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

Re: Bug#902797: lintian: check latest changelog entry for duplicate contributor information



Hi,

On Sun, Jul 01, 2018 at 11:37:21AM +0200, Mattia Rizzolo wrote:
> On Sun, Jul 01, 2018 at 12:39:47PM +0800, Paul Wise wrote:
> > I recently saw a changelog entry (quoted below) for a Perl team package
> > where several contributors to that version had their names mentioned
> > multiple times with one or more changes below each instance of their
> > name. This made the changelog harder to read. I think it would be
> > useful for lintian to emit a pedantic or info-level warning for
> > duplicate contributor information in the latest changelog entry.
> 
> I personally agree with you here, but apparently not everybody does.
> ISTR to remember in the past suggesting to change the default of dch
> --multimaint-merge but somebody complained that that way it would lose
> the chronological order of the changes or something like that (I don't
> really buy it).

I'm probably guilty of uploading a fair number of packages with such
changelogs over the last few days, as I'm working through a list of perl
team packages that haven't had an upload in over six years and thus still
point to SVN with their version in the archive.

Most of those changelog entries stem from (semi-)automatic mass commits
made over the years to our several thousand packages; and some of them
change the very same thing a previous entry modified. In the example
that Paul cited, Ansgar first changed the Vcs-* lines from svn.d.o to
git.d.o in 2011, then Salvatore changed git.d.o to anonscm.d.o in 2013,
git:// to https:// in 2016, and anonscm.d.o to salsa.d.o in 2018, each
time adding a commit with the change and one that modified d/changelog
using dch. If that last commit was made by Ansgar again, wouldn't
--multimaint-merge result in a confusing changelog that lists final
modifications before intermediate ones?

Having said that, I don't have a strong feeling about how changelogs
should look like, other than that the default behaviour of our tools
should agree with what lintian wants to see. In my use of d/changelog,
both as a packager and occasionally as a user looking at an installed
package, it's a mirror of the packaging repository's git history, very
likely auto-generated, and a bit redundant as a source of information...

Florian


Reply to: