Re: Original sources, or not

[Bob Hilliard <hilliard@debian.org>]

Matthias Urlichs <smurf@debian.org> writes:

> Hi,
>
> many packages seem to contain .orig.tar.gz files which may or may not be 
> directly related to the files actually available from upstream. That is 
> unfortunate.
>
> I think that it would make sense to add a requirement to Policy that 
> the .orig.tar.gz file should be an unmodified copy from upstream.

     This would be virtually impossible for most of my dict-*
 packages.  There are several categories of problems with including
 true .orig.tar.gz files with these packages:

1- dict-wn.  The original package shipped as a tarball containing the
sources of the formatting software plus partial sources of the WordNet
database.  When new versions of WordNet have been released, I have
extracted the requisite files from the upstream tarball and inserted
them into the ../data directory, replacing the ones originally
distributed with the package.  I create a ".orig.tar.gz" from this
directory.

2- dict-gcide.  The original package was shipped as two tarballs - one
for the formatting software and one for the dictionary database.  The
data tarball had to be unpacked into a subdirectory of the directory
contained in the first tarball.  When new versions are released, I put
the new database in the directory and create a ".orig.tar.gz" from
this.

3- dict-misc (original).   The original package contained the sources
for an early version of dictfmt and three dictionary databases -
jargon, foldoc, easton, hitchcock, and elements.  The first two are
revised regularly, the last two seldom or never.  Retaining the
original structure would have required rebuilding all 5 dictionaries
whenever one was revised.  To deal with this, I created a separate
package for dictfmt, which has subsequently been included in the dict
package, and treated the individual dictionaries as described in the
next section.

4- My other dictionary packages are distributed as .tgz, .zip, or
plain text files.  Typically, the upstream editors/compilers of these
databases are interested in lexicography or in the specific subject
the database relates to, and not in programming.  Generally, these
unpack in ./, not in a directory named for the package.  For these, I
create a directory <package_name>-<version>, unpack the data sources
into it, and make a ".orig.tar.gz" from this.  This is the closest I
come to pristine sources in any of the dict-* packages.

     dictd, of course, is a conventional program, containing a source
tree, with auto-configuration files, makefiles, header files, etc. so
creating the .orig.tar.gz consists of renaming the upstream tarball.

     All of this is, of course, documented in my changelogs.  A strict
requirement that the .orig.tar.gz file should be an unmodified copy
from upstream would be impossible to satisfy for most of my packages.

Regards,

Bob
-- 
   _
  |_)  _  |_    Robert D. Hilliard        <hilliard@debian.org>
  |_) (_) |_)   1294 S.W. Seagull Way     <bob@bobhilliard.net>
                Palm City, FL 34990 USA   GPG Key ID: 390D6559