[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-devel
Subject:    Re: [patch] fix for compiling arts with Red Hat gcc 2.96
From:       Stefan Westerfeld <stefan () space ! twc ! de>
Date:       2001-04-15 13:53:28
[Download RAW message or body]

   Hi!

The patch is in the CVS already (Waldo has applied it). I see that it would
be inconvenient to remove it again to quite some KDE users. In an ideal
world I would revert the patch (because it is /not/ the right thing to do).

But well, the world isn't ideal, and the workaround isn't that bad for the
code, so I'll leave it as it is now.

   Cu... Stefan

On Wed, Apr 11, 2001 at 09:16:56PM +0000, Hetz Ben Hamo wrote:
> I've upgraded to gcc-2.96-80 which compiles now the arts/mcop dir (tested 
> yesterday).

It's hard to be sure if it is really the compiler that is fixed, as Waldo
applied the patch just about then:

----------------------------
revision 1.47
date: 2001/04/10 19:57:48;  author: waba;  state: Exp;  lines: +11 -11
Seems like Red Hat's gcc can't access anonymous unions through
references.
Patch by Daniel Karlsson <chaotica@home.se>
----------------------------

Here is a test case for the compiler bug:

struct F {
	int i;
	F& operator=(const F& other) { i = other.i; }
};

struct G {
	union { char c; char d; }; // if you name the union u, it will work
	F f;					   // if you remove this line, it will work, too
};

main()
{
	G g1,g2;
	g1 = g2;
}

> You can't switch GCC just like that - I tried to move back to GCC 2.95 and 
> recompile the entire KDE with it - it fails in the configure script.
> 
> So that leaves us - the Redhat users - with a big problem. I cannot compile 
> KDE with kgcc, and you refuse to put patches for letting us compiling mcop.
> 
> Do you think we're alone at this? think again! Mandrake 8.0 will have the 
> EXACT gcc that Redhat 7.1 will have (gcc 2.96) - so add this into account 
> when you refuse to add support for gcc 2.96.
> 
> I wish the GCC team would move their butt and release 3.0 (they even created 
> a 3.1 branch) - but they're not doing this - this makes gcc development speed 
> makes Mozilla development fast as a rabbit!
> 
> Hetz
> 
> > It's a compiler bug (which present in gcc 2.96, but not in gcc 2.95.2
> > or latest gcc CVS). I don't think it's a good idea to put workarounds
> > for other peoples bugs into my code, because bugs should be fixed where
> > they are caused, not where they break something else.
> >
> > The whole idea of RedHat releasing an immature gcc-2.96 should be getting
> > the remaining issues fixed rather than adapting other code to a new broken
> > subset of C++.
> >
> >    Cu... Stefan
>  
> >> Visit http://master.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<

-- 
  -* Stefan Westerfeld, stefan@space.twc.de (PGP!), Hamburg/Germany
     KDE Developer, project infos at http://space.twc.de/~stefan/kde *-         
 
>> Visit http://master.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<

[prev in list] [next in list] [prev in thread] [next in thread]