[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Scientific software packaging

Mandag 23 mai 2011 18.08.32 skrev Steffen Möller :

> Hello Yngve,


> On 05/23/2011 04:42 PM, Yngve Inntjore Levinsen wrote:

> > On Monday 23 May 2011 14.19.44 Steffen Möller wrote:

> >> Hello,

> >>

> >> On 05/23/2011 01:30 PM, Yngve Inntjore Levinsen wrote:

> >>> On Monday 23 May 2011 12.29.15 Gürkan Sengün wrote:

> >>>> That sounds fantastic, do you already have ready packages that I could test?

> >>> You can test the ROOT packages which we distribute from CERN (unofficially) here:

> >>>

> >>> http://cern.ch/lcg-heppkg/debian/

> >>>

> >>> I assume Lifeng have used similar packaging scripts (and his efforts to get them into proper Debian channels are highly appreciated!).

> >> I just read through those pages. You mention source and binary packages,

> >> packaging for sid,

> >> explained backports, reprepro, reads all very nice.

> >>

> >> Have you considered sharing your 'debian' folders with the Debian science

> >> repository? There is no need to upload all the program code. And you do not

> >> even need to have anything ready to be uploaded. This way you would share

> >> you insights on how to compile on the various platforms through sharing the

> >> build instructions. This should also allow to work with Lifeng together

> >> on those

> >> packages and everyone would feel exceptionally well about those, even

> >> though your support would remain not to be official. You can e.g. use

> >> README.Debian for a disclaimer.

> >>

> >> Best regards,

> >>

> >> Steffen

> > Hi Steffen,

> >

> > I did not start this project, Axel Naumann and Kevin B. McCarty did. Since Kevin left the game I took over the Geant4 packaging (according to best efforts of course, I cannot give guarantees on the quality of my packaging from my level of expertise). People are more than welcome to download packaging scripts and suggest improvements/report bugs of course. The reason we keep it internally at CERN is, as explained, because of licensing problems and similar.

> >

> > I have not considered sharing the debian folders with Debian Science. I currently do not have much time on my hand to work on this, and I would expect that some quality control would have to be done etc (actually the latest Geant4 patch does not compile at the moment and I did not have time to fix). You are all free to download the sources using "apt-get source" of course, and I will be happy to try to explain what I have done. I am unsure if I understand your question though, you mean to give you the link of the folders from the server, or is there some "formal" way to publish the packaging scripts without the source code? I have no objection to distributing the packaging scripts (the respective authors would have to agree first), but the sources like CLHEP and Geant4 have difficult licensing so from what I understand we cannot distribute the sources outside CERN.

> >

> > Cheers,

> > Yngve


> my field is computational biology but once helped to get Christian's

> ROOT packages into the distribution when I was visiting Copenhagen more

> frequently. In the subversion repositories it is considered good style

> to only publish the debian folder, no source code. That should be

> retrieved via the instructions in the debian/watch file or the

> get-orig-source target in debian/rules or from the information in the

> debian/copyright file. That debian folder is commonly GPLed and easily

> comaintained, though this may differ if CERN has some policy I am not

> aware of when you started it. With git, there is no technical

> requirement but there it is common practice to indeed upload the source

> code in a separate branch. I dislike that immensely, but nobody seems to

> care about my aversion, so there are just a limited number of packages

> that I co-maintain with git.


> I would not mind apt-get sourcing and uploading the debian folders from

> there. But this would make sense only when you also use them. The

> quality does not matter for a start. Just say that it does not

> build/run. If the community cares then it will be fixed with or for you.

> If not, then not. I could talk you through the process of manually

> building with the debian folder in subversion and/or on how to use

> svn-buildpackage for some package that you feel more comfortable with.

> The redistribution of the binaries is a very different issue from the

> sharing of packaging/build instructions. And to have only the latter for

> some packages will still be helpful as an open invite to the community

> to contribute. I blogged about this at

> http://debianmed.blogspot.com/2011/04/debian-med-individuals-expertize-and.html


> Many greetings


> Steffen

Hi Steffen,

Thanks for your reply. Your suggestion sounds very sane, and I was not aware that this was a

possibility. I did already try to convert a less complex package to git-buildpackage system to see how

that worked (namely madx/madx-dev). If I understand there is a possibility to use either svn or git, and in case I do not want to

include the source files I should use svn?

If I can find some spare time I will have a look at this more in detail, I would VERY much appreciate to have

the build instructions stored in a more central place ( perhaps someone could even help me fix/maintain :) ). The only two that have worked on them are me and Kevin, and

he actually asked me to see if I could potentially manage to publish the built packages in Debian Science (ie. deal with all the politics that comes with non-standard licensing). Hence the build-instructions have no licensing issues along with it.



Reply to: