Re: [atomic-devel] How to deploy old commits/trees

[Colin Walters <walters verbum org>]

On Sun, Feb 1, 2015, at 12:01 PM, Dusty Mabe wrote:
> 
> Moving around to different commits should really be easier than this. Maybe I am doing it wrong? 

Agreed =) See https://bugzilla.gnome.org/show_bug.cgi?id=732526

This sort of stuff is a bit blocked behind what I'll call "command line design".  I made a major
mistake with ostree in having "ostree admin" be the sysadmin part.  Then there's
the distinction between rpm-ostree and ostree, which is itself confusing.  However,
I was trying to use the fact that there was a new command to refocus it around
system tasks first.

This case is one where it'd be great to have some "user stories" to run through, like
what happened here.  What should the commands that the admin runs look like?
How would they discover the commit ID?  

It'd be pretty easy to implement a command which would just fetch the commits
and not the rest of the metadata, so "ostree log fedora-atomic/rawhide/x86_64/docker-host"
would work.  But how much should admins reference the branch name versus
say just "atomic fetch-history [--depth=X]", "atomic deploy <commit>" or
"atomic deploy ~5" for the 5th previous commit, etc.