David Widmann

Among other things, it is simple, memory-efficient, supported by base and does not require any external packages, easy to manipulate, easy to save, supported by many other (Bayesian) analysis tools,...

You never sample `x[1]` and hence in every step `m` will be missing. Have you tried to specify `x[1]` in the `predict` step, i.e., not pass a vector full of...

> Though we need to fix the buildkite to not give us ARM. I don't know who added that to the ecosystem queue but that's the most breaking thing I've...

The tutorials didn't even start running since one can't use bash (and probably other programs) on these machines, so I don't know if there are additional Julia-specific problems.

Shouldn't these cases be handled by the algorithm that pops tstops? Even if the initial tstops are unique callbacks might add tstops that introduce non-uniqueness later during the integration process...

A lot should be covered by downstream tests such as https://github.com/SciML/DiffEqBase.jl/blob/master/test/downstream/event_detection_tests.jl.

Hmm no I don't remember any discussion about this.

> Also, unrelated, but defining `f` without argument `p` also seems to break the program, although `p` is never used (?) That [is intentional](https://diffeq.sciml.ai/stable/types/ode_types/), ODE problems require you to use...

Have you seen [ComponentArrays.jl](https://jonniedie.github.io/ComponentArrays.jl/dev/)? Maybe you could use it instead of DEDataArrays.

Just copied from my discussion on Slack with @freemin7: > Due to https://github.com/SciML/OrdinaryDiffEq.jl/blob/a5fe013d4c377ef4a39fd7d12bdc54e4558e11a2/src/integrators/integrator_utils.jl#L251 and similar line, you have to change integrator.vector_event_last_time as soon as you return an event_idx that is...