,
>> std::char_traits<char>, std::allocator<char> >,
>> std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>,
>> std::allocator<char> > > > const&, std::__cxx11::basic_string<char,
>> std::char_traits<char>, std::allocator<char> > const&,
>> std::__cxx11::basic_string<char, std::char_traits<char>,
>> std::allocator<char> > const&, std::__cxx11::basic_string<char,
>> std::char_traits<char>, std::allocator<char> > const&) ()
>> #9 0x00005555555bb2cd in main ()
>> =============================
>>
>>
>> I guess should be easiest to go from here, but if you want to reproduce,
>> need MiniZinc @ 4802031dc17742e6978fc683ef68324c73a66796 and the
>> following command:
>>
>> minizinc -v -s --solver scip
>> challenge_from_2012/mznc2017_probs/opd/opd__reduce_find_SCIPbug.mzn
>> challenge_from_2012/mznc2017_probs/opd/small_bibd_06_50_25__reduced.dzn
>> --writeModel opd__verySmall.cip
>>
>> with the .mzn and .dzn as follows (reduced from original ones in MZN
>> Challenge 2017):
>> .MZN:
>> include "lex_greatereq.mzn";
>>
> %------------------------------------------------------------------------------%
>> % Parameters
>> int: v; % Number of rows
>> int: b; % Number of columns
>> int: r; % Row sum
>> set of int: rows = 1..v; % Set of rows
>> set of int: cols = 1..b; % Set of columns
>>
> %------------------------------------------------------------------------------%
>> % Lower bound computation for lambda
>> int: rv = r*v;
>> int: rvmodb = rv mod b;
>> int: floorrv = rv div b;
>> int: ceilrv = rv div b + bool2int(rv mod b != 0);
>> int: num = (ceilrv*ceilrv*rvmodb+floorrv*floorrv*(b-rvmodb)-rv);
>> int: denom = v*(v-1);
>> int: lb_lambda = num div denom + bool2int( num mod denom !=0);
>>
> %------------------------------------------------------------------------------%
>> % This line is there for debugging purposes, it can be safely removed
>> %int: tmp = trace("% Computed lower bound for lambda:
>> "++show(lb_lambda)++"\n",lb_lambda);
>>
> %------------------------------------------------------------------------------%
>> % Variables
>> array [rows, cols] of var 0..1: m;
>> var lb_lambda..b: objective; % The 'lambda' variable
>> constraint symmetry_breaking_constraint(
>> forall(j in {min(cols), min(cols)+1} diff {max(cols)})(
>> lex_greatereq([m[i, j] | i in rows], [m[i, j+1] | i in rows])
>> )
>> );
>>
>> .DZN:
>> v = 5;
>> b = 2;
>> r = 2;
>>
>>
>> I also tried compiling current scipoptsuite/master but my ssh seems to
>> invalid for gcg, and then cmake canot find CMakeLists.txt for soplex for
>> example.
>>
>> Best regards,
>> Gleb
>>
>> *Gesendet:* Freitag, 05. Februar 2021 um 22:01 Uhr
>> *Von:* "Christopher Hojny" <c.hojny at tue.nl>
>> *An:* scip at zib.de
>> *Betreff:* Re: [SCIP] isModelCons in Orbisack -- Crash
>> Hi Gleb,
>>
>> you are right. If you want to use orbisacks to "just" handle symmetries
>> in your problem, you should set the "isModelCons" flag to FALSE. In this
>> case, SCIP will also except solutions that do not adhere to the orbisack
>> constraint. Only in case you want to enforce that a solution adheres to
>> the orbisack structure, the "isModelCons" flag should be set to TRUE. If
>> SCIP detects orbisacks on its own, it always adds them with the flag
>> being FALSE to not hinder heuristics to find solutions.
>>
>> Regarding your crash report, it is difficult to find the reason without
>> further information. Could you provide further information on how you
>> add orbisacks, a stack trace of the fail, or a minimum crashing example
>> please?
>>
>> Best,
>> Christopher
>>
>>
>>
>> On 05-02-2021 06:49, Gleb Belov wrote:
>>> Hi,
>>>
>>> I have a model where anti-symmetry constraints use orbisacks. When
>>> setting isModelCons==0, SCIP 7.0.2 crashes i SCIPconsTransform(). Do I
>>> understand correctly that pure anti-symmetry constraints should set
>>> isModelCons==0?
>>>
>>> Cheers,
>>> Gleb
>>>
>>>
>>>
>>> _______________________________________________
>>> Scip mailing list
>>> Scip at zib.de
>>> https://listserv.zib.de/mailman/listinfo/scip
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C7140fc0d520a4a9000f808d8cccf592d%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637484538214761188%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=URzKSc06BJS2pHt5Ww3b8W76rjH%2BJz6jpH5xoOYhgNE%3D&reserved=0>
>>
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C626785edf96840c9b5b308d8cc3485cb%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637483873252578462%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=5BfGlxui22qs1pCgs8btUzlMSqmXE%2BoL7g1VMYelhgg%3D&reserved=0
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C7140fc0d520a4a9000f808d8cccf592d%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637484538214771182%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=YvQ3Hg8Zkqc46C2nX9Ohotfq%2FPkDgonbyDGgZanFTWo%3D&reserved=0>>
>>>
>> _______________________________________________
>> Scip mailing list
>> Scip at zib.de
>> https://listserv.zib.de/mailman/listinfo/scip
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C7140fc0d520a4a9000f808d8cccf592d%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637484538214771182%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=YvQ3Hg8Zkqc46C2nX9Ohotfq%2FPkDgonbyDGgZanFTWo%3D&reserved=0>
>>
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C626785edf96840c9b5b308d8cc3485cb%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637483873252578462%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=5BfGlxui22qs1pCgs8btUzlMSqmXE%2BoL7g1VMYelhgg%3D&reserved=0
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C7140fc0d520a4a9000f808d8cccf592d%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637484538214781179%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=5%2BIcHz1whvSn0uIw1gx2Aci0AnixPm7go47Icy%2Bl2uQ%3D&reserved=0>>
>>
>>
>>
>> _______________________________________________
>> Scip mailing list
>> Scip at zib.de
>> https://listserv.zib.de/mailman/listinfo/scip
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C7140fc0d520a4a9000f808d8cccf592d%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637484538214781179%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=5%2BIcHz1whvSn0uIw1gx2Aci0AnixPm7go47Icy%2Bl2uQ%3D&reserved=0>
>>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
> <https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flistserv.zib.de%2Fmailman%2Flistinfo%2Fscip&data=04%7C01%7C%7C7140fc0d520a4a9000f808d8cccf592d%7Ccc7df24760ce4a0f9d75704cf60efc64%7C1%7C0%7C637484538214791170%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C2000&sdata=hx7L3KO6hb9Ku%2FSEfD6a08uNfFCr1NH%2F4R6PzwKnKxM%3D&reserved=0>
>
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
>
More information about the Scip
mailing list