Bug in ropp_io_vlist_size.f90

[Ian Culverwell]

ropp_io/ropp/ropp_io_vlist_size.f90, the overloading of the fortran intrinsic SIZE function for the various (sub-) types of vlists, contains a bug:

FUNCTION size_vlist(vlist) RESULT(n),
...
  nn(1) = size_vlistD0d(vlist%VlistD0d)
  nn(2) = size_vlistD1d(vlist%VlistD1d)
  nn(2) = size_vlistD2d(vlist%VlistD2d)

  n = SUM(nn)

should of course read

FUNCTION size_vlist(vlist) RESULT(n
...
  nn(1) = size_vlistD0d(vlist%VlistD0d)
  nn(2) = size_vlistD1d(vlist%VlistD1d)
  nn(3) = size_vlistD2d(vlist%VlistD2d)

  n = SUM(nn)

[Ian Culverwell]

According to the word of

grep -ir vlist $ROPP_SRC |grep -i size |grep -v svn | grep f90

SIZE(vlist) is never used by itself: ROPP only calls SIZE(vlist%VlistD0d) etc. So fixing this should have no effect. (Which is also why getting it wrong all these years didn't cause any problems.)

But to confirm, when we run ropp_apps_tph_tool on some (artificial) data with 0, 1, and 2D extra data, we find, at ROPP8.1:

 SIZE(ro_data1%vlist%vlistd0d) =            2
 SIZE(ro_data1%vlist%vlistd1d) =            6
 SIZE(ro_data1%vlist%vlistd2d) =            4
 SIZE(ro_data1%vlist)          =            6

which is obviously wrong, while at ROPP9.0:

 SIZE(ro_data1%vlist%vlistd0d) =            2
 SIZE(ro_data1%vlist%vlistd1d) =            6
 SIZE(ro_data1%vlist%vlistd2d) =            4
 SIZE(ro_data1%vlist)          =           12

which is obviously correct. (But the results are the same, thankfully.)

That's good enough: implementing at r4892.

Testing script also attached.

[Ian Culverwell]

test2.sh

[Ian Culverwell]

Closing ticket as fixed.