On 02/19/2011 06:00 PM, Lamar Owen wrote:
On Saturday, February 19, 2011 05:13:05 pm JohnS wrote:
maby "lzma" will attach to a few peoples built binaries then they want install.
I'm assuming you meant "won't install" so taking that assumption....
Echoing what has been said; read up on mock, and see how to go about bootstrapping from a set of source RPM's.
Then, download the buildsys-build from dev.centos.org, and run something like: rpm -qilp --provides --requires buildsys-build-0.5-6.el5.centos.7.noarch.rpm
Study the requires output, and note carefully that the RPM has no files in it.
Now, imagine you have a seed set of the basic binaries available to you (I don't know, from the RHEL6 public beta ISO's, maybe), and also imagine that the new C6 buildsys-build rpm has a requires for rpmlib(PayloadIsXz) <= 5.2-1 (that requires is from F14, but close enough).
You do need a version of mock that can unpack the seed binary packages and the source RPM's, and by extension all the built packages in the source package's buildreqs. And that means xz for C6.
However, it is true that the version of mock (and by extension the version of rpmbuild) used on the buildhost can impact the built package; see Karanbir's post on his blog on the subject: http://www.karan.org/blog/index.php/2010/07/16/rpms-built-on-el6beta2-might-...
And the kernel running on the buildhost is still the kernel running the chroot, and if a kernel interaction to the build process happens, well, that is a side-effect.
In a perfect world the buildhost environment outside the mock-constructed buildroot would be side-effect free; but is this true in practice? Karanbir's example seems to indicate that side-effects can happen inside the buildroot based on the buildsystem host's packages.....
There is a lot to understand when setting up a buildhost that's churning from SRPMS, and it's useful to read up, and then try it out and see what happens when you feed it a list of packages.....
I certainly have learned a few things during these recent threads that I wasn't aware of previously.
And JohnS, before you ask who Lamar is (he also does not have a centos.org e-mail address) ... he was the maintainer of the PostGreSQL RPMs that were distributed in Red Hat Linux for quite a few years before he turned those over, so he has also rebuilt a few SRPMS in his time.