Updated buildscripts (2.2.7)

Matthew Burgess ca9mbu at hermes.sunderland.ac.uk
Sun Apr 13 03:08:50 PDT 2003


On Sun, 13 Apr 2003 09:07:55 +0000 (UTC)
gschafer at zip.com.au (Greg Schafer) wrote:

> On Sun, Apr 13, 2003 at 09:45:06AM +0100, Matthew Burgess wrote:
> > No it wasn't from the script, and there's the problem.  I have been
> > using both the hint and the script itself to make my own scripts up
> > from (in a bid to actually learn something rather than just have a
> 
> Me too :)
> 
> > ready-made plfs system).  I've now just found where PATH gets amended
> > in the scripts (at the end of binutils-pass2) but looking at the hint
> > it makes no mention of having to explicitly add /stage1/sbin to the
> > PATH (it only adds /stage1/bin).  Now here's a question - can this be
> > set in  the
> 
> My view is that /stage1/sbin does not need to be in the PATH. Chapter 5
> is meant to be built as non-root (tho' I know Ryan has indicated in the
> past he builds Ch 5 as root) so IMHO the PATH should reflect a non-root
> user.
> 
> In actual fact, ldconfig is not needed . Glibc "knows" where it is
> installed and all libs in /stage1/lib will be found by binaries linked
> against the glibc in /stage1/lib because the embedded dynamic linker in
> those binaries is the one in /stage/lib.

Alright, so glibc knows where it is installed, and we know that the
binaries will link against the correct glibc.  So you're now saying that
any app linked against that glibc will automatically attempt to find any
other library it needs (e.g. ncurses or bzip2) in /stage1/lib as well? 
Cool- wonderfully nifty stuff these GNU tools :)

> > One other thought - when someone get's around to updating the hint
> > could they sync up the use of patches and seds (I'm thinking of
> > ncurses here in
> 
> Not possible. I've written the hint to be somewhat in line with style of
> the book i.e. it assumes interactive typing of the commands. Ryan's
> scripts are exactly that, scripts. Lots of stuff gets done in scripts
> that one would never do interactively. Also keep in mind that the
> scripts are really just a PLFS test harness (as well as being bloody
> good examples of how to script an LFS build :-)

Hmmm maybe I should have said "could the patches be updated to include the
additional sed's that are in the scripts".  Off the top of my head I
think the ncurses script has a sed which replaces a non-standard C++
header #include with its standard equivalent.  The ncurses-5.3.etip.patch
doesn't affect that, or is this just a case of Ryan being as anal as
possible :) I half considered going through a couple of the smaller
packages and trying to remove some of the gcc invoked warnings from them
but the cold light of day has instilled a bit of sanity to me!

Thanks again,

Matt.
-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe lfs-dev' in the subject header of the message



More information about the lfs-dev mailing list