Dan Smith wrote:
DV> # HG changeset patch DV> # User Daniel Veillard <veillard@redhat.com> DV> # Date 1193656844 -3600 DV> # Node ID 5b601a0035127544c54feaf086fb7065f873e9f9 DV> # Parent aaad430ea766d9a3bc0be035e595d8da17ee2122 DV> RPM Packaging changes and associated cleanups: DV> - Augment some Makefiles variables with missing files DV> - add a spec file template DV> - extend the configure script to generate the spec file DV> - add a 'make rpm' target DV> - fix src/Makefile.am to use late binding of providerdir variable
Works for me. +1
DV> There is still one thing missing I guess in the spec file: the DV> registration step, which I think should be run as a post-install DV> script. The spec file saves the provider-register.sh and DV> register_base.sh scripts in /usr/share/libvirt-cim/ so they are DV> available at that point, but I don't understand how to best call DV> them, so it will need some attention by someone who knows how that DV> process should work.
I *think* the rpm needs to do that in the postinstall step, although I'm not sure if/how the determination of which CIMOM is to be used should be made. If the rpm assumes pegasus is installed, then perhaps it should be hard-coded and the user can re-run the provider registration if they are using something non-standard?
Heidi would probably be the one to comment on the convention here.
Anyone have a problem with putting this in the tree now and fixing up the registration step in a subsequent patch?
I'm all for "mostly works" over "not in the tree at all," especially given how relatively immature the code is. It's not like we've got a finished and polished product yet; we've got a little time to straighten out the postinstall, and having the rpm stuff in there would be nice. :) -- -Jay