Possible bug: Missing matches in some tests of gipsl.all.build.filter

[maxkratz]

See: https://github.com/Echtzeitsysteme/gips-tests/blob/main/gipsl.all.build.filter/src/gipsl/all/build/filter/Model.gipsl

Sometimes, the test testMap2to1NoIsDoubled() (https://github.com/Echtzeitsysteme/gips-tests/blob/main/test.suite.gips/src/test/suite/gipsl/all/build/GipslAllBuildFilterTest.java#L46-L56) fails because some of the matches are missing.

ILP solver output if test passes:

\ LP format - for model browsing. Use MPS format to capture full model detail.
Maximize
  n2n#0 + n2n#1
Subject To
 PatternConstraint1OnvnodeNotMapped_0: n2n#0 = 1
 PatternConstraint1OnvnodeNotMapped_1: n2n#1 = 1
 PatternConstraint0OnvnodeNotMapped_0: n2n#0 + n2n#1 = 1
 PatternConstraint0OnvnodeNotMapped_1: n2n#0 + n2n#1 = 1
Bounds
Binaries
 n2n#0 n2n#1
End

ILP solver output if test fails:

\ LP format - for model browsing. Use MPS format to capture full model detail.
Maximize
  n2n#0 + n2n#1
Subject To
 PatternConstraint0OnvnodeNotMapped_0: n2n#0 + n2n#1 = 1
 PatternConstraint0OnvnodeNotMapped_1: n2n#0 + n2n#1 = 1
Bounds
Binaries
 n2n#0 n2n#1
End

AFAIK, this behaviour is independent from the used pattern matcher (I've also tested it with Democles) and the used ILP solver (GUROBI and GLPK show the same error).

[maxkratz]

See tests: https://github.com/Echtzeitsysteme/gips-tests/blob/main/test.suite.gips/src/test/suite/gipsl/all/build/GipslAllBuildFilterTest.java

[maxkratz]

Should be unrelated to eMoflon/emoflon-ibex-updatesite@3515a60.

[Flipy91]

I also observed this behavior - it seems to be a concurrency issue in the GipslEngine l.37:

constraints.values().parallelStream().forEach(constraint -> constraint.buildConstraints());

Workaround is building the constraints serial instead of parallel:

constraints.values().stream().forEach(constraint -> constraint.buildConstraints());