Brownian bridge not working properly after first solve

[rmsrosa]

If I solve an SDE with a Brownian bridge noise, the first solve is okay, but subsequent solves seem to be of a regular Wiener process, without bridging. I don't know yet where is the problem. Just registering the issue for now.

This happens in either the current versions DiffEqNoiseProcess v5.14.1 and StochasticDiffEq v6.55.0 or in the older versions DiffEqNoiseProcess v5.12.0, StochasticDiffEq v6.51.0.

MWE:

using StochasticDiffEq, DiffEqNoiseProcess

f = (u, p, t) -> 1.0
g = (u, p, t) -> 0.2
tspan = (0.0, 1.0)
x0 = 1.0

W = BrownianBridge(0.0, 1.0, 0.0, 1.0, nothing, nothing)

prob = SDEProblem(f, g, x0, tspan, noise = W, save_noise = true)

I get

julia> sol = solve(prob, EM(), dt=1/10); sol.W
t: 11-element Vector{Float64}:
 0.0
 0.1
 ⋮
 0.8999999999999999
 0.9999999999999999
u: 11-element Vector{Float64}:
 0.0
 0.26543511070295556
 ⋮
 1.0863290648610924
 1.0

julia> sol = solve(prob, EM(), dt=1/10); sol.W
t: 11-element Vector{Float64}:
 0.0
 0.1
 ⋮
 0.8999999999999999
 0.9999999999999999
u: 11-element Vector{Float64}:
  0.0
 -0.08632906486109237
  ⋮
 -0.11088269055897593
 -0.12475534714376024

[rmsrosa]

This is apparently due to W.S₁ being emptied at some point along the solve, and that is the only place where the info about the last point of the bridge is stored. Problem is that I haven't still understood the role of W.S₁ and W.S₂.

[rmsrosa]

fixed by #132