Glibc and installation of locale

Ken Moffat ken at linuxfromscratch.org
Sun Nov 20 15:37:14 PST 2005


On Tue, 15 Nov 2005, Ken Moffat wrote:

> On Tue, 15 Nov 2005, go moko wrote:
>
>> 
>> Excuse me, I thought it was a following, but no...
>> Architecture: x86_64 multilib, and effectively, it's
>> the 32 bits built which make errors during the check.
>> With locales, some errors stay, but a lot fewer.
>
> Thanks.  My multilib x86_64 is still severely bogged down trying to build 
> some 32-bit bclfs [oh, how I hate libtool! ] ;)  so I've added this to my 
> list of things to investigate (along with alsa, revisiting the hfsutils patch 
> for ppc, and G5 issues - don't hold your breathe waiting!).  It appears as if 
> the testsuites for each architecture/flavour are different, which I find 
> somewhat strange.
>
> The questions are: which locales do we need to install into /tools, for 
> which architectures, purely so that the final glibc testsuites will not fail 
> specific tests.  Worst case, this ought to be far fewer locales than are 
> needed for all the tests in the final system.
>
  A further datapoint - on powerpc64 multilib there are somewhat similar 
failures in glibc-32 (first in catgets/de, then all over the place). 
After installing that *without any locales* the glibc-64 testsuite 
finishes without any problems.  This seems to be similar to the x86_64 
multilib situation.  So, we might need DE_de in /tools for multilib.

  What really puzzles me is why the 32-bit tests are different - or has 
somebody trashed those too in the few weeks since I last built a 32-bit 
system ?

  And also, why is 'hello world' built with /tools/bin/gcc -m32 *bigger* 
than when it's built with -m64 ?  Weird.

Ken
-- 
  das eine Mal als Tragödie, das andere Mal als Farce


More information about the cross-lfs mailing list