Bug#606158: refreshed patch

[Adam Borowski]

On Sun, Dec 16, 2012 at 05:49:02PM +0000, Thorsten Glaser wrote:
> Adam Borowski dixit:
> 
> >Please apply, even on ext4 it makes a drastic speed-up.
> 
> It’s buggy:

I admit, I haven't reviewed it beyond rebasing Adrian's work and checking
that the basic functionality works.

> pbuilder_old_LD_PRELOAD is not initialised in all
> cases, and this can break (for example when already called with
> eatmydata and pbuilder’s automatic one was disabled).

So pbuilder_old_LD_PRELOAD="$LD_PRELOAD" needs to be moved before the "if".

I'm not sure what should happen if eatmydata was already enabled but
pbuilder is set to not use it.  There's no regression, at least -- you might
get some spam about the library not being present.

> It also doesn’t catch the case where eatmydata isn’t available
> outside of the build chroot but inside very well.

Again a simple fix, but considering eatmydata's size, a simple dependency
would be safer -- less moving parts.