Re: Cmake help needed (Was: Unresolved ${java:Depends} (Was: IGV))

[Andreas Tille <andreas@an3as.eu>]

Hi Olivier,

On Wed, Oct 24, 2018 at 01:45:25PM +0200, Olivier Sallou wrote:
> 
> On 10/24/18 1:12 PM, Andreas Tille wrote:
> > Hmmmm, anyway - I'd love if we could find a (hopefully clean) way to
> > update discosnp and minia to the latest upstream versions.  Its probably
> > not good to lag heavily behind upstream.
> 
> minia 3.2 needs a gatb-core more recent than code in 1.4.1
> 
> 
> the minia code do some calls to gatbcore functions with different number
> of parameters.
> 
> Means gatbcore ABI has been modified since, though no new release was done.

Thanks a lot for your investigation.
 
> example:
> 
> 
> in core v1.4.1, src/gatb/bcalm2/bglue_algo.hpp,  line ~137 , we have
> this definition
> 
> 
> void bglue(gatb::core::tools::storage::impl::Storage* storage,
>         std::string prefix,
>         int kmerSize,
>         int nb_threads,
>         bool verbose
>         );
> 
> }}}}
> 
> with 5 arguments, ut minia calls it with 6 argiments:
> 
> 
> 
> so minia depends on an unreleased gatb core ......
> 
> methods switched to 6 arguments ~9 month ago, but gatb has not been
> released since Dec 2017.....
> 
> 
> 
> Remaining option is to include the gatb-core source code from minia
> (etc...) and compile code with this one.

I see the following additional options:

  1. Open an issue in gatb-core bug tracker to tag a new release
     (you said you lost personal contact, but may be you past the
      result of your investigation in the bug tracker?)
  2. We are packaging latest Git HEAD using git mode in watch file.
     I see no real risk here since the motivation to package gatb-core
     was originally to package minia (no idea what effect this might
     have on discosnp)
 
> In minia git repo, gatbcore is in thridparty dir as a sub module.

I realised that more and more upstream entertain this habit. :-( 

Kind regards

         Andreas.

-- 
http://fam-tille.de