qpms/BUGS.rst

Known bugs
===========

Wrong factor on B coefficient (single coeff version)
-----------------------------
(Probably fixed in the "calculator object" versions!)
Under Kristensson normalisation (with CS = -1), my code gives
B(1,0,n,n)/B(1,0,n,-n) == -(2n)! at (x,y,z) = (x,0,0)
(expected plus or minus 1).
A-coefficients seem to behave correctly.
BUT THE CSPHASE MIGHT BE WRONG ON BOTH SINGLE A AND B (TODO TEST)

Xu's antinormalisation
----------------------
"Xu's antinormalisation" is broken (most likely in legendre.c and maybe
also in qpms_types.h) – the plane wave test fails and the spherical wave
reconstruction as well (but the translation coefficients match the 
Xu's tables).

Longitudinal waves
------------------
Plane wave decompositions gives wrong value on the longitudinal part.
The implementation of the L coefficients OR the longitudinal waves
is thus probably wrong.


Singular value asymmetry
------------------------
Similar as the scattering result asymmetry, although not necessarily 
only in the y-direction?

Overflows etc.
--------------
Assertion failed in gaunt_xu for test_vswf_translations.c and high values of LMAX
(LMAX=25)

Moreover, the non-normalized legendre functions that are used in translations.c
are likely to overflow for even smaller values of l. 
Therefore TODO: modify translations.c in a way that uses normalised legendre functions everywhere.