[Debian-science-sagemath] Bug#835027: pari: please provide PARI 2.8.0[.alpha] in experimental

[Ximin Luo]

Jerome BENOIT:
> On 06/09/16 15:52, Bill Allombert wrote:
>> PARI 2.8.1 will be released in about one month and will be API/ABI
>> incompatible with 2.8.0. Then PARI 2.9.0 will be released.
>> We should focus on PARI 2.9.0 and not be distracted with 2.8.0.
> 

It's not usual for FOSS projects to make API/ABI changes across minor versions. And *especially* not to keep installation paths the same whilst doing so. What's the rationale for upstream to do this?

> Can you consider PARI 2.8.1 for experimental ?
> Do you plan to bring PARI 2.9.0 to Debian before the big freeze ?
> 
> 
>> Making valid Debian packages for PARI 2.8.0 is difficult, even in
>> experimental.
>> First the PARI library is not versionned the same way in stable and
>> unstable release, so the packaging need to be adjusted to take that into
>> account (something pari-sage does not do).

Could you explain what is meant by "not versioned the same way"? It seems that these two lists are identical:

https://packages.debian.org/jessie/amd64/libpari-dev/filelist (2.7.2)
https://packages.debian.org/sid/amd64/libpari-dev/filelist (2.7.6)

and the versions (2.7.2, 2.7.6) seem very similar to me...

>> Secondly, such experimental packages must not interfer with the packages
>> for PARI 2.7.6, so different package names and filepath needs to be
>> used. pari-sage does not do that either.
> 
>> And then you have to change Sage to use the new names and new paths.
> 
>> You should just bundle pari 2.8.0 in the Sage tarball in the meantime. This
>> would be far less disruptive. Then when PARI 2.9.0 is released, you can
>> move to that.
> 
> It is not tenable because then a full bunch a library must be packaged against pari-sage.
> Please note that pari-sage is rather an ad hoc package to step forward.
> We strongly wish to work with the stable PARI material.
> 

I think what Jerome means here is that we have other packages like cysignals that we need to build against pari-2.8.0. This makes it hard to "just bundle pari 2.8 with sage". Or if this isn't what you meant, could you clarify?

I think, the best situation for now is just for us (debian sagemath) to package pari at version 2.8 and put it in our sid-sage APT repo. This would unblock our progress for the other items. We are nowhere near making a "proper Debian package" for sage yet, so we don't need to worry about co-installation and paths at this time. The important thing is to *unblock* other items so we can work on those.

By the time that we are ready to upload Sage 7.3 (or 7.4), hopefully they will have upgraded to pari 2.9. Or if not, *then* we can think about how to solve this issue with 2.8 being an old unsupported version.

X

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