Bug#650601: transition: libpng 1.5
On Mon, 2011-12-05 at 16:34 +0900, Nobuhiro Iwamatsu wrote:
> 2011/12/1 Adam D. Barratt <adam@adam-barratt.org.uk>:
> > On Thu, 1 Dec 2011 09:16:41 +0900, Nobuhiro Iwamatsu wrote:
> >>
> >> Libpng maintainers want to update libpng from 1.2 to 1.5.
> >> libpng of ABI and API has been changed by change of 1.2 to 1.5, so it
> >> needs a transition from libopng12 to libpng15.
> >> We tested building of the package depending on libpng12.
> >> FTBFS by this change is reported and is summarized below.
> >>
> >> http://bugs.debian.org/cgi-bin/pkgreport.cgi?tag=libpng15-transition;users=libpng@packages.debian.org
> >> Almost all packages have not been corrected yet.
That's quite a blocker; see below.
> >> And it is necessary to change Build-depends of almost all packages
> >> into libpng-dev from libpng12-dev.
> >
> >
> > Is there a good reason why the new libpng-dev couldn't at least Provide
> > libpng12-dev in the short term?
>
> libpng12-dev provides libpng-dev.
> This is provided from before.
Was this communicated to affected maintainers? Looking at unstable's
Sources list, there appear to be around 100 packages already
build-depending on libpng-dev, but there's no easy way of telling how
long they've been doing so.
> > Would this allow some (most?) packages to
> > be binNMUed?
>
> No, almost all packages have described libpng12-dev to Build-Depends.
I'm not sure if something's getting lost in translation here, or if I
wasn't clear enough in my question. If libpng12-dev was still Provided,
is there any reason we couldn't then binNMU the 100-or-so packages
marked as "ok" in your list?
> First, we had better upload libpng15, after changing libpng12-dev into
> libpng-dev.
> Surely, I think that this method is easy for shift.
We appear to have different definitions of "easy". Anything that
involves changes to and uploads of over 300 packages is not what the
release team classifies as easy.
Furthermore, your list indicates that you're aware of nearly 130 build
failures with the new library, and that less than a quarter of those
have patches in the BTS; that's really too large a number to be starting
a transition with. How many of the failures which don't have patches
filed are directly attributable to the libpng changes?
Regards,
Adam
Reply to: