Re: casacore-data uploaded to mentors.debian.net

[heroxbd@gentoo.org]

Dear Ole,

Ole Streicher <olebole@debian.org> writes:

> heroxbd@gentoo.org writes:
>> This package is needed by casacore at build (for test) and run time.
>> The main concern is the license.  Gijs has contacted upstream and was
>> told the data were released as LGPL.  It would be nice if the email
>> could be referred to.
>
> The major problem, is that (according to debian/README) there is no
> reproducible way to really build the package from an (upstream) source:
> it is unclear (to me) what measures_data.csh downloads at the end, since
> there are no URLs in it.

I have seen four data sources to get runtime data in casa table format:

  1. ftp://ftp.atnf.csiro.au/pub/software/asap/data/asap_data.tar.bz2
  2. ftp://ftp.atnf.csiro.au/pub/software/measures_data/measures_data.tar.bz2

     These two are identical now, by md5sum.

  3. rsync://rsync.aoc.nrao.edu::casadata

     This is outdated.

  4. Following the outline by Gijs in debian/README, I have reproduced
     the procedure as:

     + measuresdata
     
       produce measuresdata.rc
       
     + mkdir geodetic
     + measuresdata.csh | tee measuresdata.log
     
       measuresdata.csh is from casacore tarball. measuresdata.log is
       uploaded to http://neutrino.d.airelinux.org/measuresdata.log.xz

       As seen in the log, it downloads ascii data from several ftp's
       and aggregate them into casa table format. 

       The generated tables are
       geodetic/{IERSeop{2000,97},IERSpredict{,2000},TAI_UTC}. They are
       only part of needed casacore-data.
       
     + cd geodetic; showtable -in TAI_UTC gives
     
> showtable: Version 2013Oct16GvD

> Structure of table /home/heroxbd/dev/tmp/casacore-data/geodetic/TAI_UTC
> ------------------ IERS (leapSecond)
> 39 rows, 4 columns (using 1 data managers)

>  IncrementalStMan file=table.f0  name=IncrementalStMan  bucketsize=32768
>     ActualCacheSize=1 PERSCACHESIZE=1
>   MJD        double   scalar unit=d
>   Multiplier double   scalar unit=s
>   Offset     double   scalar unit=d
>   dUTC       double   scalar unit=s

      which is the same as 1. and 2. above.  I don't know how to make a
      dump of the casa table to compare them exactly. Chances are they
      are identical, or the 1. and 2. source might be a little less
      cutting edge than 4.

> I think, a clean solution would be:
>
> 1. Extract all files that are downloaded by measures_data.csh, and put
>    them as sources for the casacore-data package (there may be more than
>    one source file per Debian package).
>
> 2. In debian/rules, create the files that go into the package by using
>    casacore's measuresdata program, and build-depend on casacore for
>    this.
>
> 3. Don't build-depend casacore from casacore-data for the first upload
>    and disable the tests that depend on it
>
> 4. After casacore-data is in Debian, we can re-enable the last build
>    dependency and the tests. Circular dependencies seem not to be a
>    problem in our case according to [1].


So I think the 0-th solution is to use

  ftp://ftp.atnf.csiro.au/pub/software/asap/data/asap_data.tar.bz2

as upstream and ship the data as-is, as recommended by the casacore
team.

  http://code.google.com/p/casacore/wiki/FrequentlyAskedQuestions

Then there is no build dependence of casacore-data on casacore.  

What's your opinion?

Cheers,
Benda

> [1] https://lists.debian.org/debian-astro/2014/09/msg00016.html