[Debian-science-sagemath] Sage 7.5 and Singular patches

[Ximin Luo]

Tobias Hansen:
> On 03/11/2017 03:22 PM, Ximin Luo wrote:
>> Tobias Hansen:
>>> On 03/10/2017 12:57 PM, Ximin Luo wrote:
>>>> Tobias Hansen:
>>>>> Thanks! All the singular related test failures I reported last time are
>>>>> gone except the one attached to this mail. That means we don't have any
>>>>> timeouts or segfaults anymore, just normal failures. Mostly ones that we
>>>>> already had for 7.4.
>>>>>
>>>>> [..]
>>>>
>>>> Hey, shall we release this soon? I can do a test build over the weekend as well.
>>>>
>>>> X
>>>>
>>>
>>> Yes, sure. I'm traveling until Monday and can't do much until then.
>>>
>>> The new 3D plotting with threejs works, but it loads threejs from the
>>> web. Can we leave it like that for experimental? Otherwise we need to
>>> either disable it or backport the patches from Sage 7.6 for using a
>>> local version of threejs. And update the threejs package.
>>>
>>
>> I think if upstream do that then we should stick with it, but I'll see if I can trivially patch in a "warning notice" somewhere in the UI.
>>
>> Afterwards, we can immediately begin packaging 7.6.beta directly after we release 7.5.1, and possibly have 7.5.1 in unstable with 7.6.beta in experimental.
>>
>> I think this is better, since it means we'd stay mostly-ahead rather than mostly-behind. (I'm also suggesting this approach for the Debian rust team.)
>>
>> X
>>
> 
> I think we should not upload new versions to unstable during the freeze,
> to keep the possibility to fix RC bugs in testing via unstable, no?
> Whether we want to package beta versions depends a lot on how much
> manpower and motivation we have, since that is even more work than just
> uploading every stable release. I'd be already happy if we upload every
> stable release (to unstable, during the freeze to experimental),
> possibly using beta or rc versions when working on a package for a new
> sage version.
> 

I had thought there was a "testing-proposed-updates" mechanism but I can't find documentation on it now. I think it does exist though. Since sagemath does not have any build-dependencies on it at the moment I think uploading to unstable should not interfere with the stretch release, even for other packages.

In terms of manpower and motivation, I think we can make it work with no extra overhead - just do exactly the same thing we have been doing, except target the current beta/rc version of Sage, and build+upload it to experimental. When that is released upstream, for Debian it should simply be a case of rebuild+upload to unstable, with minimal extra effort since we already did that for experimental.

Then we can release 7.7 in Debian (perhaps even 7.6) within a week or so of it being release upstream, which is much better for everyone.

X

-- 
GPG: ed25519/56034877E1F87C35
GPG: rsa4096/1318EFAC5FBBDBCE
https://github.com/infinity0/pubkeys.git