Crash multithreading

[ghost]

What version of OR-tools and what language are you using?
Version: 7.6
Language: Python

Which solver are you using (e.g. CP-SAT, Routing Solver, GLOP, BOP, Gurobi)
CP-SAT

What operating system (Linux, Windows, ...) and version?
Ubuntu 18.04

What did you do?
Run a model with num_search_workers = 8

What did you expect to see
No crash, maybe MODEL_INVALID if there is a problem with the proto.

What did you see instead?

Check failed: it != collection.end() Map key not found: 2970
*** Check failure stack trace: ***
    @     0x7f2568cd6100  google::LogMessage::Fail()
    @     0x7f2568cd605c  google::LogMessage::SendToLog()
    @     0x7f2568cd59c0  google::LogMessage::Flush()
    @     0x7f2568cd8df6  google::LogMessageFatal::~LogMessageFatal()
    @     0x7f25661dfb99  gtl::FindOrDie<>()
    @     0x7f25662e0e16  operations_research::sat::PresolveContext::SubstituteVariableInObjective()
    @     0x7f25661c9d20  operations_research::sat::CpModelPresolver::RemoveSingletonInLinear()
    @     0x7f25661d84e9  operations_research::sat::CpModelPresolver::PresolveOneConstraint()
    @     0x7f25661da306  operations_research::sat::CpModelPresolver::PresolveToFixPoint()
    @     0x7f25661ddf3b  operations_research::sat::CpModelPresolver::Presolve()
    @     0x7f25661df121  operations_research::sat::PresolveCpModel()
    @     0x7f256620460b  _ZZN19operations_research3sat12_GLOBAL__N_19LnsSolver12GenerateTaskElENKUlvE_clEv
    @     0x7f256631612c  _ZNSt17_Function_handlerIFvvEZN19operations_research3sat20NonDeterministicLoopERKSt6vectorISt10unique_ptrINS2_9SubSolverESt14default_deleteIS5_EESaIS8_EEiEUlvE_E9_M_invokeERKSt9_Any_data
    @     0x7f2566018208  operations_research::RunWorker()
    @     0x7f256665428f  execute_native_thread_routine
    @     0x7f256cfd26db  start_thread
    @     0x7f256d30b88f  clone
    @              (nil)  (unknown)

Does not crash in v7.4.
model.zip

May be a dup of #2001

Anything else we should know about your project / environment

[lperron]

Thanks.

I could simplify the model that crashes in the lns subsolver.

[lperron]

I pushed the fix on master.

Thanks again for the bug report.