[PATCH] Add checking that during RTL bbs don't mix EH and non-complex predecessor edges

Jakub Jelinek jakub@redhat.com
Fri Jun 15 19:19:00 GMT 2018


On Fri, Jun 15, 2018 at 09:06:16PM +0200, Jakub Jelinek wrote:
> The following patch fixes it by making sure we don't have EH landing pads
> that are reachable also by the crossing edges by splitting the EH landing
> pad bb we want to jump into and jumping (EDGE_CROSSING) only to the second
> half, leaving the first part with no actual instructions to be the EH
> landing pad.

The following patch adds verification that we don't have such basic blocks
during RTL optimizations.

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

Eric, I wonder if sjlj doesn't need a similar fix to the PR86108 patch.
If it does, I think with the following patch it would trigger a checking
failure.  Not familiar with sjlj EH enough to try it myself.

2018-06-15  Jakub Jelinek  <jakub@redhat.com>

	* cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb_has_eh_pred,
	verify all incoming edges are complex.

--- gcc/cfgrtl.c.jj	2018-04-09 20:15:51.237631514 +0200
+++ gcc/cfgrtl.c	2018-06-15 18:15:37.479333131 +0200
@@ -2540,15 +2540,15 @@ rtl_verify_edges (void)
 	    n_abnormal++;
 	}
 
-        if (!has_crossing_edge
-	    && JUMP_P (BB_END (bb))
-	    && CROSSING_JUMP_P (BB_END (bb)))
-          {
-	    print_rtl_with_bb (stderr, get_insns (), TDF_BLOCKS | TDF_DETAILS);
-            error ("Region crossing jump across same section in bb %i",
-                   bb->index);
-            err = 1;
-          }
+      if (!has_crossing_edge
+	  && JUMP_P (BB_END (bb))
+	  && CROSSING_JUMP_P (BB_END (bb)))
+	{
+	  print_rtl_with_bb (stderr, get_insns (), TDF_BLOCKS | TDF_DETAILS);
+	  error ("Region crossing jump across same section in bb %i",
+		 bb->index);
+	  err = 1;
+	}
 
       if (n_eh && !find_reg_note (BB_END (bb), REG_EH_REGION, NULL_RTX))
 	{
@@ -2606,6 +2606,18 @@ rtl_verify_edges (void)
 	  error ("abnormal edges for no purpose in bb %i", bb->index);
 	  err = 1;
 	}
+      if (bb_has_eh_pred (bb))
+	{
+	  FOR_EACH_EDGE (e, ei, bb->preds)
+	    {
+	      if (e->flags & EDGE_COMPLEX)
+		continue;
+	      error ("EH incoming edge mixed with non-complex incoming edges "
+		     "in bb %i", bb->index);
+	      err = 1;
+	      break;
+	    }
+	}
     }
 
   /* If there are partitions, do a sanity check on them: A basic block in


	Jakub



More information about the Gcc-patches mailing list