Evgeni Burovski
Evgeni Burovski
Note that numpy 2.0 redefines the complex sign to return `z/|z|`. https://github.com/cupy/cupy/pull/8314/commits/cdf5ca83100adace0da42ba008312ea3f392b85f
The case of `PPoly.extend` is a curious one indeed. - the method mutates the state of `self`; - the only way to get into trouble in the nogil world is...
> not sure why you're asking though - can you give an example of a SciPy function/method that is relevant here? The original comment ( motivated by an offline discussion...
Now that it's testing threaded workloads, it'd be great to pick up tests cases from https://github.com/scipy/scipy/issues/14162 and https://github.com/scipy/scipy/issues/11828 This is basically to double-check these issues do not make a comeback...
This now LGTM, modulo the last small ask: could you squash the commits somewhat? Esp the debugging back-and-forth on alpine etc.
Can you also remove the change to `cobyqa`? I _think_ a `git submodule update --init` would fix it.
As a quick note, I believe it's better to add an ASV memory benchmark (https://github.com/scipy/scipy/tree/main/benchmarks) than a unit test.
... and CI is green, and so is this PR. Thanks @Matryoskas for the patch and @Kai-Striega for the review.
`COMMON` blocks are relatively rare: the only non-legacy submodule which does not have a more modern replacement is ODRPACK (`integrate/dop` and `odepack` have an alternative in `solve_ivp`; `mvndst` is likely...
Needs a rebase now. Would be great to decide on this one either way.