summaryrefslogtreecommitdiff
path: root/gcc/testsuite/gfortran.dg/keyword_symbol_1.f90
diff options
context:
space:
mode:
Diffstat (limited to 'gcc/testsuite/gfortran.dg/keyword_symbol_1.f90')
-rw-r--r--gcc/testsuite/gfortran.dg/keyword_symbol_1.f9057
1 files changed, 57 insertions, 0 deletions
diff --git a/gcc/testsuite/gfortran.dg/keyword_symbol_1.f90 b/gcc/testsuite/gfortran.dg/keyword_symbol_1.f90
new file mode 100644
index 000000000..2661897ff
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/keyword_symbol_1.f90
@@ -0,0 +1,57 @@
+! { dg-do compile }
+! This tests the fix for PR28526, in which a public interface named
+! 'end' would be treated as a variable because the matcher tried
+! 'END INTERFACE' as an assignment and left the symbol modified in
+! failing. The various pitfalls that were encountered in developing
+! the fix are checked here.
+!
+! Contributed by Paul Thomas <pault@gcc.gnu.org>
+!
+module blahblah
+ public function, end
+
+! The original PR from Yusuke IGUCHI <iguchi@coral.t.u-tokyo.ac.jp>
+ interface end
+ module procedure foo1
+ end interface
+
+! A contribution to the PR from Tobias Schlueter <tobi@gcc.gnu.org>
+ interface function
+ module procedure foo2 ! { dg-error "is neither function nor" }
+ end interface
+
+ interface function
+ module procedure foo3
+ end interface
+
+ interface
+ function foo4 ()
+ real foo4
+ x = 1.0 ! { dg-error "in INTERFACE" }
+ end function foo4
+ end interface
+
+ interface
+ x = 2.0 ! { dg-error "in INTERFACE block" }
+ function foo5 ()
+ real foo5
+ end function foo5
+ end interface
+
+ x = 3.0 ! { dg-error "in MODULE" }
+
+contains
+
+ subroutine foo1
+ end subroutine foo1
+
+ function foo2 ! { dg-error "Expected formal argument list" }
+ foo2 = 0 ! { dg-error "already been host associated" }
+ end function foo2 ! { dg-error "Expecting END MODULE" }
+
+ function foo3 ()
+ real foo3
+ end function foo3
+
+ x = 4.0 ! { dg-error "in CONTAINS section" }
+end module blahblah