> I was reworking some of the pkg.actions code as part of a pkgdepend > rework and thanks to Dan's help, discovered that validate_pkg in the > onbld package depended on a function that I had removed (get_variants). > > As it turns out, we can probably keep the function around for the moment > to keep compatibility. I'd like to file the bug to migrate validate_pkg > to the new function and I'd do that in bugster right?
Thank you. Yes, under solaris/consolidation/os-net-tools. > In general though, there's a larger concern that validate_pkg is using > interfaces that we haven't declared stable. Had Dan not happened to test > these bits with a full ON build, I would've happily putback and then ON > builds would've started breaking soon when build machines got upgraded. > > If validate_pkg is going to use internal apis of pkg, then maybe it > belongs in the IPS gate w/ tests so that we'd catch breakage before > putback. Dan had an idea where IPS provides an interface for doing > proto-comp like activities which the consumers providing a call back > function which is used to do the checking. Alternatively, we could go > the contract/documentation route. Whatever we decide, the current > situation seems likely to promote build breakage going forward. It seems like the IPS gate makes sense, as the functionality it provides is potentially useful to folks outside of ON. The feature set might or might not be rich enough for generic usage, as it was developed specifically as a protocmp replacement. I'm happy to fix it in the short term, and migrate it when that becomes appropriate. Do you think we've reached that point, and would the current state of the tool (modulo the current fix) be a reasonable first cut? --Mark _______________________________________________ on-ips-dev mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/on-ips-dev
