On Mon, Jun 06, 2005 at 04:55:52PM -0500, James Chacon wrote:
> 
> Then it should stay in-tree somehow. A "script" to allow someone to do that
> will mean it works the day the script goes in and then immediately rots and
> never works/fits again.

I have thought that the ramdiskbin programs (if not the entire ramdisk?)
should be built earlier in the build process - eg at the same time as
'rescue'.

That, together with a script I have to determine the sizes of all its
components, a target size, and a saved list of the sizes of a 'good'
version would make it much more likely that any size increase will
be detected in a timely manner.

Of course, the actual recent bloat is in the kernel part.
(The previous one - 8 weeks ago - was the change to uvm_swap to use blists.
I haven't located this one yet, comparing against netbsd-3 might give a clue.)

Note also that disposing of rescue-tiny doesn't actually solve anything,
it just moves the goalposts.

Until/if we change the way boot CDs are generated, the amd64 install will
be the next to explode.

	David

-- 
David Laight: david@l8s.co.uk