I have tried the suggested script (with a small modification to count number of failures) running for a few hours on one of the packages that had failed on the buildd and never saw cmake fail.
But then building a few packages locally with sbuild, apbs failed first time to build with the cmake failure. Second try at building apbs with sbuild succeeded. So it appears that there is something different about running in sbuild (or possibly under dpkg-buildpackage) that is triggering this failure on occassion. Cheers, Michael.
