TESTING/LIN: align ISEED declarations with documented dimension in four test routines

[nakatamaho]

This PR updates four TESTING/LIN routines:

• clatsp
• clatsy
• zlatsp
• zlatsy

In all four files, the parameter documentation says that ISEED is an
INTEGER array of dimension (4), but the actual dummy argument
declaration is currently ISEED(*).

These routines pass ISEED to CLARND/ZLARND, whose interfaces use
ISEED(4) and whose documentation states that the fourth element must
be odd.

This patch changes both the commented argument declarations and the
actual dummy argument declarations from ISEED(*) to ISEED(4) so
that the routine declarations match the documented contract and the
callee interfaces.

No functional change is intended.

[langou]

A comment.

If, for example, an array is LDA-by-N, then LAPACK will declare it as

      DOUBLE PRECISION   A( LDA, * )

LAPACK will not declare it as

      DOUBLE PRECISION   A( LDA, N )

Similarly, if ISEED is a array of size 4, then LAPACK will declare it as

      INTEGER            ISEED( * )

and not

      INTEGER            ISEED( 4 )

Both ways of doing it are correct in FORTRAN, but I agree that the latter is better software practice. (Albeit not followed by LAPACK).

All in all, all good with me.

Thanks!

[nakatamaho]

Thanks, I agree. Both forms are valid Fortran, and I did not mean to imply that LAPACK’s (*) style is wrong.

My comment was only from the downstream translation side: in our FABLE-based Fortran-to-C++ flow, explicit extents like ISEED(4) preserve fixed-size array information more reliably than ISEED(*). We first ran into this in generated WRITE/I/O handling, where missing extent information can lead to overly generic treatment on the C++ side.

So I fully agree this is not a LAPACK correctness issue — just a tooling convenience on our side.

[langou]

Note: I would be in favor to propage these types of change more broadly. But this needs to be done with care.

I appreciate the effort and contribution for this specific four routines. Thanks!