https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87084
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |accepts-invalid
Status|UNCONFIRMED |NEW
Last reconfirmed| |2018-08-24
Component|other |c
Version|unknown |8.2.1
Ever confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
It's because of
rtx
expand_builtin (tree exp, rtx target, rtx subtarget, machine_mode mode,
int ignore)
{
...
case BUILT_IN_VA_ARG_PACK:
/* All valid uses of __builtin_va_arg_pack () are removed during
inlining. */
error ("%Kinvalid use of %<__builtin_va_arg_pack ()%>", exp);
return const0_rtx;
case BUILT_IN_VA_ARG_PACK_LEN:
/* All valid uses of __builtin_va_arg_pack_len () are removed during
inlining. */
error ("%Kinvalid use of %<__builtin_va_arg_pack_len ()%>", exp);
return const0_rtx;
and diagnostics of valid/invalid uses of these builtins should really happen
in the FE (they may be only used in variadic functions and those functions
need to be marked always-inline either by the user or the frontend).
So - marking as C-family FE bug. We should reject this testcase and
the above code should be changed to ICE.
