This is an old problem with C++ parsing of function attribute that is
fixed in gcc-2.95.3. However the fix has not been incorportaed in
3.0.x branch. Here is Bob Wilson's patch that has been applied to
2.95.3 (at least in the cygwin and mingw instantiations):
http://gcc.gnu.org/ml/gcc-patches/2001-01/msg01751.html
Have a look at the assembly for the first example you sent. _malloc@4
looks a bit strange, doesn't it. That is why, in the past, windows.h
had to be included before
gcc/cp/decl.c has undergone some serious reorganisation. Following is
the above patch refreshed for gcc-3.0.1.
--- decl.c-3.0.1 Tue Sep 04 19:29:10 2001
+++ decl.c Tue Sep 04 19:38:45 2001
@@ -10469,7 +10469,16 @@ grokdeclarator (declarator, declspecs, d
ignore_attrs = 0;
else if (inner_attrs)
{
- decl_attributes (type, inner_attrs, NULL_TREE);
+ /* Create a dummy decl to pass to decl_attributes. The
+ attributes will be added to a variant of type, and this
+ new variant type can be retrieved from the dummy decl.
+ Passing type directly causes the attributes to be added
+ to type, which is wrong because type may be used
+ elsewhere without attributes. */
+
+ tree dummy = build_decl (TYPE_DECL, NULL_TREE, type);
+ decl_attributes (dummy, inner_attrs, NULL_TREE);
+ type = TREE_TYPE (dummy);
inner_attrs = NULL_TREE;
}
@@ -10988,7 +10997,18 @@ grokdeclarator (declarator, declspecs, d
if (inner_attrs)
{
if (! ignore_attrs)
- decl_attributes (type, inner_attrs, NULL_TREE);
+ {
+ /* Create a dummy decl to pass to decl_attributes. The
+ attributes will be added to a variant of type, and this
+ new variant type can be retrieved from the dummy decl.
+ Passing type directly causes the attributes to be added
+ to type, which is wrong because type may be used
+ elsewhere without attributes. */
+
+ tree dummy = build_decl (TYPE_DECL, NULL_TREE, type);
+ decl_attributes (dummy, inner_attrs, NULL_TREE);
+ type = TREE_TYPE (dummy);
+ }
else if (attrlist)
TREE_VALUE (attrlist) = chainon (inner_attrs, TREE_VALUE (attrlist));
else
Danny
Re: Problems executing programs which includes rpcndr.h
To: cygwin at cygwin dot com
Subject: Re: Problems executing programs which includes rpcndr.h
From: David Salotti <david at proximity dot com dot au>
Date: Tue, 4 Sep 2001 16:28:00 +1000
Organization: Proximity
References: <[EMAIL PROTECTED]>
--------------------------------------------------------------------------------
Ok, this is pretty weird - but if I rearrange the #include's, I am able
to run
the program successfully??? Does anyone know what's in rpcndr.h that's
conflicting with iostream??
the arrangement that works is:
#include <rpc.h>
#include <rpcndr.h>
#include <iostream>
ta
Dave
_____________________________________________________________________________
http://messenger.yahoo.com.au - Yahoo! Messenger
- Voice chat, mail alerts, stock quotes and favourite news and lots more!
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Bug reporting: http://cygwin.com/bugs.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
