From: Thomas Heller <theller@ctypes.org>
Date: Tue, 6 Jun 2006 18:50:46 +0000 (+0000)
Subject: Add a hack so that foreign functions returning float now do work on 64-bit
X-Git-Tag: v2.5b1~224
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=d8714dedfa9bf45ce11330c006c9479b418136f9;p=python

Add a hack so that foreign functions returning float now do work on 64-bit
big endian platforms.
---

diff --git a/Modules/_ctypes/callproc.c b/Modules/_ctypes/callproc.c
index 8163f49c2b..31c5f1028a 100644
--- a/Modules/_ctypes/callproc.c
+++ b/Modules/_ctypes/callproc.c
@@ -964,7 +964,12 @@ PyObject *_CallProc(PPROC pProc,
 	   address cannot simply be used as result pointer, instead we must
 	   adjust the pointer value:
 	 */
-	if (rtype->size < sizeof(ffi_arg))
+	/*
+	  XXX I should find out and clarify why this is needed at all,
+	  especially why adjusting for ffi_type_float must be avoided on
+	  64-bit platforms.
+	 */
+	if (rtype->type != FFI_TYPE_FLOAT && rtype->size < sizeof(ffi_arg))
 		resbuf = (char *)resbuf + sizeof(ffi_arg) - rtype->size;
 #endif