Re: PING: gimplifier ICE fix / broken GCC 4.0

[Helge Hess]

On Jan 25, 2005, at 9:53, Marcin Dalecki wrote:
> Just a note please don't forget about opengroupware, which is in fact 
> using ObjC as well.

Yes, we are very much concerned about the current discussion of getting 
a broken GCC 4.0. It basically implies that you won't be able to 
compile OpenGroupware.org without significant additional effort on 
upcoming Linux distributions.

I think there are currently three important free software projects 
using cc1obj (without an own fork, like Apple):
- Swarm
- OpenGroupware.org
- GNUstep
Together several millions of lines of free software code rendered 
useless by the breakage.

A comparison with gcj is a bit ridiculous since cc1obj was working just 
fine since something like 1990? And now, in 2005 a whole language is 
dropped and several projects are left in the rain just because the GCC 
core developers are unwilling to spend some thoughts and a few hours 
per year on fixing an issue raised by changes in C/C++?

Honestly I don't understand the technical issue at hand but I can 
hardly believe that the issue involves something which isn't trivial to 
fix by one of the core maintainers who broke cc1obj in the first place 
(absolutly *NO* offense intended here, don't get me wrong).

There was a discussion some months back on why the ObjC++ patches can't 
be applied to GCC. The (IMHO) correct justification was that the one 
who patches has to ensure that other frontends still work - just the 
usual don't-break-other-peoples-stuff which is regular practice in free 
software projects. I fail to see why this doesn't hold true in the 
other direction.

Having said that, I understand the points of Mark as well. I just 
assume that we are not talking about several days of development but 
about some good will by the C/C++ developers to acknowledge that there 
is another frontend.

Creating a seperate GCC community just for cc1obj is obviously 
nonsense. Getting into GCC is a life-time project while the fixes are 
just minor issues. Remember that cc1obj was working for years. We don't 
need to have it extended, we don't want to add additional 
functionality, we just want to ensure that changes in the C frontend do 
not break the ObjC one. That shouldn't be too much to ask for.

I *very* much hope that FSF will take some authority and step in and 
resolve this issue. Either by paying someone to fix the issue at hand 
or by modifying the release criteria for GCC. After all GCC and GNUstep 
are both FSF projects. Releasing GCC 4.0 without ObjC will hurt the 
free software world.

best regards,
  Helge
--
http://docs.opengroupware.org/Members/helge/
OpenGroupware.org