Re: Testing out jhbuild

[Allan Day <allanpday gmail com>]

Owen Taylor <otaylor redhat com> wrote:
> To me, the criterion for success is that someone can start from scratch,
> without knowing much about Linux development and have a working build
> within an hour or so, without having to babysit it. Any sort of
> babysitting makes things much longer for everybody, and basically
> impossible for the novice.

I'm really happy to see JHBuild getting some attention. These seem
like noble aims indeed. :)

> * I'm a bit skeptical of the existence of meta-gnome-core-shell,
>  which, as I understand it, is supposed to contain the set of things
>  that need to be built for gnome-shell to work at runtime - so,
>  e.g., dconf is just a run-time dependency, since the build-time
>  depdendency is just gsettings. But what if I wanted to hack on
>  nautilus or gnome-control-center rather than gnome-shell? Shouldn't we
>  shoot for:
>
>   jhbuild build <X>
>   jhbuild run <X>
>
>  working?

meta-gnome-core-shell (or something similar) seems useful to me, since
it gives you a minimal (ie. quick) version of the essential GNOME
experience (the shell, control center and everything that they
require). This is typically what early adopters and testers want, and
we are doing ourselves a favour if we try and get everyone using and
testing the core anyway. I'd actually prefer it if
meta-gnome-core-shell gets built by default rather than
meta-gnome-core and meta-gnome-apps-tested...

I'd also like to add that our current modulesets are rather
impenetrable to newcomers. Not only are their names confusing, but
they're undocumented. Fewer modulesets (two per release, perhaps?)
with saner names (gnome-core-3.2 and gnome-world-3.2, for example)
would be an improvement. An explanation of what each moduleset is
along with a list of the modules and metamodules in each one would
also help.

Allan
--
IRC:  aday on irc.gnome.org
Blog: http://afaikblog.wordpress.com/