Reproducer from a doctest in src/sage/matrix/matrix_double_dense.pyx sage: sage: A = matrix(CDF, [[1,2,4],[5,3,9],[7,8,6]]) sage: A.det() # wrong 0.0 sage: sage: A = matrix(QQ, [[1,2,4],[5,3,9],[7,8,6]]) sage: A.det() 88 sage: sage: A = matrix(RDF, [[1,2,4],[5,3,9],[7,8,6]]) sage: A.det() 88.0
On Tuesday, March 28, 2023 at 2:57:19 PM UTC-7 Matthias Koeppe wrote: > The error can also be reproduced by running the testsuite of > src/sage/matrix/matrix2.pyx > > No mention of openblas 0.3.22 in the numpy/scipy issue trackers yet, as > far as I could see. > > > On Tuesday, March 28, 2023 at 2:47:17 PM UTC-7 Matthias Koeppe wrote: > >> I agree that the openblas upgrade is to blame here. This error also >> reproduces with openblas 0.3.22 built from source - see >> https://github.com/sagemath/sage/pull/35371 >> >> >> On Tuesday, March 28, 2023 at 11:35:03 AM UTC-7 John H Palmieri wrote: >> >>> I ran "brew upgrade" recently on two different Macs, one Intel and one >>> Apple Silicon, and now the Sage documentation fails to build: >>> >>> [hyperboli] from >>> /Users/jpalmier/Desktop/Sage/sage_builds/TESTING/clean/sage-10.0.beta5/src/doc/en/reference/hyperbolic_geometry/sage/geometry/hyperbolic_space/hyperbolic_geodesic.rst: >>> [hyperboli] Traceback (most recent call last): >>> [hyperboli] File "sage/matrix/matrix_double_dense.pyx", line 345, in >>> sage.matrix.matrix_double_dense.Matrix_double_dense.__invert__ >>> [hyperboli] M._matrix_numpy = scipy.linalg.inv(self._matrix_numpy) >>> [hyperboli] File >>> "/Users/jpalmier/Desktop/Sage/sage_builds/TESTING/clean/sage-10.0.beta5/local/var/lib/sage/venv-python3.11/lib/python3.11/site-packages/scipy/linalg/_basic.py", >>> >>> line 975, in inv >>> [hyperboli] raise LinAlgError("singular matrix") >>> [hyperboli] numpy.linalg.LinAlgError: singular matrix >>> [hyperboli] During handling of the above exception, another exception >>> occurred: >>> [hyperboli] Traceback (most recent call last): >>> [hyperboli] File >>> "/Users/jpalmier/Desktop/Sage/sage_builds/TESTING/clean/sage-10.0.beta5/local/var/lib/sage/venv-python3.11/lib/python3.11/site-packages/matplotlib/sphinxext/plot_directive.py", >>> >>> line 517, in _run_code >>> [hyperboli] exec(code, ns) >>> [hyperboli] File "<string>", line 3, in <module> >>> [hyperboli] File >>> "/Users/jpalmier/Desktop/Sage/sage_builds/TESTING/clean/sage-10.0.beta5/src/sage/geometry/hyperbolic_space/hyperbolic_geodesic.py", >>> >>> line 1625, in perpendicular_bisector >>> [hyperboli] isom_mtrx = S.inverse() * (T1 * T2) * S >>> [hyperboli] ^^^^^^^^^^^ >>> [hyperboli] File "sage/matrix/matrix2.pyx", line 9964, in >>> sage.matrix.matrix2.Matrix.inverse >>> [hyperboli] return ~self >>> [hyperboli] File "sage/matrix/matrix_double_dense.pyx", line 347, in >>> sage.matrix.matrix_double_dense.Matrix_double_dense.__invert__ >>> [hyperboli] raise ZeroDivisionError("input matrix must be >>> nonsingular") >>> [hyperboli] ZeroDivisionError: input matrix must be nonsingular >>> >>> (I saw this first with 10.0.beta6 but then went and tried with >>> 10.0.beta5, which had succeeded earlier. Now it fails.) >>> >>> Perhaps this is due to an upgraded openblas? That's the most obvious >>> among the recently upgraded packages, at least to me: homebrew just >>> upgraded these: >>> >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 gnupg >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 pinentry >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 npth >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 libusb >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 libksba >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 libgcrypt >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 libassuan >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:21 libgpg-error >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:00 gh >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 10:00 imagemagick >>> drwxr-xr-x 3 jpalmier staff 96 Mar 27 09:59 tox >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 09:59 primecount >>> drwxr-xr-x 3 jpalmier staff 96 Mar 27 09:59 tcl-tk >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 09:59 qt >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 09:59 gobject-introspection >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 09:59 netpbm >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 09:59 glib >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 09:59 mpg123 >>> drwxr-xr-x 3 jpalmier admin 96 Mar 27 09:59 cmake >>> drwxr-xr-x 3 jpalmier staff 96 Mar 27 09:59 ghostscript >>> drwxr-xr-x 3 jpalmier staff 96 Mar 27 09:59 sqlite >>> drwxr-xr-x 3 jpalmier staff 96 Mar 27 09:59 openblas >>> >>> "brew info openblas" says that it's version 0.3.22, whereas Sage comes >>> with 0.3.21. Everything works after building with `./configure >>> --with-system-openblas=no`, but that flag triggers some other packages to >>> be built by Sage rather than come from the system, so I'm not 100% sure >>> it's the issue. >>> >>> Has anyone else seen this? >>> >>> -- >>> John >>> >>> -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/ee3c60ad-46d7-4718-92d5-12bfeab602a8n%40googlegroups.com.
