Re: [Pkg-octave-devel] Sundials is way outdated

Sorry about the late reply, was flying to the US and didn't really have internet.

To answer a specific question first:
> Do they generate a CMake package configuration at least?

No, they don't, and they dropped sundials-config (http://sundials.2283335.n4.nabble.com/sundials-config-removed-td4653478.html#a4653481), which was the main way to get the correct linking args.

About my packaging:
I started back at the end of 2015 (see http://sundials.2283335.n4.nabble.com/Packaging-sundials-for-debian-td4653658.html for my email to upstream, I've sent them the associated patches, but heard no reply. They did however make it easier to download sundials in 2.7, and the watch file in that version works), where I used gbp-import-dscs to build the history. I originally tried just packaging 2.6.2 (when this was the latest release), but to try to understand the ABI breaks I ended up packaging 2.6-2.6.2, and 2.7 later when it came out. I vaguely remember trying to create an alioth account, but there was some problem, which is why I never ended up uploading it to Debian Science git (I uploaded it to github as the developer of scikits.odes asked me to, which is why it's there). I also couldn't get the svn-to-gbp converter working which is mentioned on the wiki. I was planning on replying ITA when I got the reply from upstream with the patches, but they didn't and it slipped of my radar, so the mixup with different version is my fault. I haven't done anything with pthreads or openmp, as I was going to query how we wanted to restructure the package with the new changes/additions.

Future plans:

I'm in the US for a conference until next week, and don't have ready access to a system setup for packaging, so I wouldn't be able to do much before the end of next week. I don't mind if you choose to reimport everything, or use Dima or my repositories. I would point out though that most of the debian/ folder is not that useful (ABI breaks, build system change which invalidates existing patches, changes to watch file, etc.), but you should be able to see from what I did what's worth keeping and what's not. When I get back home, I'll create packages for each of the different run methods (openmp, pthreads, plus any other new additions which upstream has added in 2.6/2.7), and properly drop octave-sundials, which I'll do off my work unless someone will have already cleaned up the debian science git repo by then. I'll also try again to recreate an alioth account, and report here if I encounter any problems. I'm happy to maintain this in the long term, I was (and still am partly) unsure how drastic the changes I should make are (e.g. in relation to octave-sundials which we've decided on, and what actual numbers I should use for the ABIs).

James

On 2 February 2017 at 19:00, Dima Kogan <dima@secretsauce.net> wrote:

Andreas Tille <andreas@fam-tille.de> writes:

> On Wed, Feb 01, 2017 at 11:48:08PM +0000, Ghislain Vaillant wrote:
>> > I admit I would have prefered if you would have left you as owner and
>> > would take over the lead here.
>>
>> I thought I was stepping on someone else's work (Dima or James) and
>
> If this would have been the case Dima or James should have responded in
> the last three days, right?

Right. If you want to do the required work in the near term, then by all
means, please do it.

>> that an official update to a more recent release was already underway.
>> Hence myself leaving the RFA up for whoever was in charge.
>>
>> I should state upfront that I am not comfortable with the way this
>> packaging effort has been performed so far. Looking at the history,
>> Dima started the work on 2.7 and imported the Debianization from svn,
>> without any consideration for past work nor policy. The very minimum
>> would have been to start with a `gbp import-dscs` if a proper import
>> from svn was too hard to achieve, despite the process being documented
>> in different places on the Debian wiki.

I don't remember exactly what I was thinking when I did this. Normally,
I'd import the old repo, but clearly I did not do that here. Yes, that
is bad. Sorry. The existing new git repo isn't really used by anybody,
and rebasing on top of a new svn import would be fine. I don't think
you'll be confusing any existing users.

>> Finally, considering the complexity of this software, it would be nice
>> to clarify upfront who is intending to maintain sundials long-term.
>
> +1
>
>> Again, looking at the history, Dima kickstarted the update effort for a
>> week in October last year, and apparently gave up afterwards. That does
>> not sound like a confident statement for long-term maintenance to me.

I needed packages for a new version of this library, so I did some of
that work in the git repo yall have seen. At the same time I wrote that
email to the debian-science mailing list. Talking to James on that
thread, we decided to leave this alone until after stretch was out the
door, and to pick it up again then. Plan for the future: if someone here
wants to do the packaging and maintenance, please do that. If nobody
wants to, I'll do it.

>> I hope you guys understand my concerns here.
>
> I fully share your concerns. Dima and James claimed that they are happy
> with their work for local usage but this does not help from a Debian
> point of view at all. @Dima & @James: Could you please explicitly lay
> out your plans here until weekend. In case we do not hear from you I
> think the best plan is to redo the SVN-Git conversion.

Agree completely. Sorry for the mess.

Don't send me files in proprietary formats (.doc(x), .xls, .ppt etc.). It isn't good enough for Tim Berners-Lee, and it isn't good enough for me either. For more information visit http://www.gnu.org/philosophy/no-word-attachments.html.

Truly great madness cannot be achieved without significant intelligence.
- Henrik Tikkanen

If you're not messing with your sanity, you're not having fun.
- James Tocknell

In theory, there is no difference between theory and practice; In practice, there is.