On Mon, Oct 03, 2005 at 10:46:00AM +0200, Josselin Mouette wrote:
> Le dimanche 02 octobre 2005 à 23:55 -0700, Steve Langasek a écrit :
> > The latest version of libpng12-0 includes a shlibs bump with only this
> > changelog entry:

> >   * libpng12-0.shlibs: bump the shlibs version.

> > I have compared the symbol tables from libpng12-0_1.2.8rel-4 and
> > libpng12-0_1.2.8rel-1.  There are *no* symbols exported by -4 that were not
> > already in -1.  Indeed, as we both know there are a number of symbols that
> > were *removed*, but that should have no impact on the shlibs: any binary
> > built against -4 *will* work with -1, so the shlibs should not have been
> > changed from their previous value.

> Due to the build of the MMX optimised routines, there is a new symbol,
> named png_pass_width. (No, don't ask me why.)

Gar, I overlooked this because it's a .rodata symbol, not a .text symbol.
Of course, a sane library doesn't *export* data structures as part of its
ABI, but I guess I should know better...

> As I don't think it is used by any applications, maybe the shlibs change
> can indeed be removed. Another option is to remove the optimisations
> until the package is in sarge, or to simply remove this package. What do
> you think?

I think removing the symbol from the export table and dropping the shlibs
change would be best here.

Ideally, PNG_USE_GLOBAL_ARRAYS should go away, but that would itself be an
ABI change.

Thanks,
-- 
Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
[EMAIL PROTECTED]                                   http://www.debian.org/

Attachment: signature.asc
Description: Digital signature

Reply via email to