This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: Powerpc disassembly and gdb breakpoints
- From: Alan Modra <amodra at bigpond dot net dot au>
- To: Andreas Schwab <schwab at suse dot de>
- Cc: binutils at sourceware dot org
- Date: Wed, 2 Jul 2008 09:43:14 +0930
- Subject: Re: Powerpc disassembly and gdb breakpoints
- References: <20080514022043.GA11328@bubble.grove.modra.org> <jek5g51u47.fsf@sykes.suse.de>
On Tue, Jul 01, 2008 at 05:50:32PM +0200, Andreas Schwab wrote:
> Alan Modra <amodra@bigpond.net.au> writes:
>
> > See the fixme I added to explain why putting the synthetic symbol on
> > the stub is difficult for ppc64.
>
> Apparently there is a similar problem on ppc. The number of PLT stubs
> does not necessarily correlate with the number of .rela.plt relocations,
> which may be related to the use of -Bsymbolic-functions. For example,
> in libQtNetwork.so.4.4.0 from libqt4-4.4.0-12.1.ppc in openSUSE 11.0
> there are 1617 PLT stubs, but only 457 .rela.plt entries. As a result
> all the foo@plt symbols are misnamed. When disassembling starting from
> offset 0x96620 there is a call to _Z7qgetenvPKc@plt, when in fact this
> function calls ioctl.
Sigh. Yes indeed. I forgot that -msecure-plt can require more than
one glink stub per plt entry when -fPIC and -shared/-pie.
--
Alan Modra
Australia Development Lab, IBM