FreeBSD 10.0-RC3 Now Available

[Warner Losh]

On Dec 26, 2013, at 9:20 PM, Glen Barber wrote:

> On Thu, Dec 26, 2013 at 08:08:32PM -0800, Tim Kientzle wrote:
>> On Dec 26, 2013, at 7:25 PM, Glen Barber <gjb at FreeBSD.org> wrote:
>> 
>>> Well, no. :(  The time-consuming part is the dependency chain for the
>>> build.  Such as, RPI-B needs python, gsed, and world+dog.  BeagleBone
>>> needs cross-gcc, and I'm sure other stuff.
>>> 
>>> It would be pretty cool if crochet could have a '-D' flag to 'show
>>> missing dependencies for board specified’.
>> 
>> I’ve not yet come up with a particularly clean way to
>> do that within Crochet, but there are a few ideas
>> I’ve not yet tried.
>> 
>> It does occur to me that at some point crochet
>> is trying to do package management and maybe
>> that’s a bad thing.
>> 
>> But here’s an idea that might get exactly that:
>> I’ve considered literally building board images as
>> ports/packages.
>> 
>> E.g.,
>> 
>> $ cd /usr/ports/freebsd/raspberry-pi
>> $ make
>> ... builds/installs python, gsed, boot bits, etc as necessary via port dependencies
>> ... builds world/kernel (using non-root path…)
>> $ sudo make install
>> ... creates disk image and installs world/kernel/boot bits
>> 
> 
> Hmm.  I think you're on to something here.

I'd include '-build' in the name, so we can still have port that's to install on the rpi...

>> I don’t think this actually requires much effort
>> to get this working, and it would have some
>> interesting side-effects (like having RPi images
>> spit out regularly by the package cluster).
>> 
> 
> Maybe we don't need to go that far.  But building a port that includes
> the dependencies needed to do the actual image build would be a *huge*
> bonus from the releng side.  Meaning, if there was a port that I could
> install that would give me all the necessary bits to create an image for
> a specific board, that's a big step forward.
> 
> Plus, the package builders don't actually 'buildkernel', so offloading
> releng tasks to portmgr is not really scalable (meaning, if head/
> arm/armv6 is broken, etc...).

I'm not sure I understand what you are saying here...

Warner