This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH] ld: provide __ehdr_start magic symbol
On Thu, Jun 21, 2012 at 09:48:39AM -0700, Roland McGrath wrote:
> I think this notion is quite misguided, but that issue is entirely separate
> from what this change is about. So I'm happy to make my change comport
> with the new norm you've established, departure though it is from prior
> behavior and specifications. I'll discuss the section-vs-absolute issue in
> a separate thread.
Please do read some old threads first. You'll find me trying to justify
existing ld.so behaviour even though I knew it was wrong.
> Is this version OK to commit?
Yes, but without the following comment, which presumes we've had our
discussion and found absolute symbols quite appropriate for virtual
addresses.
> + /* The headers never sit inside any section, so this is really a
> + case for which an absolute symbol (SHN_ABS) is appropriate.
> + But the norm for linker-provided symbols now is that they are
> + associated with some section, even if the address is outside
> + the bounds of that section. So pick an arbitrary section to
> + associate this symbol with. */
--
Alan Modra
Australia Development Lab, IBM