This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: Stripping empty sections
On Fri, Jul 08, 2005 at 05:32:07AM -0400, Hans-Peter Nilsson wrote:
> A few bugs.
>
> On Fri, 8 Jul 2005, Alan Modra wrote:
>
> > ld/testsuite/
> > * ld-cris/pv32.s: Give expobj a size.
>
> Why, here? Previously it was an undefined symbol defined
> elsewhere (see below). Why does it have to have a size *here*?
> Why doesn't that affect the test?
It needs a size *somewhere*. A zero size doesn't allocate space in
.dynbss. I'm minded to put an assert in adjust_dynamic_symbol,
because a zero size sym with a copy reloc just doesn't make sense.
Either that, or say we don't need a copy reloc.
> Mmph, why not in expdyn1.s, where it's defined?
Yeah, that would probably be better.
> > RCS file: /cvs/src/src/ld/testsuite/ld-mmix/sec-7m.d,v
> > retrieving revision 1.3
> > diff -u -p -r1.3 sec-7m.d
> > --- ld/testsuite/ld-mmix/sec-7m.d 12 May 2003 05:45:38 -0000 1.3
> > +++ ld/testsuite/ld-mmix/sec-7m.d 8 Jul 2005 03:26:07 -0000
> > @@ -20,19 +20,19 @@ Sections:
> > Idx Name[ ]+Size[ ]+VMA[ ]+LMA[ ]+File off Algn
> > 0 \.text[ ]+0+4 0+ 0+ 0+ 2\*\*2
> > [ ]+CONTENTS, ALLOC, LOAD, CODE
> > - 1 \.di 0+27ffb 2000000000000000 2000000000000000 0+ 2\*\*2
> > + 1 \.di +0+27ffb +0+4 +0+4 +0+ +2\*\*2
>
> This whole section moved? Is that right? Hm, I don't set it to
> writable anywhere in the input, so maybe moving it to the code
> area (0... not 2...) is something that can be expected. Still,
> it subverts the original test which expected to get a (separate)
> sequence of bytes in a new section, not .text.
Well, it is in a new section! The section moved because previously you
had an intervening zero size .data section that was located at 2...
--
Alan Modra
IBM OzLabs - Linux Technology Centre