Re: [C++ PATCH] Avoid -Wreturn-type warnings if a switch has default label, no breaks inside of it, but is followed by a break (PR sanitizer/81275)

[Jakub Jelinek <jakub@redhat.com>]

On Fri, Nov 24, 2017 at 10:59:53PM +0100, Jakub Jelinek wrote:
> The testcase below has a useless break; that causes a bogus -Wreturn-type
> warning.  The C++ FE already has code to avoid adding a BREAK_STMT
> after a return or similar sequence that is known not to return.
> The following patch extends block_may_fallthrough to also return false
> for SWITCH_STMTs that can't fall through.
> 
> Those are ones with non-empty body where the whole body can't fallthrough,
> additionally they need to have a default: case label (or cover the whole
> range of values, but that is not what this patch can compute, that would
> be too big duplication of the gimplification processing) and no BREAK_STMT.
> 
> For the default: case label we need to look in all SWITCH_BODY children
> except for nested SWITCH_STMTs, for BREAK_STMTs also not in
> {FOR,DO,WHILE}_BODY.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

Actually, thinking about it some more, maybe it would be more efficient
to gather this information during construction of the SWITCH_STMT in some
new flag on the tree, so cxx_block_may_fallthru would just:

      case SWITCH_STMT:
	if (SWITCH_STMT_CANT_FALLTHRU_P (stmt)
	    && SWITCH_STMT_BODY (stmt)
	    && !block_may_fallthru (SWITCH_STMT_BODY (stmt)))
	  return false;
	return true;

and
/* Set if the body of a switch stmt contains a default: case label
   and does not contain any break; stmts, thus if SWITCH_STMT_BODY
   is not empty and doesn't fallthru, then the whole switch stmt
   can't.  */
#define SWITCH_STMT_CANT_FALLTHRU_P(NODE) \
  TREE_LANG_FLAG_0 (SWITCH_STMT_CHECK (NODE))

Seems the C++ FE already has switch_stack, so we could just add there
a has_default_p, has_break_stmt_p and inside_loop_p flags and both
inside of templates and outside in finish_case_label, in
finish_break_stmt if actually adding BREAK_STMT and when entering
a body of a FOR/DO/WHILE loop tweak those flags.  Seems switch_stack
is also maintained during pt.c, but we should compute it both during
parsing and during pt.c (start with the bit clear on a new SWITCH_STMT).

Thoughts on this?

Guess for the C FE we should similarly handle SWITCH_EXPR, though a break;
in there is a goto outside of the switch, so all we need to note is a
default: case label.

	Jakub