On 2012-08-27 03:20, Ken Moffat wrote:
> On Sun, Aug 26, 2012 at 06:51:37PM -0700, Bryan Kadzban wrote:
>>
>> Uh, isn't that backwards? :-)
>>
>> BUILD_CC is the CC to use to build programs that will be executed on
>> the
>> cross-compilation *source* architecture, not the destination.
>> ${LFS_TGT}
>> is the destination arch. The glibc sources are trying to build this
>> program for the source.
>>
> And we think we don't want it, because "we never used to need this,
> grumble, grumble" ;-)
>
>> This actually also means that host includes (and libs) are
>> *absolutely*
>> required to be here. This output doesn't get linked into the cross
>> glibc, so this is *not* actually a contamination from the host
>> system.
>> (Only the output of rpcgen, after it gets built and run, is used.)
>>
>> This does mean, however, that the host does really need the
>> rpc/types.h
>> header file installed, so on systems that don't have it, glibc won't
>> build...
>>
>>
> I'm glad someone here understands this stuff (I certainly don't),
> but
> I'm still expecting to see moans about this new requirement,
>
> I've just had another go, reverting first
> commit 656416c96528437f54b345b1470aaa40f145ed30
> Author: Andreas Schwab <sch...@linux-m68k.org>
> Date: Sat May 12 08:40:20 2012 +0200
>
> Fix dependency tracking on cross-rpcgen-objs
>
> and then the other commit. Strangely, the Changelog parts of the
> commits don't revert, but the rest does. I've now got in to gcc
> pass 2 (without the extra sed or the CC override). Still with the
> rpc headers pushed out of the way.
>
> I'll keep this running to see if it completes.
>
> I don't _think_ that the host's rpc headers are critical at this
> stage, but I'm not a toolchain expert.
>
> Even if my revert works, we don't really want to keep carrying this
> for our own "idiosyncratic" pure build - it's the sort of thing that
> will be fun to fix whenever that part of the code gets changed.
> But, FWIW, I'll attach it - it's the sort of thing that will be fun
> to fix whenever that part of the code gets changed.
>
> ĸen
I'll be posting logs and other thoughts later on, but for now we need
to be looking at *why* any part of ch5 after and including glibc is
using host headers full stop.
--
Jasmine Iwanek
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
