c++/6074 is a regression from 2.95

c++/6074 is a regression from 2.95

[Wolfgang Bangerth]

c++/6074 is about the following:
--------------------------------------
extern
void foo(void);
void foo(void) __attribute__ ((weak)) __attribute__ ((alias ("__foo")));
--------------------------------------

In 2.95 and present 3.3 CVS, this is accepted by both the C and C++ 
compilers. In 3.2, C++ rejects this:
    weak.cc:3: `void foo()' defined both normally and as an alias
The C compiler accepts it.

Technically, this is a regression, but a very weak (pun intended) one. 
Since it's already fixed in 3.3, I'll leave it up to others whether this 
merits fixing, or what else should happen to this report.

Regards
  Wolfgang

-------------------------------------------------------------------------
Wolfgang Bangerth              email:           bangerth@ticam.utexas.edu
                               www: http://www.ticam.utexas.edu/~bangerth

Re: c++/6074 is a regression from 2.95

[Nathan Sidwell]

Wolfgang Bangerth wrote:
> c++/6074 is about the following:

> Technically, this is a regression, but a very weak (pun intended) one. 
> Since it's already fixed in 3.3, I'll leave it up to others whether this 
> merits fixing, or what else should happen to this report.
If you want to, see if you can track down the patch which fixed it. I'm
fairly certain this bug report has popped up more than once, indicating
real code contains such stuff.

nathan

-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
          'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org

Re: c++/6074 is a regression from 2.95

[Mark Mitchell]

--On Friday, October 25, 2002 06:38:30 PM -0400 Jim Wilson 
<wilson@redhat.com> wrote:

>> Would you mind committing it there?
>
> Done.

Thanks!

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: c++/6074 is a regression from 2.95

[Jim Wilson]

>Would you mind committing it there?

Done.

Jim

Re: c++/6074 is a regression from 2.95

[Nathan Sidwell]

Mark Mitchell wrote:
>> 6930 is a definite duplicate

> I've marked 6074 as high-priority and attached the fix you found; once
> that gets tested we'll check it in.
These email's are a little vague as to who's doing/done this. For the
record, I'm building and testing it now, and will commit if it is ok.

Wolfgang, thanks for finding the right one.

nathan

-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
          'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org

Re: c++/6074 is a regression from 2.95

[Mark Mitchell]

--On Friday, October 25, 2002 01:40:21 PM -0400 Jim Wilson 
<wilson@redhat.com> wrote:

> As far as I know, this patch is safe for gcc-3.2.1, and would be useful
> there.

Jim --

Would you mind committing it there?

Thanks,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: c++/6074 is a regression from 2.95

[Jim Wilson]

As far as I know, this patch is safe for gcc-3.2.1, and would be useful there.

Jim

Re: c++/6074 is a regression from 2.95

[Mark Mitchell]

> 6930 is a definite duplicate

I've marked 6074 as high-priority and attached the fix you found; once
that gets tested we'll check it in.

Thanks,

-- 
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com

Re: c++/6074 is a regression from 2.95

[Wolfgang Bangerth]

This patch (thanks, Jan, for the pointer)
  revision 1.944
  date: 2002/10/10 22:29:05;  author: wilson;  state: Exp;  lines: +0 -1
  Fix C++ FE attribute alias failure.
  	  * decl.c (duplicate_decls): Don't call decl_attributes.

fixed c++/6074 and some duplicates on the mainline. I append it below. Can 
someone consider moving it over to the 3.2 branch? I confirmed that it 
fixes the regression there also, but I don't know whether it might have 
some other adverse effects there.

Regards
  Wolfgang

Index: decl.c
===================================================================
RCS file: /cvsroot/gcc/gcc/gcc/cp/decl.c,v
retrieving revision 1.943
retrieving revision 1.944
diff -u -r1.943 -r1.944
--- decl.c      9 Oct 2002 21:27:37 -0000       1.943
+++ decl.c      10 Oct 2002 22:29:05 -0000      1.944
@@ -3422,7 +3422,6 @@
      except for any that we copy here from the old type.  */
   DECL_ATTRIBUTES (newdecl)
     = (*targetm.merge_decl_attributes) (olddecl, newdecl);
-  decl_attributes (&newdecl, DECL_ATTRIBUTES (newdecl), 0);

   if (TREE_CODE (newdecl) == TEMPLATE_DECL)
     {


-------------------------------------------------------------------------
Wolfgang Bangerth              email:           bangerth@ticam.utexas.edu
                               www: http://www.ticam.utexas.edu/~bangerth

Re: c++/6074 is a regression from 2.95

[Jan Vroonhof]

[This is a hand crafted reply based on the mailing list archive]

 > If you want to, see if you can track down the patch which fixed it. I'm
 > fairly certain this bug report has popped up more than once,

Yes, I even used a patch included in one of them to "fix" it on 3.2. I
found it through one of the eCos lists. I can't find back to
though. Maybe it was from a message on gcc-patches?

I still have the GCC tree for that though. I have attached the diff at
the end.

6930 is a definite duplicate

 > indicating
 > real code contains such stuff.

This code is used in the eCos C library (which is built using C++)
.There is a work around on the eCos side as well, but AFAIK that is
not in the eCos CVS.

Jan

cvs server: Diffing .
Index: decl.c
===================================================================
RCS file: /cvsroot/gcc/gcc/gcc/cp/decl.c,v
retrieving revision 1.866.2.36.2.2
diff -u -u -r1.866.2.36.2.2 decl.c
--- decl.c	6 Oct 2002 23:02:09 -0000	1.866.2.36.2.2
+++ decl.c	25 Oct 2002 10:26:05 -0000
@@ -3500,7 +3500,6 @@
      except for any that we copy here from the old type.  */
   DECL_ATTRIBUTES (newdecl)
     = (*targetm.merge_decl_attributes) (olddecl, newdecl);
-  decl_attributes (&newdecl, DECL_ATTRIBUTES (newdecl), 0);
 
   if (TREE_CODE (newdecl) == TEMPLATE_DECL)
     {