pkgme-devs team mailing list archive

[Jonathan Lange <jml@xxxxxxxxx>]

Hello pkgmers,

The other day I had a bit of a muck-around trying to write a packaging
backend for pkgme. I found the documentation really useful, and very
quickly got the hang of what I needed to do. So that was nice.

However, there were a whole bunch of niggles. I've filed all of these
as bugs and am listing them below. If there are themes, it's that
pkgme raises tracebacks where it should error more gently and that it
doesn't give backend authors enough information.

809422	Traceback when 'want' is not executable
809423	When no backend is available, pkgme tracebacks
809424	Traceback when only 'want' exists
809429	No way to list available backends
809431	No way to determine backend search paths
809440	No way of telling what pkgme thinks it's doing
809444	Traceback when 'all_info' not executable but 'want' is

I don't think any of these are particularly nasty.

I had a bit of a chat with James yesterday, and we both thought that
pkgme might have to grow sub-commands to deal with some of this.

We both agreed that it would be nice if 'pkgme' by itself kept doing
what it does today (i.e. packaging something). We also agreed that
would probably make discoverability hard.

James mentioned that options should be kept to a minimum, that pkgme
should be opinionated. I concurred.

jml