Edwin Beasant wrote: > Hi there: > > On 06/14/12 09:59 PM, david.co...@oracle.com wrote: > >>Thanks for the help - I've put all the suggestions together and respun > >>and tested the results. Danek has put his seal of approval on it too: > >> > >>http://jurassic.us.oracle.com/~ebeasant/webrevs/7168548/ > > > >components/ksh93/developer-astdev.p5m > > > > Why do you have this under components/ksh93 rather than > > components/meta-packages/history where all of the other > > obsoleted/renamed package manifests live? > > > > It would appear that moving them to there in this case results in dupaction > problems during pkglint of the ksh93 manifests: > I'm figuring that without the presence of the package rename manifests in > the component directory, the publish of the ksh93 component results in the > pkglint not being able to 'replace' the cannonical repo version with the > stub version during the lint run, resulting in the dupaction errors that > I've just seen. > > Do you know of any way of successfully forcing these manifests to be built > and considered at pkglint time? I did a full publish of the > meta-packages/history prior to the ksh93 publish. It would appear that the > pkglint is effectively performed against the (cached) cannonical repo data > for the packages.
Yeah, I think we need something better -- either from pkglint, or from the way we use it -- in order to get this right. I think it would be okay to keep all the ksh93 bits in the ksh directory for now, and in the next build move the renamed package manifests over to the history directory. The same should probably happen for the other renamed packages. Danek _______________________________________________ userland-discuss mailing list userland-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/userland-discuss
