Comment 9 for bug 1088393

I'm chasing the test case failure on i386/lpia which caused a build failure:

***************
*** 432,443 ****
  select * from numeric_table
    where num_col in (select float_col from float_table);
           num_col
! -------------------------
                         1
- 1.000000000000000000001
                         2
                         3
! (4 rows)

  --
  -- Test case for bug #4290: bogus calculation of subplan param sets
--- 432,442 ----
  select * from numeric_table
    where num_col in (select float_col from float_table);
   num_col
! ---------
         1
         2
         3
! (3 rows)

The 8.3.21 -> 8.3.22 diff does not touch this part, and indeed building the current hardy-updates version 8.3.21 on current hardy now fails in exactly the same way. This happens in a hardy i386 chroot on raring amd64. So this could be a buildd change or a regression from a hardy-updates package. I'll do some bisecting to see what caused this.