perl package split - if you maintain a perl-* module, read this message.

Ralf Corsepius rc040203 at freenet.de
Wed Apr 18 16:28:03 UTC 2007


On Wed, 2007-04-18 at 10:27 -0500, Chris Adams wrote:
> Once upon a time, Robin Norwood <rnorwood at redhat.com> said:
> > Chris Adams <cmadams at hiwaay.net> writes:
> > > The best way to come up with a list probably is to see all the binary
> > > RPMs that depend on perl or a perl module.  Not all of them need
> > > perl-devel to build, but I'd bet most of them do.  A quick look at
> > > rawhide/i386 finds over 100 packages that don't have perl in their name
> > > that require perl.
> > 
> > Well, as Ralf points out elsewhere in the thread, this will break
> > packages upon rebuild, at which point they will need the appropriate
> > BuildRequires added...So I think this will be a problem only if the
> > package owner hasn't read this thread and isn't aware of the change.
> 
> There _could_ be some that don't break obviously.  If the package is
> autoconfed and perl is optional, it would just leave it out of the
> build.  Depending then on how the file list is specified, it could just
> end up being built without perl support.
Right, this case can't be excluded.

Nevertheless I am inclined to consider this to be a rarely met corner
case, because the perl-split primarily touches perl modules which are
typically used at built time of perl-modules and are rarely used at
run-time by other packages.

> That's probably not a big deal; at most there's probably only a few such
> packages, and for them at worst it would then result in a bugzilla
> if/when someone tries to use the missing perl support (and then it is a
> simple rebuild).
Exactly. ATM, I am not aware of any such package.

The only package I am aware about with weird module deps probably
needing deeper investigation is mod_perl (The FC6 version pulls in most
of the split-out modules - My gut feeling without having looked into
mod_perl's sources is "This probably isn't right".

Ralf





More information about the fedora-devel-list mailing list