Daniel Karrasch
Daniel Karrasch
I went ahead and opened #54476, since it's orthogonal to this PR, but allows for more options.
Hm, failures look real. Something about reference to undefined objects in `Matrix{BigFloat}`...
I don't have time right now to work it out, but as an option, can we catch the potentially bad cases in `adjoint!` and `transpose!` by types? Like overload these...
Awesome, how many further improvements this has triggered through failing tests. :rofl:
I guess this can be closed?
Judging by, e.g., https://en.wikipedia.org/wiki/Mahalanobis_distance, that seems to be correct. Could you perhaps submit a PR which clarifies the terminology?
The fix has been backported to v1.8.4.
We currently don't have a pre-release of v1.9, so tests obviously fail. Once we have, then we should restrict testing to v1.9 and exclude nightly, to avoid visual clutter and...
Perhaps we should finalize #335, include it here and then go.
We don't squash this, right? Just create a merge commit. And how do we bump it over at `backports-release-1.9`? With the bot?