[PATCH] Segfault in AArch64 backend with -g and -mbig-endian

[Renato Golin]

So, in this case, the revert need only to happen if the section is .eh_frame. If the section is empty, it's not .eh_frame and the local logic is not wrong.

That's not saying there isn't a problem and that it doesn't need fixing, it just says that this is not the point of *this* patch. What'd be the problem of fixing this and creating a new bug for the empty section? What guarantees that this is the only case where the section is empty? I believe that the right thing to do is to fill the section code with asserts, so we can pinpoint the origin of the problem. But none of that would make this fix wrong in the future.

Furthermore, the function does return null when it doesn't find the section, so if that was really wrong it should assert in the first place. I don't understand what the problem is in this case, can you elaborate?

http://reviews.llvm.org/D4870