Re: [BUG] Regression in 2.14.90 (relative to 2.13.90)

[Daniel Jacobowitz <drow at mvista dot com>]

On Thu, Nov 13, 2003 at 11:12:40PM +0100, Carlo Wood wrote:
> On Wed, Nov 12, 2003 at 10:20:17AM -0800, H. J. Lu wrote:
> > > The change is intentional. The testcase looks normal to me since
> > > the discarded function is identical to the remained. Is there a
> > > problem with gdb? Please follow
> > > 
> > > http://sources.redhat.com/ml/binutils/2003-06/msg00471.html
> 
> I fail to understand how .eh_frame is related to this .debug_info
> problem.
> 
> > I took a look at the original bug report:
> > 
> > http://gcc.gnu.org/ml/gcc/2003-10/msg00657.html
> > 
> > The issue here is 2 functions have the same size, but slightly
> > different. I don't think we can solve all those link-once debug
> > problems without SHT_GROUP.
> 
> You cannot assume that functions with the same name, from
> different compilation units, are 100% equal; not even if
> the same is equal (that can be used as safety-gard, but it
> is a very poor one then).
> 
> Therefore, if the linker chooses a function from one compilation
> unit, it should at all times use the .debug_line info for that
> function from the same compilation unit.
> 
> Now, the way thinks work - that is not possible yet (without
> SHT_GROUP).
> 
> But Jason Merrill says
> http://gcc.gnu.org/ml/gcc/2003-10/msg00682.html
> 
> "I don't see why there would be a problem with the current
> situation; only the right .debug_line info should match the current PC."
> 
> and in http://gcc.gnu.org/ml/gcc/2003-10/msg00689.html
> 
> "... the relocs in the .debug_line info for the discarded copy should
> resolve to 0.  That's how we deal with duplicate .debug_info ..."
> 
> And then he came with the test case
> http://gcc.gnu.org/ml/gcc/2003-10/msg00723.html
> to show how that works - and it turned out not to work anymore
> with binutils-2.14.90.
> 
> Is there any reason why this can't be changed back?

Have you tried this using CVS binutils?

Nick could not reproduce your problem.  It could be a patch in Rawhide,
unlikely as that seems.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer