Stuff that I could fix myself

Some thoughts on usability from experiences of the last few days include: yum vs apt; and Arch with more grabability.

I tried to use yum to upgrade from Fedora Core 1 to 3 rather than walk ten metres to ask our operations staff where the CDs were. This didn’t completely break my system as I was informed it would, although I did end up trying to upgrade packages in sets of 10. I gave up because I broke X so badly that starting it caused the machine to reboot. But at least it was booting in the first place, which is better luck than James had going from Fedora Core 2 to 3. (My story ends with the CD upgrade by the way, which went fine.)

So, why was I doing this? First I’m used to Debian-like systems which are more fussy about having good upgrade paths (to be fair, they don’t usually support skipping a release like I was trying to do). And second, I’m used to apt.

Two things that apt does that I wish yum did too:

  1. When asked to upgrade some enormous number of packages, and two of the upgrades are set to fail due to dependency issues, apt will install the enormous number of packages, minus the broken ones, thereby avoiding you having to manually construct a list of packages which don’t appear to be broken.
  2. apt does not re-index the repositories every time you invoke it, it needs a special command. Very useful when you need to walk it through several steps (and I have to say, I’ve never had to do as much of that as I have over the last day, but that’s partly because the office mirror of FC3 is incomplete) and don’t want to waste thirty seconds on each step waiting for it to rebuild its meta-data.

I’ve also been messing with Arch. Well, I branched Planet again and I’m going to put my old changes in by hand, I can’t face walking it though 60 revisions while it tries to merge my old old old changes in. While I was messing with doing this, I felt clicky urges. I’m not a hugely spatial user normally, but I had a sudden desire to have a big screen full of colourful blocks representing Arch branches, and I would merge code between them by grabbing branches with the mouse and dragging them to the right place. I don’t have a good system yet for maintaining a mental model of what branches come from and should be merged into what points in the parent branch.