uncomplicate.neanderthal.linalg/org error in 0.25.3

[sparkofreason]

The following code worked in 0.24.0:

  (require '[uncomplicate.neanderthal.linalg :as linear-algebra]
           '[uncomplicate.neanderthal.native :as native]
           `[uncomplicate.neanderthal.core :as neanderthal])
  (let [a (native/dge [[1.0 2.0] [3.0 2.0] [6.0 2.0] [5.0 3.0]])
        qr (linear-algebra/qrf a)
        q (linear-algebra/org qr)]
    q)

Updating to 0.25.3 gives the following error:

Intel MKL ERROR: Parameter 2 was incorrect on entry to DORGQR.
Syntax error compiling at (/home/davedixon/Projects/Compute/math/src/provisdom/math/neanderthal_matrix.clj:249:3).
There has been an illegal argument in the native function call.
    at clojure.lang.Compiler.load(Compiler.java:7648)
clojure.lang.ExceptionInfo: There has been an illegal argument in the native function call. {:arg-index 3}
    at uncomplicate.neanderthal.internal.host.mkl.DoubleGEEngine.gqr(mkl.clj:904)
    at uncomplicate.neanderthal.internal.api$eval13429$fn__13654$G__13403__13658.invoke(api.clj:59)
    at uncomplicate.neanderthal.internal.api$eval13429$fn__13654$G__13402__13663.invoke(api.clj:59)
    at uncomplicate.neanderthal.internal.common.OrthogonalFactorization.org(common.clj:288)
    at uncomplicate.neanderthal.linalg$org.invokeStatic(linalg.clj:419)
    at uncomplicate.neanderthal.linalg$org.invoke(linalg.clj:409)
    at provisdom.math.neanderthal_matrix$eval31890.invokeStatic(neanderthal_matrix.clj:251)
...

Looking at the commits from 0.24.0 to 0.25.3, I guessed that native/dge as used above would return the transpose of what it did previously, and indeed the following works as before:

(let [a (neanderthal/trans (native/dge [[1.0 2.0] [3.0 2.0] [6.0 2.0] [5.0 3.0]]))
        qr (linear-algebra/qrf a)
        q (linear-algebra/org qr)]
    q)

Not sure why linear-algebra/org fails in the first case, when linear-algebra/qrf runs, perhaps something I'm misunderstanding about the underlying implementation.

[blueberry]

This is caused by the intended change in the transfer! multimethod implementation between nested sequences and matrices. Previously, the nested sequence was always considered to consist of the rows of a matrix, and now it depends of the :layout. Since the default layout in ge constructors is :column, in your example the values are transposed. The solution is to either rearrange the values in the source sequence, or add {:layout :row} optional argument to the dge call.

[sparkofreason]

Why does org throw an error while qrf seems to return an answer?