Re: [davidm@napali.hpl.hp.com: Re: [PATCH] i386 vsyscall DSOimplementation]

["H. J. Lu" <hjl at lucon dot org>]

On Sat, Jun 14, 2003 at 03:50:47PM -0700, Jim Wilson wrote:
> The patch looks OK to me.  Normally this works because unwind info and 
> text end up in the same segment.  If they end up in different segments, 
> then it matters whether we use the segment containing the reloc, or the 
> segment containing the symbol.  This stuff was written by me, and I 
> didn't really know what I was doing at the time.
> 
> There does seem to be a technicality here though.  The IPF psabi says 
> that the reloc is valid only if the symbol and reloc end up in the same 
> segment, but they are in different segments here, so technically this is 
> invalid.  However, there is no problem making this work because the two 
> segments have fixed load addresses, so the offsets between the two 
> segments will never change.  If one of the segments could be loaded at a 
> different address, then the offset would change, and the unwind info 
> would be incorrect.

Under what condition will the offset between 2 segments change? Even
with shared library, I don't think the offset between 2 segments
will change.


H.J.