Failing tests with MOI and JuMP

[JiazhengZhu]

Convex.jl tests with BigFloat numeric type

• No tests for satisfy, maximize or minimize poblem constructors. They do not process type information yet.

• No tests for lambda_max or lambda_min atoms. This is due to this issue with GenericLinearAlgebra.

MOI.Test

• Compared with SDPA.jl, we further exclude tests 'norminf2' and 'normone2' in MOIT.contconictest(). We get MOI.UNKNOWN_RESULT_STATUS instead of MOI.INFEASIBILITY_CERTIFICATE.

JuMP/examples

• diet.jl does not work.

ERROR: LoadError: In `MathOptInterface.ScalarAffineFunction{Float64}`-in-`MathOptInterface.EqualTo{Float64}` constraint: Constant -1800.0 of the function is not zero. The function constant should be moved to the set. You can use `MOI.Utilities.normalize_and_add_constraint` which does this automatically.

Code to reproduce these failures: failing_tests.zip

[ericphanson]

On the Convex#MathOptInterface branch, I added support for numeric types in minimize etc (via a keyword argument), and updated the Problem Depot problems to use it. I just updated the tests here to test Convex.jl problems with BigFloats. I tested locally with just sdpa-gmp and it was OK; let's see what Travis says for the rest. We may need to update the excludes.

[ericphanson]

I updated the tests a bit in Convex.jl#MathOptInterface to add "socp" to the problems that use SOCP, and to move a test that solve! to utilities, since otherwise it would check handle_problem! and sometimes we return stuff from that. So CI for this is here actually: https://travis-ci.com/ericphanson/SDPAFamily.jl/builds/127229390 (triggered after the last of those changes).