On Thu, Jan 23, 2014 at 12:12:23PM -0800, Jonathan Nieder wrote: > Jeff King wrote: > > On Thu, Jan 23, 2014 at 11:52:06AM -0800, Jonathan Nieder wrote: > > >> My main worry about the patches is that they will probably run into > >> an analagous problem to the one that v1.7.12-rc0~1^2~2 > [...] > > I think this probably works in practice because align_ntohl is inlined, > > and any sane compiler will never actually load the variable. > > I don't think that's safe to rely on. The example named above didn't > pose any problems except on one platform. All the relevant functions > were static and easy to inline. GCC just followed the standard > literally and chose to break by reading one word at a time, just like > in this case it could break e.g. by copying one word at a time in > __builtin_memcpy (which seems perfectly reasonable to me --- > optimization involves a lot of constraint solving, and if you can't > trust your constraints then there's not much you can do). I wasn't disagreeing with you. I was guessing at why it did not fail out of the box when I tested it. What do you think of the alternative I posted? -Peff -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html