Tim Foster wrote: > On Sun, 2011-03-06 at 21:05 -0800, Danek Duvall wrote: > > I'm doing some work in userland, and have found that some unnecessary > > attributes are making it through the publication process. In particular, > > I've seen pkg.linted* and pkg.depend.bypass-generate. > > I could certainly see a case for removing pkg.depend.bypass-generate: > you're unlikely to ever want to re-run pkgdepend on a package. > > Removing linted would be more contentious, for example in the case where > a user uploads a new package, then tries to lint the entire repository: > the repo would report errors on unrelated packages that were previously > intentionally bypassed, for whatever reason.
Well, I think of a repository as a place for packages where clients can access them. I suppose there's a case to be made that they're also an intermediate holding area for packages while they're still under development. Perhaps we stress that stripping these attributes may not be desirable when publishing to the latter, but strongly encourage doing so when publishing to the former (possibly by using pkgrecv with a mogrify filter to do that publication). > > Another option might be to give pkgsend the ability to read in transforms > > from one or more files. > > Yes! I'd wanted that in the svr4 convert work to allow transforms to > automatically tag SMF services and the like -- can I resurrect that bit > of code too? I thought (not having looked at your code) that you were doing that, but hardcoding the ones you knew about, rather than providing a hook into pkgsend for people to provide arbitrary transforms. Danek _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
