440a5b570f
A routine in Semantics/resolve-directives.cpp was overwriting a symbol table pointer in a parse tree Name, thereby removing the AssocEntity with the correct type for a TYPE IS or CLASS IS clause that had been placed there. I don't really understand why resolve-directives has to overwrite symbol table pointers in the first place, but it definitely shouldn't be replacing these.
34 lines
748 B
Fortran
34 lines
748 B
Fortran
! RUN: %python %S/test_symbols.py %s %flang_fc1 -fopenacc
|
|
!DEF: /MAIN MainProgram
|
|
program MAIN
|
|
!DEF: /MAIN/t ABSTRACT DerivedType
|
|
type, abstract :: t
|
|
end type
|
|
!REF: /MAIN/t
|
|
!DEF: /MAIN/t2 DerivedType
|
|
type, extends(t) :: t2
|
|
!DEF: /MAIN/t2/y ObjectEntity REAL(4)
|
|
real :: y
|
|
end type
|
|
contains
|
|
!DEF: /MAIN/s (Subroutine) Subprogram
|
|
!DEF: /MAIN/s/d ObjectEntity CLASS(t)
|
|
subroutine s (d)
|
|
!REF: /MAIN/t
|
|
!REF: /MAIN/s/d
|
|
class(t) :: d
|
|
!DEF: /MAIN/s/a ObjectEntity REAL(4)
|
|
real a
|
|
!$acc data create(a)
|
|
!REF: /MAIN/s/d
|
|
select type (d)
|
|
!REF: /MAIN/t2
|
|
class is (t2)
|
|
!DEF: /MAIN/s/OpenACCConstruct1/OtherConstruct1/d AssocEntity CLASS(t2)
|
|
!REF: /MAIN/t2/y
|
|
d%y = 1.
|
|
end select
|
|
!$acc end data
|
|
end subroutine
|
|
end program
|