Hi, I put this in bugzilla: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=123005 . I really do not appreciate Jeff pulling a wontfix on me on this one before having a proper discussion on the issue. The issue is that the two macros involved in doing a --root --aid install are interpreted in a different way. %_solve_dbpath is being interpreted as relative to the destination system, but %_solve_pkgsdir is being interpreted as relative to the source system. The argument for this is that you cannot assume anything of the chroot system. This argument doesn't hold for two reasons: 1) If this is a valid argument, then why is %_solve_pkgsdir not interpreted as relative to the chroot? (Think a minute about the consequences if it would be.) If you still stick to this argument then the macros should be taken from /etc/rpm/ in the chroot anyway and not from the source system. That would be a consequent implementation. 2) Not only do you know to what system/version you are installing, you will have to consider this before you install using --root and --aid, because if you are installing to a different version you will have to override both %_solve_pkgsdir and %_solve_dbpath (the latter only if the rpmdb is in a different location in the target system than on the source system). Further a different version installation is dangerous anyway. Consider installing RHL 9 from FC 1 this way. The version of rpm on the newly installed RHL 9 will not know how to handle the rpmdb FC 1 created. All in all this new behaviour is crippling a perfectly useful option. A --root --aid install to an empty chroot is now impossible without first installing the rpmdb. Nothing is gained as both macros have to be overridden when doing a different version installation anyway. If you argue that %_solve_dbpath usually doesn't need to be overridden in most cases I'd say this slight benefit does not compensate for the necessity to install the rpmdb to the chroot first. (Apart from the arguments against a different version install (incompatible rpmdb being created).) I'd rather override two macros than that I need to install the rpmdb beforehand. Finally I would like to know if this new behaviour was intended. If so, please explain why the macros are not taken from the chroot in the first place, or at least why the involved macros are not interpreted in the same way (both relative to the chroot). Leonard. -- mount -t life -o ro /dev/dna /genetic/research _______________________________________________ Rpm-list mailing list Rpm-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/rpm-list
