[PATCH] Fix yet another expand_debug_expr ICE (PR middle-end/83694)

[Jakub Jelinek]

Hi!

The recently added testcase for PR83666 ICEs on powerpc*/sparc*
and perhaps other targets, where get_inner_reference gives us
VOIDmode for a 512-bit field and smallest_int_mode_for_size
ICEs on it because there is no such integer mode.

Fixed by giving up above MAX_BITSIZE_MODE_ANY_INT, that would have been
BLKmode and not really useful for debug info generation anyway.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2018-01-05  Jakub Jelinek  <ja...@redhat.com>

        PR middle-end/83694
        * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
        and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.

--- gcc/cfgexpand.c.jj  2018-01-04 12:43:38.199537090 +0100
+++ gcc/cfgexpand.c     2018-01-05 12:10:46.960037601 +0100
@@ -4534,8 +4534,12 @@ expand_debug_expr (tree exp)
        if (MEM_P (op0))
          {
            if (mode1 == VOIDmode)
-             /* Bitfield.  */
-             mode1 = smallest_int_mode_for_size (bitsize);
+             {
+               if (maybe_gt (bitsize, MAX_BITSIZE_MODE_ANY_INT))
+                 return NULL;
+               /* Bitfield.  */
+               mode1 = smallest_int_mode_for_size (bitsize);
+             }
            poly_int64 bytepos = bits_to_bytes_round_down (bitpos);
            if (maybe_ne (bytepos, 0))
              {

        Jakub