Jos Vos wrote:
On Sat, Feb 12, 2005 at 10:14:14AM -0500, Jeff Johnson wrote:
THe issue with pango is perhaps subtle.
In a multilib environment, pango was one of about 6 or so packages that had to change because of path collisions.
The fix was to create a per-platform directory.
How that directory is named, and how the path is determined, have nothing
whatsoever to do with the original problem that needed to be solved, i.e.
changing the path so that components were not clobbered in a multilib environment.
All fine, but my answer is still not answered (unless I'm missing the point in your answer): why is %_host evaluated to i386-redhat-linux-gnu in the RH build environment and to i686-redhat-linux-gnu on a standard Fedora/RHEL system.
Is this an intentional, explicit setting in the RH environment, or is it done for a few packages only, or...
All depends on how build systems are configured. Get a copy of rpm --showrc from a RH build, and configure your build machine the same way.
There are as few overrides and "secret sauce" as possible in beehive, by intent.
OTOH, arch happens to be one configuration parameter that is overridden frequently.
There are certainly no per-package unique build configuration flags by intent. OTOH,
configurations can and do rot, and there is the occaisional need to expedite a package
build with a hack to get the job done.
HTH
73 de Jeff
