Gdb backtrace of the failure can be found at $URL. While this is not a regression, I really thing we need to get rid of it for good. Plus, it's not just a test failure but a proper SEGV. If access to the Alpha dev machine is needed, don't hesitate to ask.
Please report this problem to upstream [1] and attach the backtrace in upstream issue. [1] http://bugs.python.org/
Added as Issue 9447 upstream: http://bugs.python.org/issue9447
This is now fixed (i.e. our b0rken patches have been backed out). @hppa can you confirm this for your arch?
Reverting the sed change in the ebuild fixes test_ctypes on hppa for python-2.6.5-r2. No more segv.
Marking this as resolved.