[Debian-science-sagemath] <DKIM> Re: Library transition freeze coming up on November 5

[Julien Puydt]

Hi,

On 10/10/2016 15:37, Tobias Hansen wrote:
> On 10/10/2016 02:19 PM, Julien Puydt wrote:
>> Hi,
>>
>> On 10/10/2016 15:12, Jerome BENOIT wrote:
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA512
>>>
>>> Ok,
>>>
>>> On 10/10/16 14:04, Julien Puydt wrote:
>>>> Hi,
>>>>
>>>> On 10/10/2016 13:55, Jerome BENOIT wrote:
>>>>> On 10/10/16 06:01, Julien Puydt wrote:
>>>>>
>>>>>> Jérôme, can you get in touch with upstream and get things
>>>>>> straight?
>>>>>
>>>>> I have contacted the upstream Sollya team about this very issue
>>>>> (and others): 0] they are very cooperative; 1] they are working on
>>>>> a Release Candidate that may support fplll version 5, this RC is
>>>>> not public (private ?) but they give me access to it and I am
>>>>> working on it with them.
>>>>>
>>>>> Is Sollya already part of Sage ?
>>>>
>>>> That isn't about sage: sollya is part of Debian unstable and depends
>>>> on fplll, so to push fplll 5.x.y in unstable, we need a library
>>>> transition, and that has to happen before the freeze.
>>>
>>> I understood this part of the story; I was just curious.
>>>
>>> Here Sollya may not be the issue, but gap-float should be as the
>>> upstream maintainer
>>> does not have the extra time necessary to migrate to fplll 5:
>>> is it possible to maintain a libflll4-dev for a while (read: beyond
>>> the freeze) ?
>>
>> Well, there was a soname bump going from fplll 4 to fplll 5, so there is
>> a libfplll0 and a libfplll2 ; that part is compatible.
>>
>> I'll need some mentoring : I have no clue how to make a new version
>> 4.0.4-3 with a new binary libfplll4-dev, while there is already a
>> 5.0.1-1 and a 5.0.2-1 :-/
>>
>
> If this is supposed to be temporary, one could upload a new version of
> fplll 4 with libfplll4-dev (conflicts/breaks libfplll-dev) and fplll4
> (conflicts/breaks fplll), upload gap-float using that, and then do the
> fplll 5 transition for sollya, introducing libfplll-dev (conflicts
> libfplll4-dev) and fplll (conflicts fplll4).
>
> In principle one could avoid the transition altogether this way. That's
> probably not what the release team wants to achieve with the transition
> freeze...

Uh... basically creating a new libfplll4 repository? That seems overkill...

Snark