BUG: dgbmv gives "(len(x)>offx+(trans==0?m-1:n-1)*abs(incx)) failed for 7th argument x"

[lciti]

I get an error that the size of the x argument is incorrect when using dgbmv.
See the following MWE where I am creating a 5×7 matrix in BLAS general band format with zero subdiagonals and two superdiagonals. When the matrix is multiplied by a vector of length 7 it works fine. When dgbmv is asked to compute the product of the transpose of the matrix by a vector of length 5, it gives an error.

>>> import numpy as np
>>> from scipy.linalg.blas import dgbmv

>>> A = np.ones((3, 7))
>>> x1 = np.ones(7)
>>> x2 = np.ones(5)

>>> print(dgbmv(5, 7, 0, 2, 1., A, x1, trans=0))
[3. 3. 3. 3. 3.]
>>> print(dgbmv(5, 7, 0, 2, 1., A, x2, trans=1))
error                                     Traceback (most recent call last)
Cell In[97], line 10
      6 x2 = np.ones(5)
      9 print(dgbmv(5, 7, 0, 2, 1., A, x1, trans=0))
---> 10 print(dgbmv(5, 7, 0, 2, 1., A, x2, trans=1))

error: (len(x)>offx+(trans==0?m-1:n-1)*abs(incx)) failed for 7th argument x

If I pass an unnecessarily longer argument as x, it works fine and produces the correct answer:

>>> x1[-2:] = np.NaN
>>> print(dgbmv(5, 7, 0, 2, 1., A, x1, trans=1))
[1. 2. 3. 3. 3. 2. 1.]

Note that the two extra elements of x1 (compared to x2) are not used, otherwise the NaNs would propagate to the outputs.
It looks like the code doing the work is fine but there is a problem with the code checking the size of the arguments.

I am using numpy=1.23.5 and scipy=1.10.1.

[lciti]

I think the error is at line 95 of fblas_l2.pyf.src. Basically the same check is used for y and x while one is pre-multiplied and the other one is added (hence their length must be different for rectangular a).

90  check(len(y)>offy+(trans==0?m-1:n-1)*abs(incy)) :: y
...
95  check(len(x)>offx+(trans==0?m-1:n-1)*abs(incx)) :: x)

Unless I am missing something, the fix is easy and boils down to replacing == with != (or swapping m-1 and n-1) in line 95.