On Friday 18 August 2006 14:05, Jesse Keating wrote:
On Friday 18 August 2006 14:48, Clark Williams wrote:
> If I'm correct here, the problem is that you need both 32-bit and 64-bit
> glibc's in the same chroot to build gcc for a sparc. I looked on a
> x86_64 FC5 system and it has two glibc-2.4-8 packages, but if you list
> them both, one is the 64 bit version and the other is the 32-bit
> version. This implies to me that they were both "installed" as opposed
> to "updated". I can't really see an easy way to force mock/yum to do
> this, so I think the only way will be multiple invocations of mock and
> rpm. Something like:
We handle this at Red Hat by having a file buildrequirement on a file that
is provided by the 32bit glibc, as well as a buildrequirement on just
glibc. This way no matter what arch you're on, you'll get the 32bit glibc.
Now, at Red Hat we had to create a fake 'glibc32' package that is built
x86_64 so that it is available in the x86_64 repo. Perhaps you can do the
same for sparc?
perhaps, but it needs to be the other way around. that is i need
the 64 bit
glibc in a 32bit chroot to build gcc. so i guess i need to build a
glibc64 'fake' package gcc has a BuildRequires
/lib64/libc.so.6 /usr/lib64/libc.so so its file based
I will work out what i need to do for the glibc64 package
Dennis