On Fri, Oct 4, 2013 at 3:58 PM, Andrew Gregory <andrew.gregory.8@gmail.com> wrote:
pmdb is responsible for populating /var/lib/pacman/local/. In fact, you'll need to update pmdb so it will create the mtree file, which I seem to have neglected to mention before.
D'oh! Yes, indeed, it does! How embarrassing that I somehow missed that. I assumed incorrectly based on its name and by analogy to what pacman does that pmpkg.install_package would be responsible for populating /var/lib/pacman/local. Oops. Obviously this patch set is D.O.A.
The snapshot is used to determine if the file was modified after being installed. None of your tests needed a post-snapshot hook, and I can't think of any that would. Of course, if such a need does arise, we can revisit the idea of post-snapshot hooks at that time.
Correct, none of the -Qk(k) tests need that hook. But I mentioned in the cover-letter a scenario that I thought would: reinstalling to repair damaged files. Perhaps it's just that I'm still not completely understanding how the snapshot gets used, but that scenario still seems to me to need something more than what the local pmdb does. I am quite content, however, to put that aside until the day a test is written that actually requires it. So, I'll get to work on an improved version of this patch set. My hooks code is out. The mtree generation gets moved to where it belongs. And I need to figure out how to call the install script - which (without looking at the code at all, so I could be wrong) involves forking sh as a final step in pmpkg.install_package. And roll in your other suggested improvements as well. Piece o' cake! :) Thanks, Jeremy