Re: [Tails-dev] Please start reviewing bugfix/7345-upgrade-f…

Supprimer ce message

Répondre à ce message
Auteur: intrigeri
Date:  
À: The Tails public development discussion list
Nouveaux-sujets: [Tails-dev] "Upgrade from ISO" path from 1.0.1 to 1.1 [Was: Please start reviewing bugfix/7345-upgrade-from-iso-from-1.0-to-1.1]
Sujet: Re: [Tails-dev] Please start reviewing bugfix/7345-upgrade-from-iso-from-1.0-to-1.1
Hi,

sajolida@??? wrote (21 Jun 2014 12:57:00 GMT) :
> Did you consider issuing an IUK that would only update liveusb-creator
> in 1.0.1? It could be published in parallel with Tails 1.1 or a few days
> in advance.


I've briefly considered it, and then more or less disregarded it
silently due to the larger amount of work, our scarse resources, and
the fact we're (I'm) already late and overwhelmed with everything we
need to fix in time for 1.1.

Of course, providing such an IUK would imply a nicer UX to those who
can install it. We still have to somehow deal with the others, though:
e.g. those running from a boot medium that's not been created by Tails
Installer, or on a stick that hasn't enough free space available
anymore, or on a system that has not enough free memory.

Given the current state of the 1.1 milestone, and the announced
release schedule, I'm not sure that we have the resources to
build/test/provide the IUK *and* to prepare/test/document the apt-get
update+install steps. If we have, fine. Else, we have to decide for
which users we want to optimize.

Personally, on the principle I'm fine with optimizing for those who
have a boot device created by Tails Installer, *but* it's way more
work to validate the IUK (see below), and 1.0.1 still has the
crazy-high memory limits when checking for upgrades that'll be fixed
in 1.1, so I'm afraid we could leave too many users on the "Upgrade
from ISO is unsupported" side if we go this way.

> The shitty part of this would be the ISO building and testing of that
> IUK, but maybe we can find a trick to create an IUK with that upgrade
> only, without having to go through the whole RM process.


We can probably imagine a few such tricks, and it would be good to
have on the long term. I don't see how we can validate these tricks
without going through the whole test suite, but at least we can skip
most of the release process.

Cheers,
--
intrigeri