See http://koji.fedoraproject.org/koji/taskinfo?taskID=810128 and http://koji.fedoraproject.org/koji/taskinfo?taskID=810005 for examples.
I've seen these failures on the xenbuilders but not on the x86 hosts. Possibly an infrastructure issue, possibly cmake using much more memory than it should (or both). For now, re-submit and hopefully you won't get a xenbuilder.
You seem to be right. If it does not hit a xen builder the job goes through w/o problems.
Dang, I had seen this too, but had thought that the parens patch would help here, guess not. :(
This is basically caused by popen() not working on xen kernels. *** This bug has been marked as a duplicate of bug 453394 ***
Infrastructure ticket: https://fedorahosted.org/fedora-infrastructure/ticket/811