Frequently Asked Questions

Why don’t you submit these to Bugzilla?

At one time I did submit new ebuilds to Gentoo’s bugzilla, but it tended to be very unproductive for me.

The purpose of this overlay is to provide packaging for things I use, and that means they’re used today.

Resolution for new package bugs tends to take an awful lot of time(if it happens at all), and when they are resolved there are occasionally incompatible changes made. Each incompatible change means I get to make the choice of whether I attempt to shift our users to the in-tree version via some possibly hairy migration path, or simply continue to maintain the out of tree ebuild indefinitely.

Basically the end result is just far more work, with fast approaching zero benefit from upstreaming. To see the amount of work involved check out ::shadow, it only exists to handle packages with migration issues and it demonstrates the situation quite well.

Why isn’t this listed in the layman repo?

I’ve never used layman, so it isn’t something I would test very often.

A few users have asked in the past, but they normally find the Installation method for layman satisfies their needs when it is pointed out. Perhaps, it could be more discoverable(ideas how to make it so are very welcome).

Why are there so many merge commits?

As a result of the way my overlays are managed I have to maintain a bunch of different branches to work around supporting the many different configurations; on top of gentoo-x86, stacked on the EADS main tree, using upstream’s python eclasses, &c. The simplest way to do it is using a very heavy branching structure.

If you scroll back through the repository history to a time when I wasn’t supporting people using portage as their package manager or gentoo-x86 as a base, you’ll see the history was a lot cleaner.