Ian McInerney
Ian McInerney
Turns out this was the `unsafe_load` causing the segfault, not the subsequent access to the struct (which was already guarded in a null check). This was fixed in https://github.com/osqp/OSQP.jl/pull/126, but...
Was that all that had to be done to transition to MOI 1.0, or are there other parts that we don't implement that we still need to?
If you could hold off doing any major updates to OSQP.jl for a while that would be appreciated. We are working on rebasing it to work with the updated API...
Alternately, if you want to target the branch we are working on the 1.0 version on, feel free to make more changes. But we don't have the jll it refers...
Thanks for the PR, and sorry it sat for so long before merging.
Can you provide a direct link to the test code that is causing the segfault? I can't investigate the C library aspect unless I have the reproducer code.
It is probably related to https://github.com/osqp/osqp/issues/346, which after a discussion between @gbanjac @bstellato and myself was traced back to a change in the tolerance checking code between 0.6.1 and 0.6.2....
We should probably just create an `UpperTriangular` view of `P` all the time and then utilize that for future processing, that way we don't need that check anymore.
I have toyed with the idea of exposing this, but haven't quite figured out the API that would be best to use at the C library level to allow user-supplied...
icc is different from icx. icx is the newer LLVM-based Intel compiler and icc is their older one that is now deprecated (to be removed in a year or two...