Source: scipy Followup-For: Bug #957780 The error is
gfortran:f77: scipy/sparse/linalg/eigen/arpack/ARPACK/SRC/dsaitr.f gfortran:f77: scipy/sparse/linalg/eigen/arpack/ARPACK/SRC/cnaitr.f scipy/sparse/linalg/eigen/arpack/ARPACK/SRC/dsaitr.f:671:35: 369 | call dvout (logfil, 1, rnorm, ndigit, | 2 ...... 671 | call dvout (logfil, 2, xtemp, ndigit, | 1 Error: Rank mismatch between actual argument at (1) and actual argument at (2) (scalar and rank-1) likewise l.743, and cnaitr.f ll.383,666,737 Something to do with rnorm being declared as scalar (Double precision) but dvout requiring an array, albeit dimension 1 (which would be rnorm(1) ). Looks like previously gfortran treated scalars as arrays of dimension 1, but GCC-10 must have changed the handling. Some discussion at https://stackoverflow.com/questions/34613356/error-rank-mismatch-in-argument-rank-1-and-scalar https://github.com/Reference-LAPACK/lapack/issues/353 says netlib source needs -std=legacy to compile.