On 02.02.2012 20:55, Mathieu Malaterre wrote:
On Thu, Feb 2, 2012 at 9:13 PM, Adam D. Barratt <adam@adam-barratt.org.uk> wrote:On Wed, 2012-01-25 at 11:13 +0100, Mathieu Malaterre wrote:GDCM 2.2.0 introduces a new ABI, as seen on #655783 and al.Since API (whatever that means for C++) is preserved, would it be a good time to- move gdcm 2.2.0 from experimental to unstableApparently the lack of an explicit "no" - having waited less than a week- was taken as an implicit "yes". That's unfortunate, given that it means that the gdcm transition is now tied together with the mono transition which we were very close to finishing.
[...]
I completely understand your point and I will not upload any new gdcm.In any case if you decide to revert to gdcm 2.0 watch very carefully for #657288 since it introduce a change in the API without any SONAME bump.
Thanks. I suspect that's not the bug number you intended to reference though.
I initially made the very first upload of gdcm 2.2 because of #657779,which I thought would help in the mono transition.
Fixing the bug helped, yes. The SONAME bump not so much. :-)
I choose to upload directly 2.2.0 (vs a gdcm 2.0.19) since it clearly state the SONAME bump and I assume this would make the life of everybody else much easier. In particular I assumed having gdcm 2.2 would help the ITK4 transition, also debated on debian-release [1].
It may make ITK4 easier; we'll see.What it looks like we'll end up doing is pushing the new gdcm in to testing much earlier than we normally would, to get the mono transition finished; britney allows us to keep the shared libraries for both 2.0 and 2.2 in testing while we sort out the reverse-dependencies. We'll then look at finishing off the gdcm transition.
Regards, Adam