5.1-pre2 space measurements (ch5)

Ken Moffat ken at kenmoffat.uklinux.net
Sat May 8 13:07:04 PDT 2004


On Sat, 8 May 2004, Matthew Burgess wrote:

> On Sat, 8 May 2004 19:24:30 +0100 (BST)
> Ken Moffat <ken at kenmoffat.uklinux.net> wrote:
>
> >  As part of my tests, I'm reviewing the "required disk space" values.
>
> Thanks.  My builds actually log these, but I'm lazy and never actually
> look at them...maybe this will change now!
>
> > I'm assuming that the book's "MB" are what we now have to learn to
> > call"MiB", i.e. they're binary megabytes.
>
> When I were a lad, there was only one kind of megabyte (1024
> kilobytes, which was 1024 bytes).  No idea what LFS uses though, but I
> will say, when I update the package sizes (in chapter 3) I use
> bytes/1024.

 Sounds fair, up until MiB suddenly appeared in the kernel a MB was a
power of 2 except to disk manufacturers.

>
> > The space for the following packages seems wrong in the book. In most
> > cases, I blame the extra source you get with a version increment.
>
> Yep, that's likely the most common cause.
>
> > linux-2.4.26 (headers) - up to 193MiB (book: 186MB)
> >
> > glibc, with the minimal set of locales, 872MiB (book: 800MB) - we've
> > got nptl and linuxthreads both in the source now.
> >
> > dejagnu - I'm down to a smidgeon over 6MiB (book: 8.6MB), has it
> > really shrunk with version increments ?
>
> Yes, dejagnu-1.4.4 was largely rewritten, which could well have reduced
> it's size considerably.
>
> >  If these figures are worth maintaining, nobody should feel bad about
> > not recording and updating them every time there's a version change -
> > changing the toolchain also affects these.  It's a pre-release task,
> > which is why I'm mentioning it now.
>
> Thanks very much Ken, I'll update them tonight.  BTW: If anyone's got
> accurate and up-to-date lists of files installed I'd appreciate them :)
>

Never got around to that part.  Thanks for picking these up Matt, as
ever the numbers are E&OE.  I'll look at chapter 6 later.

Any view on what sort of host we should be using for the quoted SBUs ?
I'm thinking primarily about the version of gcc, even 3.2.2 will make a
difference on the bigger packages.

Also, of course, do we run the tests or not when we produce the SBUs and
space measurements ?  As I said the other day when belgarath was acting
up, for non-toolchain chapter 5 I think we ought to do the figures
without the tests, but for chapter 6 we should be running the tests
(i.e. what we recommend to ordinary builders).

Ken
-- 
 das eine Mal als Tragödie, das andere Mal als Farce




More information about the lfs-dev mailing list