Initial checkin of GCC 4.9.0 from trunk (r208799).

Change-Id: I48a3c08bb98542aa215912a75f03c0890e497dba

1 files changed, 67 insertions, 0 deletions

diff --git a/gcc-4.9/gcc/testsuite/gfortran.dg/c_by_val_5.f90 b/gcc-4.9/gcc/testsuite/gfortran.dg/c_by_val_5.f90
new file mode 100644
index 000000000..3a8bc3bf7
--- /dev/null
+++ b/gcc-4.9/gcc/testsuite/gfortran.dg/c_by_val_5.f90
@@ -0,0 +1,67 @@
+! { dg-do run }
+! Overwrite -pedantic setting:
+! { dg-options "-Wall" }
+!
+! Tests the fix for PR31668, in which %VAL was rejected for
+! module and internal procedures.
+! 
+
+subroutine bmp_write(nx)
+  implicit none
+  integer, value :: nx
+  if(nx /= 10) call abort()
+  nx = 11
+  if(nx /= 11) call abort()
+end subroutine bmp_write
+
+module x
+ implicit none
+ ! The following interface does in principle
+ ! not match the procedure (missing VALUE attribute)
+ ! However, this occures in real-world code calling
+ ! C routines where an interface is better than
+ ! "external" only.
+ interface
+   subroutine bmp_write(nx)
+     integer, value :: nx
+   end subroutine bmp_write
+ end interface
+contains
+   SUBROUTINE Grid2BMP(NX)
+     INTEGER, INTENT(IN) :: NX
+     if(nx /= 10) call abort()
+     call bmp_write(%val(nx))
+     if(nx /= 10) call abort()
+   END SUBROUTINE Grid2BMP
+END module x
+
+! The following test is possible and
+! accepted by other compilers, but
+! does not make much sense.
+! Either one uses VALUE then %VAL is
+! not needed or the function will give
+! wrong results.
+!
+!subroutine test()
+!    implicit none
+!    integer :: n
+!    n = 5
+!    if(n /= 5) call abort()
+!    call test2(%VAL(n))
+!    if(n /= 5) call abort()
+!  contains
+!    subroutine test2(a)
+!      integer, value :: a
+!      if(a /= 5) call abort()
+!      a = 2
+!      if(a /= 2) call abort()
+!    end subroutine test2
+!end subroutine test
+
+program main
+  use x
+  implicit none
+!  external test
+  call Grid2BMP(10)
+!  call test()
+end program main