Bug#802222: transition: gdal

[Emilio Pozuelo Monfort <pochu@debian.org>]

Control: tags -1 confirmed

On 21/10/15 21:30, Sebastiaan Couwenberg wrote:
> On 21-10-15 21:19, Emilio Pozuelo Monfort wrote:
>> On 18/10/15 16:38, Bas Couwenberg wrote:
>>> For the Debian GIS team I'd like to transition to the recently released
>>> GDAL 1.11.3 as soon as possible.
>>>
>>> GDAL 2.0.1 was released along with 1.11.3 but several reverse dependencies
>>> still need patches to support GDAL 2.0, as recently discussed on the
>>> debian-gis list: https://lists.debian.org/debian-gis/2015/10/msg00022.html
>>>
>>> gdal (1.11.3+dfsg-1~exp1) is ready in experimental for about a month
>>> now.
>>>
>>> Because of the problematic mix of C & C++ symbols provided by libgdal,
>>> as discussed in the previous transition (#756867), the virtual ABI package
>>> provided by libgdal1i has changed to libgdal.so.1-1.11.3.
>>>
>>> The ben file used to prepare this transition is attached.
>>
>> I think you forgot to attach it? Anyway I've created one.
>> Please check if it looks correct.
> 
> I guess I did, the ben file used to prepare the transition is now
> attached, but the one you created will do too.

OK.

>>> Despite only marking the packages relying on C++ symbols as bad, I think
>>> all affected reverse dependencies should be binNMUed as part of this
>>> transition.
>>
>> Why is that?
> 
> Mostly to be better safe than sorry.
> 
>> If the C ABI is stable, then there's no need to binNMU the
>> rdeps. If it isn't, then you should change the package name.
>>
>> If we binNMU them now, the binNMUs will migrate to testing before the new
>> gdal, which wouldn't be good if there were ABI changes...
> 
> That's a good point.

Yeah, so let's not do that. Hopefully you've tested the new gdal with some C
apps that didn't break, that'd be good enough. If something turns out to break
in the end, then you'll have to rename the package... but let's hope that's not
necessary.

You can start this.

Cheers,
Emilio