[Bug fortran/58771] [Regression] ICE in transfer_expr, at fortran/trans-io.c:2164

[pault at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58771

--- Comment #2 from Paul Thomas <pault at gcc dot gnu.org> ---
This one is going to drive me mad! It appears that the derived type is not
being built as Steve reported in comment #1.  If I force it to be built by
using a type 't' variable, the ICE goes away.  However, I have not found the
offending point where it is not being built... if you see what I mean.

A slightly more elaborate and legal testcase to illustrate this:
module m
  type t
    integer g
  end type
end

function f() result(ff)
  use m
  type(t) :: ff
  ff%g = 1
end

  use m
  type(t) :: z
  interface
    function f() result(ff)
      use m
      type(t) :: ff
    end function
  end interface
  print *,f()
!  z = t(1)       ! Does not ICE if this is uncommented
end

Paul