I also see something like this for tokyocabinet, see #813880 for an analysis...
I see those failures on powerpc, ppc64 and s390x, but not on mips. So this might be indeed a differnt bug as well, but I guess related to optimization. After seeing this bug, I tried -fno-expensive-optimizations on partch.d.o -- and this fixes also the regression I was experiencing. Comparing the buildd-logs, I see it not failing with version gcc-5_5.3.1-7 (https://buildd.debian.org/status/fetch.php?pkg=tokyocabi net&arch=powerpc&ver=1.4.48-5&stamp=1453879293) and the last-non-failing one was using gcc-5_5.3.1-4 (https://buildd.debian.org/status/fetch.php?pkg=tokyocab inet&arch=powerpc&ver=1.4.48-4&stamp=1451741581) -- tobi