On 07/07/2010 09:49 AM, seth vidal wrote:
On Wed, 2010-07-07 at 09:39 -0400, seth vidal wrote:
>> Sorry, my bad, you're correct. rpmbuild does re-evaluate it, perhaps yum
>> does not when koji tries to populate the buildroot?
>
> I don't think yum can. It doesn't have any expansion capability for rpm
> macros.
Just did some tests on this:
In a spec file I added:
BuildRequires: foobar%{?_isa}
I rebuilt the srpm and I get:
rpm -qpR file.src.rpm
foobar(x86-32)
rpmlib(CompressedFileNames) <= 3.0.4-1
I opened the rpm using the rpm-python interface (what yum uses
internally):
hdr['requires']
['foobar(x86-32)', 'rpmlib(CompressedFileNames)']
So are you suggesting that yum should look for foobar(somearchspec) and
reinterpret those based on what the local system is?
Well, to be fair, I'm just trying to understand where the concern is here.
I'm not sure it makes sense for yum to reinterpret SRPM Requires. I
suspect there are some potential situations where a spec would want to
hardcode:
BuildRequires: foo(x86-32)
And wouldn't want yum re-evaluating it.
~spot