On Mon, Mar 02, 2009 at 07:44:32PM -0800, Bart Smaalders wrote: > We have had this conversation before....
Yes... > 1) You cannot run any piece of software you are installing in a > postinstall script. > 2) You cannot count on the architecture of the machine being the > same. > 3) You cannot count on the filesystems being mounted in the usual > manner. > 4) You cannot count on the OS version being the same. > 5) You cannot count on any system services being available aside > from single user level basics. Yup, yup. > Scripting also raises the following problems: > > 1) Packages w/ scripting are unverifiable. I cannot determine > whether or not a package is correctly installed, because I > cannot determine the effects of the scripts being run. SVR4 packaging scripts are supposed to use installf(1M) and removef(1M) to tell the packaging system what they've done so that it can be verified. It may sound like I'm picking a nit, and to a point I am, but this feature may well be important to self-assembling packages. Of course, many an SVR4 pkg, no doubt, fails to use installf/removef properly. So your point is actually right for enough pkgs (i.e., more than zero) that it is right period. The system cannot possibly judge whether an install-time packaging script is coded to correctly use installf/removef. > We're _not_ supporting, allowing, permitting or condoning arbitrary > install time scripting in packages. Agreed. Nico -- _______________________________________________ pkg-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
