illegal instruction in try/finally

[mlubin]

To reproduce:

1. Check out this PR: [release-0.18] Readd GLPK jump-dev/JuMP.jl#1498
2. Run JuMP's tests
3. Wait a few minutes
4. Expected output: https://travis-ci.org/JuliaOpt/JuMP.jl/jobs/433523251#L864

@tkoolen correctly guessed an issue with try/finally. The illegal instruction goes away when I comment out the try/finally/end lines at https://github.com/JuliaOpt/GLPKMathProgInterface.jl/blob/fdad2fb72a93be219573e48248b141380c821ee8/src/GLPKInterfaceMIP.jl#L457-L471 leaving the code in between.

Note that this code involves callbacks through C back to Julia which has caused issues with exception handling in the past (#14284).

[mlubin]

Reduced test case. Using JuMP@1fd17f17 (release-0.18) and GLPKMathProgInterface v0.4.3

using JuMP, GLPKMathProgInterface

mod = Model(solver=GLPKSolverMIP())
@variable(mod, 0 <= x <= 2, Int)
@variable(mod, 0 <= y <= 2, Int)
@objective(mod, Max, y + 0.5x)
function corners(cb)
    x_val = getvalue(x)
    y_val = getvalue(y)
    TOL = 1e-6
    if y_val + x_val > 3 + TOL
        @lazyconstraint(cb, y + 0.5x + 0.5x <= 3)
    end
end
addlazycallback(mod, corners)
@show solve(mod)

Interestingly enough, the crash goes away with JuliaOpt/GLPKMathProgInterface.jl#54. So it looks like it was the missing method errors that turned into illegal instructions.

[c42f]

This sounds familiar, I've seen the missing method error turn into an illegal instruction. I think it was over at JuliaArrays/StaticArrays.jl#493, but annoyingly I can't put my finger on the exact code now.

[ViralBShah]

Reopen if still an issue.