Le 31 mai 2013 14:39, "Salvatore Bonaccorso" <firstname.lastname@example.org> a écrit :
> Package: release-notes
> Tags: patch
> Hi Michael
> For reference, this is from http://bugs.debian.org/594273#54
> On Wed, May 29, 2013 at 03:11:06PM +0400, Michael Tokarev wrote:
> > I'm not sure it is the right place - the RM bugreport -
> > to add this information to.
> > The thing is: finally, wheezy is out without this module.
> > And it turns out that many 3rd-party code still uses the
> > old name, despite the new name being in core perl for
> > quite a while.
> > In particular, when upgrading postgresql.org to wheezy,
> > this change (removal of Digest::SHA1) was one of the main
> > issues, significantly bigger than all other together.
> > There were about 40 modules/programs which used the old
> > name, which all, at the same time, stopped working, and
> > it tool quite some efforts to figure it all out and to
> > make it all work again.
> > This pain could be reduced significantly by one of at least
> > two ways:
> > 1. By documenting the fact that this module is no longer
> > available, somewhere in release notes or something like
> > that. One of the issues faced was complete lack of any
> > information about this, and even searching does not reveal
> > anything useful easily (google finds oldstable package but
> > not this removal request).
> This sounds like a good plan as libdigest-sha1-perl was nevertheless
> still widely used. Attached is the patch proposed to apply to the
> > 2. By providing a compatibility layer, a tiny wrapper that
> > just maps one namespace into another. I don't remember how
> > this works in perl, but I guess it should be more or less
> > trivial to do - so that old Digest::SHA1 name still works
> > (maybe printing some deprecation warning or something).
> > Either as an additional package (maybe using the same old
> > name - libdigest-sha1-perl), or within core perl itself.
> IMO not the right approach as the aim was really to drop the use of
> the obsoleted Digest::SHA1. But adding a release-notes entry is surely
> more than worth.
Please add a Lintian test or ask for a test for checking obsolete module