Need some help with a patch for new command --warn-execstack for ld

[Mike Frysinger]

On Wednesday 26 May 2010 10:00:40 Nick Clifton wrote:
> Getting hold of X however is not really possible.  There is no
> information in the object file so specify which function(s) needed an
> executable stack.  That information is only known at compile time and it
> is lost by the time that the object file is created.  My recommendation
> would be to just drop references to explicit function names from the
> warning messages.

the real power of the original patch to gcc was that it gave function-level 
warnings.  when working with a huge file and only one of those funcs was 
causing a problem, it was hard to narrow down which one by hand.  yes, there 
were some spurious warnings from time to time ;), but we ignored those if the 
intermediate object and/or link didnt cause problems (require exec stack).

i'm not complaining about the linker or some bfd limitation here, just 
commenting on the original purpose.  i know that by the time the exec stack 
marking has made it to the linker, any such binding to the triggering function 
is lost.  this is true before any code is even delivered to binutils since gcc 
only outputs one .note.GNU-stack per unit.

so if we want function-level warning (and we def do), i'm pretty sure it has 
to be in gcc.  not to say we dont also want to extend the linker as there are 
cases where the exec stack is coming from asm files that gcc has nothing to do 
with, or precompiled objects that the end user is given directly.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://sourceware.org/pipermail/binutils/attachments/20100526/99a35c0c/attachment.sig>