Is `GC.@preserve` enough?

[rejuvyesh]

I am seeing a bunch of memory corruption in rejuvyesh/PyCallChainRules.jl#10 which go away if I wrap the tests inside GC.enable(false) and GC.enable(true). Still trying to figure out a MWE because this is not as consistent with simple use of dlpack.

[p-zubieta]

Can you check if after #15 you still see this issue?

[rejuvyesh]

It's possible that I am doing it wrong (I don't have any GC.@preserve in my codebase), but I still get the same issues. Everything passes with GC.enable(false) but memory corruption otherwise. I updated the code to work with the new DLPack at rejuvyesh/PyCallChainRules.jl#10.

[p-zubieta]

Fixed in #16