Bug#977477: release-notes: Update apt upgrade guidance (Was: Re: Bug#977477: apt: Adding progression indication to apt-get output)
Julian Andres Klode, le mar. 15 déc. 2020 16:15:23 +0100, a ecrit:
> > The problem is that these are not equivalent: apt upgrade will attempt
> > to install additional packages required by newer versions of existing
> > packages. That can lead to conflicts/breaks with other existing
> > packages, and thus get into all the complexity that using apt-get
> > upgrade first avoids.
> You mean that using apt upgrade upgrades more packages already, and
> hence dist-upgrade has less conflicts?
No, I mean that apt upgrade will encounter more conflicts by trying to
install newer packages, which may break/conflict other old packages.
> :D You can argue that in circles, you don't know which is going to be
Whatever is "best" or "worse", what we want is that what users actually
use is what is documented and *tested*. If we fail that we'll continue
seeing users getting trapped into conflicts etc.
> Of course people are free to apt upgrade --without-new-pkgs. The
> optimal way to upgrade likely is
> apt upgrade --without-new-pkgs
> apt upgrade
> apt full-upgrade
> which is equivalent to
> apt-get upgrade
> apt-get upgrade --with-new-pkgs
> apt-get dist-upgrade
Will the second step nicely behave in the case of a new package that
conflicts/breaks with an old already-installed package?
> > The problem is then that actual users end up in *other* situations than
> > what would typically be tested according to the release notes.
> People should test apt for interactive systems, and apt-get for
> non-interactive systems, as always.
I believe few developpers know this, and have their hands used to
apt-get, not apt. As shown by the release notes which document using
apt-get, not apt.
> Enabling progress for apt-get - the legacy scripting frontend -
> is a no-go. As is removing it from apt - the interactive user's
So we'd rather make release-notes document using apt instead of
apt-get? I'm fine with that but we *ALSO* need to make sure that debian
developers actually *test* that path, and not the apt-get path.