sjdaines

Partial fix in PR https://github.com/JuliaDiff/SparseDiffTools.jl/pull/146 This adds a fast path that avoids slow sparse indexing if J and sparsity are both SparseMatrixCSC with the same sparsity pattern. Updated timing (Julia...

I just retested our application (that uses Kinsol), putting GC.enable(false) and GC.enable(true) around the call to Sundials, and this appears to remove the segfault (an example failure was in https://github.com/PALEOtoolkit/PALEOmodel.jl/issues/25)....

The explicitly stored NVectors should of course be fine (as well as the explicitly stored memory handles). But this doesn't look like it covers everything: the root cause here I...

In fact looking at this again, I'm not sure there isn't another problem? According to the Julia manual `finalizer` is only defined for a mutable struct, so it's not clear...

Also, what requirements does the Sundials C API place on the lifetime of C N_Vector arguments to C function calls ? Can they be deleted once the C API call...

With Julia1.9.0-beta2 and our application code, CVODE segfaults in more places (and almost every time...) (edit: this case at least is fixed by https://github.com/SciML/Sundials.jl/pull/380 above, just adding the now-reproducible failure...

Yup this was my fix for the case above ! (I just got the order of issue update and submit PR wrong) @bolognam @markowkes this might not be the only...

Minimum change that I think would be needed to the generated code in `libsundials_api.jl` to make the `NVector` memory management work as in the current release (ie as it does...

Apologies, just seen this. Unfortunately I can't offer much help here as I *didn't* update the wrapper generation script or regenerate the wrappers, as just getting the clang generator toolchain...

This errors with `ERROR: UndefVarError: `NoAD` not defined` in Optimization.jl v3.25.0 See issue https://github.com/SciML/Optimization.jl/issues/747