[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: more for the mix

+++ Paul Jimenez [04-04-08 11:43 -0500]:
> One more idea for the mix (perhaps more as inspiration than
> actual use, since it's fairly specialized to one platform):  the
> 'vrdeb' system that the Agenda VR3 (a linux PDA) uses. (details at
> http://agendawiki.com/cgi-bin/aw.pl?VrdebHome)
> It's a patch-the-debian-distro method which leverages existing debian/
> files, but adds a 'romfiles' file to the debian dir that lists which
> files from the 'make install' output should be installed to the rom image.

That' more-or-less how emdebsys works (ignoring the separate dependency tree
it also maintains) - for each package you specify the files you actually
want in the final image.

> The entire output of 'make install' is installed to the dev enviroment
> (where space is presumably not a problem) allowing libs and docs and such
> to exist where they'll be useful without taking up space on the actual
> embedded image.

OK. This scheme works well when you are building a fixed filesystem image,
but maybe not so well if you are trying to build smaller packages? I do like
it though. It's simple and it's almost exactly what you want quite a lot of
the time. We could certainly very easily generate 'romfiles' files for a
pile of packages specifying the bits you actually need - this would move a
lot of such knowledgeout of the big emdebsys list into each package, which
is obviously where it should live. Can we think of a better name than

This scheme is probably compatible with fancier schemes to modify the build
rules too - you can use one or the other according to your needs (highly
emdebbed image vs. upgradeable packagified system).

> This seems like the best compromise to me - power of debian, but extended
> a bit.  Perhaps someday even standardized into the main debian distro.
> Which is also, I think, a worth long-term goal for emdebian to have.

That is certainly the goal. Any scheme we think up has to be 'saleable' to
the rest of Debian in the medium-long term. We're probably getting to the
stage where we need to float some of these ideas on debian-devel and see
what other good ideas/problems are broughtup. That needs a summary paper of
the thinking so far though (which we're working on).

Aleph One Ltd, Bottisham, CAMBRIDGE, CB5 9BA, UK  Tel +44 (0) 1223 811679
work: http://www.aleph1.co.uk/     play: http://www.chaos.org.uk/~wookey/

Reply to: