[pypy-commit] pypy default: Follow-up for dfb609e46b95? I think there are reasons for always

[arigo]

Author: Armin Rigo <arigo at tunes.org>
Branch: 
Changeset: r80813:3baa25ce21e8
Date: 2015-11-21 08:28 +0100
http://bitbucket.org/pypy/pypy/changeset/3baa25ce21e8/

Log:	Follow-up for dfb609e46b95? I think there are reasons for always
	returning an UnrolInfo() with extra_same_as instead of inventing an
	empty list. I missed this 4th place... But this is even worse
	because I can't design a test *and* I don't have a crash. (Issue
	#2185 is around there but I think it might not be related.)

diff --git a/rpython/jit/metainterp/optimizeopt/unroll.py b/rpython/jit/metainterp/optimizeopt/unroll.py
--- a/rpython/jit/metainterp/optimizeopt/unroll.py
+++ b/rpython/jit/metainterp/optimizeopt/unroll.py
@@ -180,7 +180,7 @@
 
         if not inline_short_preamble:
             self.jump_to_preamble(celltoken, end_jump, info)
-            return (UnrollInfo(target_token, label_op, [],
+            return (UnrollInfo(target_token, label_op, extra_same_as,
                                self.optimizer.quasi_immutable_deps),
                     self.optimizer._newoperations)