[lld] r254003 - Revert "ELF: Make .note.GNU-stack more compatible with traditional linkers."

[Joerg Sonnenberger via llvm-commits]

On Tue, Nov 24, 2015 at 02:42:48PM -0800, Rui Ueyama wrote:
> On Tue, Nov 24, 2015 at 2:37 PM, Joerg Sonnenberger via llvm-commits <
> llvm-commits at lists.llvm.org> wrote:
> 
> > On Tue, Nov 24, 2015 at 02:23:34PM -0800, Rui Ueyama wrote:
> > > Omitting PT_GNU_STACK when -z execstack is provided is easy, but it leads
> > > to a counter-intuitive behavior on platforms that do not recognize
> > > PT_GNU_STACK and stack is not executable by default. (You need to pass -z
> > > execstack to make it run on such platforms, but it wouldn't make the
> > stack
> > > executable.)
> >
> > That argument makes no sense to me and I am not even sure I understand
> > it correctly. A platform (like NetBSD) that doesn't recognize
> > PT_GNU_STACK and provides a non-executable stack by default will not be
> > affected at all. Yes, it won't do what is intended, but it is no worse
> > than the current situation of creating a bogus program header either.
> >
> 
> Let me rephrase. That you needs to pass "-z execstack" (although you don't
> intend to make the stack executable) to make the output compatible with a
> platform that rejects executables with PT_GNU_STACK is I think
> counter-intuitive.

I agree that it is counter-intuive from the option name, but not if you
consider the header entry as a way to only flag binaries with
non-default non-execstack. A better name would be -z noexecstackheader,
but the question really just is that *some* form should be supported.
Not emiting the header if it doesn't do anything sounds perfectly
reasonable to me, it wastes less space and allows producing binaries
fully matching the original ELF specification.

Joerg