[Pkg-exppsy-maintainers] dh_make question

[Yaroslav Halchenko]

> I'm a bit confused about this whole dh_make business.
not a surprise since imho it is a bit messy ;)


> I then had to rename the directory to netcdf4-0.7.3.1 to run dh_make,
> which immediately complained that it couldn't find the source tarball.
>  I don't understand why it should need it because we'll be making it
> with the upstream branch.
I will leave this question unanswered but you can find answer for the
logic of dh_make in its manpage... ok -- here is the quote:

       Unless --native was given, dh_make makes sure a original source archive
       (<packagename>-<version>.orig.tar.gz) exists.  If no such file  exists,
       the  file  specified  with -f is copied in place.  If no -f is supplied
       either but --createorig is, the current directory it copied into <pack‐
       agename>-<version>.orig   which   will  be  made  to  this  archive  by
       dpkg-source later automatically.  The original archive  is  needed  for
       other  Debian  tools  to  generate  the  diffs  to the original sources
       required by the Debian packaging  format.   Unless  there  are  reasons
       against it this file should be the pristine upstream archive.

so pretty much your complain is "why --createorig is not enabled
automagically"... right? ;-) imho too much of automagic is a bad thing -- many
things could go wrong without mentioning them.

> So I supplied the tarball file with the
> command line options.  Do we have to keep this orig tarball around
> even if we start pulling in new versions of the source into the
> upstream branch?

that is more of "philosophy" question. It is in good manners (I believe
or it might be even "should" in debian policy), that package provides
.orig.tar.gz which is pristine tarball. Although content of upstream might be
identical, sha or md5sum of tarball would be different, thus a user will not
have an easy way to verify that debian shipped orig is pristine (without
untarring and diffing)

Thus, even so I have upstream branch, I prefer to make use of orig.tar.gz
tarballs even with git-buildpackage. If you mind doing it that way -- you are
welcome to don't keep tarballs but you might run into problems or confusions
later on -- for instance prepping security update - you would need to upload
orig.tar.gz as well -- but if you regenerate it from upstream sha/md5 sums
would be different from what is already in the pool of main Debian. Thus Debian
will get two files with the same names and different checksums... imho it is
somewhat bad...

I hope that clears the question ;-) if not - lets debate more

> The other annoying part is that I'm going to be making python-netcdf3,
> but the source is going to be something named netcdf4-xxx, but I guess
> that's a by-product of the fact that the author put it all in one
> tarball.

not a problem imho. source package doesn't have to be 100% identical to
produced binary packages... besides that sooner or later you will also provide
us with python-netcdf4 package right?

also the thing to keep in mind that according to python policy
http://www.debian.org/doc/packaging-manuals/python-policy/ch-module_packages.html
Public modules should be packaged with a name of python-foo, where foo is the
name of the module. 
It seems that you are ok but I just wanted to mention to keep that in mind...

> above.  At this point I need to actually start editing the files in
> debian and making necessary source changes to setup.py and the like to
> make it only compile the netcdf3 wrappers.  Before I start that, do
> you think I made any mistakes with what I've done so far?
well... seems to be fine so far... ;-)
I suggest you to use 
git-dch --snapshot

to create 'pre-release debian packages' so we have proper versions
attached without collissions when you progress with the packaging.

> I just created a new git repository on alioth as netcdf4.git.  At your
> convenience, could you take a quick glance to make sure I can move
> forward from the foundation I have build?
sure -- go ahead l) why to stop?

-- 
Yaroslav Halchenko
Research Assistant, Psychology Department, Rutgers-Newark
Student  Ph.D. @ CS Dept. NJIT
Office: (973) 353-5440x263 | FWD: 82823 | Fax: (973) 353-1171
        101 Warren Str, Smith Hall, Rm 4-105, Newark NJ 07102
WWW:     http://www.linkedin.com/in/yarik