The configure script looks for a working C stack overflow detection, which currently fails on amd64. Reproducible: Always I saw that there was a bug report filed upstream on this issue a few months ago, but it still has not been corrected yet, and in order to get a fully working system using gcc-4.5 we need to get around this issue. I would expect this issue to be resolved by gcc-4.5 final, but here is the workaround nonetheless.
Created attachment 219479 [details, diff] diffutils-2.8.7-gcc45.patch workaround for this (temporary?) bug
Created attachment 219481 [details] diffutils-2.8.7-r2-build.log
Patch should be for configure.ac or configure.in, since configure is only autoconf generated file.
Created attachment 219515 [details, diff] diffutils-2.8.7-gcc45-autoconf.patch If the ebuild were to use autoconf, which it currently doesn't, this would be a workaround.
Actually this is fixed in 2.9 which was just released yesterday.
Comment on attachment 219479 [details, diff] diffutils-2.8.7-gcc45.patch this is just silly. there's no point in patching a lot of crap out when the code provides perfectly usable cache variables like ac_cv_sys_xsi_stack_overflow_heuristic. simply export it with a value of 'yes' (or 'no' as the situation requires).
diffutils-2.9 now in the tree