[Bug 1864252] Re: preseeded snap installs fail in images

[Robert C Jennings]

** Description changed:

+ [Impact]
+ 
  Images built with pre-seeded snaps contain insufficient assertion data
- causing boot to fail.  The snaps for seeding are downloaded with a
- custom snap tool for an earlier cohort API (now deprecated).  The
- assertions that it pulls are incomplete.  We could update that list and
- move to the new API but at this time the snap-tool provides no value
- compared to use of the snap CLI (cohort support has moved to the cli as
- well).  The development overhead of maintaining snap-tool in livecd-
- rootfs are not warranted.
+ causing boot to fail.
+ 
+ The snaps for seeding are downloaded with a custom snap tool for an
+ earlier cohort API (now deprecated).  The assertions that it pulls are
+ incomplete.  We could update that list and move to the new API but at
+ this time the snap-tool provides no value compared to use of the snap
+ CLI (cohort support has moved to the cli as well).  The development
+ overhead of maintaining snap-tool in livecd-rootfs are not warranted.
+ 
+ This patch removes the bespoke snap-tool and relies on the snap CLI
+ instead.
+ 
+ [Test Case]
+ 
+  * Produce images that include preseeded snaps (in
+ /var/lib/snapd/seed/*)
+ 
+  * Boot the resulting image and ensure that the snapd.seeded unit is
+ successful and the snaps (from the correct channels) show up in 'snap
+ list'
+ 
+ [Regression Potential]
+ 
+  * The interface for these two tools is consistent and the output should
+ be the same.  There's always a chance that snap-tool had quirks which a
+ move to the snap CLI uncovers, where the result would be different snaps
+ seeded from before the change. An example would be channel differences
+ before and after this change. I haven't seen issues in my testing and I
+ do think it's unlikely, mostly I'm suspicious of SRUs that don't list
+ any regression potentials.

** Changed in: livecd-rootfs (Ubuntu)
     Assignee: (unassigned) => Robert C Jennings (rcj)

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to livecd-rootfs in Ubuntu.
https://bugs.launchpad.net/bugs/1864252

Title:
  preseeded snap installs fail in images

Status in livecd-rootfs package in Ubuntu:
  New

Bug description:
  [Impact]

  Images built with pre-seeded snaps contain insufficient assertion data
  causing boot to fail.

  The snaps for seeding are downloaded with a custom snap tool for an
  earlier cohort API (now deprecated).  The assertions that it pulls are
  incomplete.  We could update that list and move to the new API but at
  this time the snap-tool provides no value compared to use of the snap
  CLI (cohort support has moved to the cli as well).  The development
  overhead of maintaining snap-tool in livecd-rootfs are not warranted.

  This patch removes the bespoke snap-tool and relies on the snap CLI
  instead.

  [Test Case]

   * Produce images that include preseeded snaps (in
  /var/lib/snapd/seed/*)

   * Boot the resulting image and ensure that the snapd.seeded unit is
  successful and the snaps (from the correct channels) show up in 'snap
  list'

  [Regression Potential]

   * The interface for these two tools is consistent and the output
  should be the same.  There's always a chance that snap-tool had quirks
  which a move to the snap CLI uncovers, where the result would be
  different snaps seeded from before the change. An example would be
  channel differences before and after this change. I haven't seen
  issues in my testing and I do think it's unlikely, mostly I'm
  suspicious of SRUs that don't list any regression potentials.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/livecd-rootfs/+bug/1864252/+subscriptions