On 24-09-15 09:14, Fabian Arrotin wrote: > On 24/09/15 09:11, Jacco Ligthart wrote: > > >> I so had a quick look and here are the buildrequires (each time > >> with > other build requires for package) for > hwloc: > > >> libibverbs > > > valgrind > openmpi > valgrind (and other ones, like > > libiverbs too) => circular > build dep > > So let me then just try > > to > * rebuild openmpi (from f19, against f19+c7) > * inject it in > > buildroot (_staging) > * try to then rebuild valgrind, libibverbs, > > hwloc, etc > > > what you probably want is: * rebuild openmpi (from *c7*, against > > f19+c7) > > > Jacco > > Hi Jacco, > > I like the idea of just rebuilding the previous version (f19) in that > case in a staging buildroot, just to satisfy the other buildrequires > for that package (but with the correct c7 version), so that the c7 > srpm itself is built against the c7-buildroot/pass-1 (and not having > to rely on f19 anymore) : so that staging repo is just temporary hmm, not sure if we mean the same or not. I'm also in favor of a staging area. So what I did when building armv5tel is (when needed) build C7 srpms against a mix of already build C7 RPMS, some hand picked F18 (latest armv5 fedora) and sometimes some manual adjustments. The result would then (hopefully) be a RPM. These RPMS would go into staging, only to be used to build release worthy RPMS. The only RPMS that got into release were those that were exclusively build against C7 content. I guess that I just don't see the point in rebuilding F19 SRPMS. It might however depend on the version differences we're talking about. if the versions are for all practical reasons the same, than it does not really matter. J.