http://codereview.appspot.com/4591066/diff/3001/gcc/tree-threadsafe-analyze.c File gcc/tree-threadsafe-analyze.c (right):
http://codereview.appspot.com/4591066/diff/3001/gcc/tree-threadsafe-analyze.c#newcode1159 gcc/tree-threadsafe-analyze.c:1159: gcc_assert (false); On 2011/06/11 17:52:51, Diego Novillo wrote:
>+ else >+ gcc_assert (false); >+
Change to gcc_unreachable ();
Done. http://codereview.appspot.com/4591066/diff/3001/gcc/tree-threadsafe-analyze.c#newcode2151 gcc/tree-threadsafe-analyze.c:2151: && gimple_call_num_args(call) > 0) On 2011/06/11 17:52:51, Diego Novillo wrote:
2147 starting from GCC-4.5.). The clones could be created as
early as when
2148 constructing SSA. Also note that the parameters of a cloned
method
could 2149 be optimized away. */ 2150 if (TREE_CODE (TREE_TYPE (DECL_ORIGIN (fdecl))) == METHOD_TYPE 2151 && gimple_call_num_args(call) > 0)
Wouldn't it be easier to make fdecl == DECL_ORIGIN (fdecl) earlier in
the
function?
It's OK either way, though.
Yes, my original fix was to make fdecl = DECL_ORIGINAL (fdecl). But I later changed it to this way because I wanted to tolerate the case where the base object (i.e. "this" pointer) is an object instead of a pointer only when fdecl is a clone. (i.e. I don't want to arbitrarily relax it.) That's why I kept fdecl intact. http://codereview.appspot.com/4591066/
