Jeremy Katz wrote:
> XO Bundles, which are the normal activity distribution mechanism
for
> OLPC are self contained.
>
It seems like they contain more than just the "binary" bits, also
including most of the source tree
Any example of this other than setup.py (and the activity python scripts)?
including things like setup.py and the
raw po files that shouldn't be needed at all at runtime. Is this
intentional or just something that should be fixed up in bundlebuilder?
I think the idea there is that kids should be able to modify activities
directly (starting from the installed .xo) and redistribute them.
The use case is obviously not relevant to activity rpms though.
> Now, when installing into the system with rpm, it would actually
make
> more sense to use the standard directory layout. I'm only a little
> worried that it would complicate the build system of activities which
> doesn't use Sugar bundlebuilder.
>
What proportion of activities don't use bundlebuilder? All the ones
I've looked at have, but that is admittedly a small-ish sample set
Vast majority uses bundlebuilder. I can only think of Etoys which
doesn't right now.
> Note that we have to extend/replace find-lang.sh regardless, to
be aware
> of localized activity.info.
>
That's a quick fix that I can get into upstream rpm once we are agreed
on going in that direction much like we already do for a few other
things.
I'm inclined to go in that direction, yeah. I'll write a patch for
bundlebuilder and see the feedback I get about it by the rest of the team.
Thanks!
Marco