Re: PING: gimplifier ICE fix

[Lars Sonchocky-Helldorf]

Sorry for the late reply, but things have evolved since then.


At first I want to thank you that you've taken the time to review this 
patch. I know that you probably did that in your spare time and I 
appreciate that. However as things turned out to be I'll ask you to 
reconsider your decision. See below.


Am Sonntag, 12.12.04 um 23:48 Uhr schrieb Richard Henderson:

> On Fri, Dec 10, 2004 at 08:37:33PM +0100, Lars Sonchocky-Helldorf 
> wrote:
> > This one was proposed about a month ago and blocks my efforts of
> > getting GNUstep build with gcc-4.0:
> >
> > http://gcc.gnu.org/ml/gcc-patches/2004-11/msg00925.html
> >
> > Gimplifier folks, please approve (or at least: comment).
>
> This patch is wrong.
>
> The problem is an inconsistency in the objc front end.
>
> It claims that (List*) and (List_linked*) are compatible
> (aka identical) types, but that List and List_linked aren't.
>
> I have no idea what the objc front end is after here, so
> I can't provide any advice.

You're absolutely right here, the ObjC Front end has a shortcoming 
here. But there currently seems to be no other solution to this problem 
according to the person who provided that patch:

http://gcc.gnu.org/ml/gcc-patches/2005-01/msg01423.html :

"RTH is right in that the ObjC type system representation is 
inconsistent, but that is really orthogonal to whether the patch I 
proposed is safe or not. A rewrite of the type representation (to bring 
it closer in line with C++/Java) is planned, but in the gcc-4.1 time 
frame at the earliest."

and:

http://gcc.gnu.org/ml/gcc-patches/2005-01/msg01434.html:

"Does that mean - in other words - gcc-4.0 will be of no use for the 
GNUstep people or will you provide an interim fix?

The interim fix is the one that we're currently discussing (and which 
has been rejected)."



In other words that means that - if this patch isn't approved - GCC 4.0 
would ship with a broken Objective-C compiler:

GCC is in it 4.0 incarnation currently not able to compile GNUstep, see:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18408

and more clearly in its Duplicates:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18771 and
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19553

Since GNUstep is an official GNU project, GNUstep's needs should get at 
least some attention.



My proposition is:

Approve the patch on a timed base, that means approve it for now (so to 
say as a temporary band aid) until the Objective-C rewrite as proposed 
by Ziemowit Laski is done , annotate it with a FIXME comment for that 
purpose.

Would that - given that this patch doesn't cause trouble anywhere else 
- be o.k. for you?

> r~

regards, Lars