From: Armin Rigo <arigo@tunes.org>
Date: Fri, 7 Jan 2005 18:10:51 +0000 (+0000)
Subject: Re-running python with/without the -Qnew flag uses incorrectly optimized
X-Git-Tag: v2.5a0~2164
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=664b43b3f4cbb7a63d5d67117e1d040a286af371;p=thirdparty%2FPython%2Fcpython.git

Re-running python with/without the -Qnew flag uses incorrectly optimized
bytecodes from the previously saved .pyc files.  Fixed by disabling the static
optimization of BINARY_DIVIDE between two constants.
---

diff --git a/Python/compile.c b/Python/compile.c
index 15c843614977..545042f87c0a 100644
--- a/Python/compile.c
+++ b/Python/compile.c
@@ -468,12 +468,9 @@ fold_binops_on_constants(unsigned char *codestr, PyObject *consts)
 		newconst = PyNumber_Multiply(v, w);
 		break;
 	case BINARY_DIVIDE:
-		if (!_Py_QnewFlag) {
-			newconst = PyNumber_Divide(v, w);
-			break;
-		}
-		/* -Qnew is in effect:	fall through to
-		   BINARY_TRUE_DIVIDE */
+		/* XXX care is needed to fold this operation statically:
+		the result might depend on the run-time presence of the -Qnew flag */
+		return 0;
 	case BINARY_TRUE_DIVIDE:
 		newconst = PyNumber_TrueDivide(v, w);
 		break;