RHEL7b1 is composed of 2520 srpms; Of these, some are arch specific to arch's we are not building (yet).
The x86_64 distro is made up of 8,520 binary rpms. Of these, 2,863 are noarch rpms and 1,919 are 32bit multilib. Leaving us with 3,738 x86_64 rpms that need to be built. Lets assume that building a clear x86_64 distro is our first goal here ( its not, but things are simpler with that assumption ).
As round-1 of the build cycle, we built all 2520 srpms against f19/32bit and then all of them again against f19/64bit; I use this result set as a backstore, to borrow from when we need to either break a dependency loop, or satisfy an interim build requirement locally. These packages are not otherwise a part of the regular buildroots used for round-2.
Round-2 is then a mass build attempt against r7b1/32bit and r7b1/64bit. Of these, nothing built against r7b1/32bit since large chunks of packages needed, even for the baseline buildroot are not published upstream. And needing to borrow from the f19/32bit backstore meant almost using a f19 buildroot, which isnt ideal. So lets put that aside for the time being.
The second part of round-2 was the x86_64 build attempt against r7b1/64bit. The results are a bit more promising there : 85% of the packages build. But only about 75% of those that build, look like the ones shipped by Red Hat as the 7b1 repos. I'm using percentages here, since using exact numbers for built-matched takes away from the aim of getting to complete. Now adding in the noarch builds, we get to : 5636 of 8520 built. Taking away the 1919 that are i686 rpms, we get a deficit of 965 that failed completely. These 965 represent 122 source rpms.
And this 122 set is going to be our next target. Alongside, I am going to start working on the blacklist/whitelists so we can start making the build roots public. Look for more info on that in the coming days, along with how everyone can contribute towards the 7 build effort.