PING: Patch: mainline bootstrap failure (libobjc)

[Kai Henningsen]

doko@cs.tu-berlin.de (Matthias Klose)  wrote on 22.02.03 in <15959.40912.737545.113842@gargle.gargle.HOWL>:

> The first part of the patch is wrong or at least incomplete. Adding
> -I$(srcdir)/$(MULTISRCTOP)../boehm-gc/include helps for compilation
> only, it misses linking of the boehm-gc and installation of headers
> and libraries.
>
> If using the included boehm-gc is correct at all, is another
> question. See the thread starting at
>
> 	http://gcc.gnu.org/ml/gcc/2001-01/msg01019.html

Well, *that* change
1. needs more knowledge about gcc's build machinery *and* boehm-gc than I  
currently have
2. seems as if it would certainly need an assignment from the patch author

so I can't do that one, at least not right now.

In the mean time, the current patch *does* at least help the compilation.

> The patch to libobjc/gc.c seems to be a workaround. At least it seems
> to fix the errors seen in
>
> 	http://gcc.gnu.org/ml/gcc/2003-01/msg00007.html

That's what it's for. It's entirely along the same lines as a similar  
patch by Zack a while back ... ah yes, thread starting with
http://gcc.gnu.org/ml/gcc/2002-12/msg00023.html; I believe any fix for  
that one should also fix this one. From readinmg that thread, it seems gcc  
needs to have a way to pass down some properties of the target to the  
library. It currently hasn't, thus the workaround.

All this makes me wonder how --with-objc-gc ever made it into gcc. It  
seems rather untested.

MfG Kai