[testsuite] Fix false FAIL in gdb.cp/casts.exp

  Hi,

gcc-6.2.1-1.fc26.x86_64

gdb compile failed, /home/jkratoch/redhat/gdb-clean/gdb/testsuite/gdb.cp/casts.cc:40:10: error: expected primary-expression before 'int'
 decltype(int x)
          ^~~
/home/jkratoch/redhat/gdb-clean/gdb/testsuite/gdb.cp/casts.cc:40:10: error: expected ')' before 'int'
/home/jkratoch/redhat/gdb-clean/gdb/testsuite/gdb.cp/casts.cc:40:1: error: expected unqualified-id before 'decltype'
 decltype(int x)
 ^~~~~~~~
/home/jkratoch/redhat/gdb-clean/gdb/testsuite/gdb.cp/casts.cc: In function 'int main(int, char**)':
/home/jkratoch/redhat/gdb-clean/gdb/testsuite/gdb.cp/casts.cc:59:14: error: expected primary-expression before 'decltype'
   double y = decltype(2);
              ^~~~~~~~

'decltype' is a registered keyword since C++11 which is now a default for GCC.

OK for check-in?


Jan
gdb/testsuite/ChangeLog
2016-09-11  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* gdb.cp/casts.cc: Rename decltype to int_to_double.
	* gdb.cp/casts.exp (whatis decltype(5)): Rename to ...
	(whatis int_to_double(5)): ... here.
  

On 09/11/2016 03:10 PM, Jan Kratochvil wrote:
> --- a/gdb/testsuite/gdb.cp/casts.exp
> +++ b/gdb/testsuite/gdb.cp/casts.exp
> @@ -112,7 +112,7 @@ gdb_test "print reinterpret_cast<A &> (*b)" " = \\(A \\&\\) @$hex: {a = 42}" \
>  
>  # Test that keyword shadowing works.
>  

This comment suggests that the test actually uses decltype on purpose.

> -gdb_test "whatis decltype(5)" " = double"
> +gdb_test "whatis int_to_double(5)" " = double"
>  

Seems to be exercising the FLAG_SHADOW bits:

...
    {"__typeof__", TYPEOF, OP_TYPEOF, 0 },
    {"__typeof", TYPEOF, OP_TYPEOF, 0 },
    {"typeof", TYPEOF, OP_TYPEOF, FLAG_SHADOW },
    {"__decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX },
    {"decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX | FLAG_SHADOW },
...

/* This is used to associate some attributes with a token.  */

enum token_flag
{
...
  /* If this bit is set, the token is conditional: if there is a
     symbol of the same name, then the token is a symbol; otherwise,
     the token is a keyword.  */

  FLAG_SHADOW = 2
};

So perhaps a better fix is to move that particular test to a
separate testcase that force-compiles with -std=c++03.

Adding Tromey, who wrote the test initially.

Thanks,
Pedro Alves
  

>>>>> "Pedro" == Pedro Alves <palves@redhat.com> writes:

Pedro> So perhaps a better fix is to move that particular test to a
Pedro> separate testcase that force-compiles with -std=c++03.

Pedro> Adding Tromey, who wrote the test initially.

Yeah, I think that's correct.  According to the patch email, I did this
because there wasn't a way to tell which version of C++ was in use:

https://sourceware.org/ml/gdb-patches/2012-07/msg00349.html

Tom