"That's the way to do it"

In taking out a bug spotted by my regression tests in t/cmd/for.t
I actually managed to disable the entire optimisation. Which means
that I didn't find the other bug. This optimisation is live, and
passes all tests.

p4raw-id: //depot/perl@23109

1 files changed, 5 insertions, 5 deletions

diff --git a/op.c b/op.c
index d4c28874a7..8ceaa095ba 100644
--- a/op.c
+++ b/op.c
@@ -6751,17 +6751,17 @@ Perl_peep(pTHX_ register OP *o)
 	    if (!theirmark || theirmark->op_type != OP_PUSHMARK)
 		break;
 
-	    ourmark = ((LISTOP *)o)->op_first;
-	    if (!ourmark || ourmark->op_type != OP_PUSHMARK)
-		break;
-
-	    if (ourmark->op_next != o) {
+	    if (theirmark->op_sibling != o) {
 		/* There's something between the mark and the reverse, eg
 		   for (1, reverse (...))
 		   so no go.  */
 		break;
 	    }
 
+	    ourmark = ((LISTOP *)o)->op_first;
+	    if (!ourmark || ourmark->op_type != OP_PUSHMARK)
+		break;
+
 	    ourlast = ((LISTOP *)o)->op_last;
 	    if (!ourlast || ourlast->op_next != o)
 		break;