upgrade stops if one package cannot be downloaded

[Rehan Khan]

Imported: http://tracker.labix.org/issue115

Reason: Review usability issue

further details: https://blueprints.launchpad.net/smart/+spec/bug-reporting-migration

msg573 (view) Author: mikus Date: 2006-06-21.04:03:12

I would like to add some urgency to this bug.

With 'apt-get' I can use the -m flag to have that download manager *continue*
installing packages (whose dependencies were met) despite not all of the package
downloads being successful. That makes it *much* easier to examine why some
package were not installed. I do not know how to elicit similar behavior from
'smart'.

--------

Today (with smart --gui) I did an 'update', then 'upgrade', then 'apply'. Since
I have a slow dial-up line, I went away for a couple of hours (to give time for
the downloads and installs). When I came back, there was an error pop-up saying
that a specific package had a public key problem -- 'smart --gui' had ABORTED
the entire "install" portion of that 'apply'. [Turned out that was not the only
package that had a problem, but I would have needed to know enough to ask for
the log to be displayed to find out the names of the other problem packages -
what I saw from 'smart --gui' was only one error pop-up for that 'apply'
attempt. [I was given another error pop-up (with another package name) after I
had "unmarked" the first package name, and tried 'apply' again.]

So I did a view -> Hide unmarked, and manually specified 'Keep' for the packages
whose install 'smart --gui' had rejected. That finally let 'apply' install the
remaining packages (those not rejected by 'smart --gui').

I then went to the /var/lib/smart/packages directory, and issued the command
'rpm -Uvh *.rpm'. That (without fuss) installed the packages that 'smart-gui'
had refused to install !!!

If there was some way for me to tell 'smart --gui': "keep going", it would have
saved me a considerable amount of time and effort (e.g., in "unmarking" packages
so 'apply' could proceed).
.

[Rehan Khan (rasker)]

A possible solution would be to notify the user of failed downloads (but not stop the downloads) and redo the packageset during the preparing phase. Is this possible?