Suggestion: Static libs policy, a draft

[Peter Jones]

On Mon, 2006-11-27 at 23:09 -0800, Ulrich Drepper wrote:
> Bill Nottingham wrote:
> >> In other words, for all statically linked executables in Core 
> >> (recovering, init/boot time etc.), all the libraries which take place in 
> >> the correspond static linkage must be present.
> > 
> > Obviously. This is sort of prerequisite for self-hosting.
> 
> Well, not so fast.  There is no reason for booting to require static 
> linking.  It might even be that the static linking meanwhile increases 
> the ramdisk size (2.7M for both binaries).  In any case, we're not 
> booting from floppy disks anymore so this is not an issue.  The static 
> linking should be removed where it is currently used for booting.

And as of mkinitrd-6.0.1-1 , it mostly is.  Some of the utilities
mkinitrd pulls in are still statically linked (basically just lvm
really).  I'll release an update to use a dynamic lvm just as soon as
there's an lvm2 package that provides dynamic executables.  Until then,
the initrd is actually significantly _bigger_ with dynamic linking.

mkinitrd-6.0.1-1 should hit rawhide tomorrow.  Testing will be much
appreciated.

-- 
  Peter