The channel ubuntu-touch/rc/bq-aquaris.en for device vegetahd appears to have no delta images

[John McAleely]

Examining:

http://system-image.ubuntu.com/ubuntu-touch/rc/bq-aquaris.en/vegetahd/index.json

There are two images (1, 2), both of type 'full'.

I would expect 2 to also refer somehow to a delta image, especially as two tarballs (ubuntu, custom) are identical

This results in an over-large OTA

[John McAleely (john.mcaleely)]

Curiously, this does not appear to impact:

http://system-image.ubuntu.com/ubuntu-touch/stable/bq-aquaris.en/vegetahd/index.json

[Steve Langasek (vorlon)]

On Thu, May 28, 2015 at 06:31:24PM -0000, John McAleely wrote:
> Curiously, this does not appear to impact:

> http://system-image.ubuntu.com/ubuntu-touch/stable/bq-
> aquaris.en/vegetahd/index.json

I ran into this behavior when creating a new test channel for validating the
fix for bug #1444347. Copying an image into the channel did not result in
the creation of a delta.

A look at bin/copy-image shows the following:

    # Generate a list of required deltas
    delta_base = []

    if args.destination_channel in conf.channels:
[...]

It turns out that copy-image will generate a delta inline if it doesn't
already exist, but *only* if the channel is configured in etc/config.

It is not otherwise required to configure channels in etc/config if they are
manual channels, so many of our manual channels (i.e., the user-facing ones
that are *not* used for smoke testing) do not have entries here.

I do not know why the ubuntu-touch/rc/bq-aquaris.en channel is impacted but
ubuntu-touch/stable/bq-aquaris.en is not. I would think it would be because
the channel where the images are initially imported -
ubuntu-touch/rc-proposed/bq-aquaris.en - declares
ubuntu-touch/stable/bq-aquaris.en in its 'deltabase' list, but it *also*
declares ubuntu-touch/rc/bq-aquaris.en in this list. So deltas for both
channels should already exist at import time, and copy-image's
non-generation of deltas should have no impact.

Since the copies of the images onto both channels happened rather recently
(27 May in the case of the rc channel, 28 May for the stable channel), we
can say with confidence that this is not a result of differences in the
configuration of the two channels in etc/config at the time of the copy.
And since both channels had the same base version, the delta is the same for
both targets, and so it also cannot be due to differences in the
configuration at the time of the image's generation.

So I'm not sure of the cause of the missing delta. I do think that
bin/copy-image's behavior of not processing deltas for target channels not
listed in etc/config is buggy.

[Steve Langasek (vorlon)]

the branch has been published to the system-image server.