Comment 7 for bug 193186

Revision history for this message
C de-Avillez (hggdh2) wrote :

I do remember that, at least at one point in the hardy buildup, I did get an E-D-S upgrade to propose Evo removal (new E-D-S had already been put available in the repositories, but not the corresponding Evo). During Hardy development, the KDE4 updates would every so often cause this.

Every so often I get this type of upgrade behaviour during new releases build-time (and also on Debian -- duh --). So, you are not alone (and I guess a lot of others have also seen/got hit by this). So yes, this is interesting, but mostly to highlight the dangers of running Alpha code: it can be very dangerous to run a blind dist-upgrade.

What I do to survive in the bleeding edge:

1. I never run 'apt-get dist-upgrade' and then say 'Y' to the changes without checking what is being removed. Every time I am using a box that is running the bleeding edge, I *never* run 'apt-get dist-upgrade'. Instead...

2. I usually run Synaptic, click on "Mark all updates", then analyse what Synaptic tells me -- and every so often I cancel *all* changes, and go hunting for the, huh, bad, update(s), until I reach the one(s) that propose to remove what I want to keep -- then I unmark the change(s), and keep on. I guess once burned...

This process can be quite time-consuming if you dist-upgrade every few weeks (a lot of changes to check on worst scenario)... but you will not lose a loved piece of your system :-)

3. Still on Synaptic, after I have have /selected/reviewed all changes I want, I go to "Apply", and -- again -- I check the proposed changes, with special attention on the removal list (usually, if the "to be removed" list header is not bold, then I just go ahead and commit). Otherwise I again check what will be removed. I guess, sigh, twice burned, etc, etc.

In fact, while I was typing this response, I went into Synaptic and ran an update/mark all changes (so that I could get the actual headers), and found one library being removed. I cancelled all changes, searched for the culprits, and found that the update proposing the library removal was doing so because it was also requesting a new package version of the library to be installed.

So... this is, I guess, it. Please tell me if you consider my answer satisfactory (or go ahead and close this bug, if this is the case).