On Tue, Oct 02, 2012 at 06:42:40PM -0500, Bruce Dubbs wrote:
> I am wondering about making a change to LFS to combine some of the root 
> directories and /usr.
> 
 My initial feeling was very negative - it doesn't seem to affect me
(/usr has always been part of / rather than separate on all my
builds), but I'm reluctant to do things which limit what others can
do with the system : separate / and /usr appears to be a
common-enough use-case that some people still want it.

 I'm also concerned about the possible effect on BLFS - for me, the
bootscript changes in LFS-7.0 were a disaster : when I stepped away
in early 2011 my desktop built had worked fine, by the time I came
back after LFS-7.0 a lot of the critical parts of BLFS (e.g. nfs)
were broken.  This time, I don't think it will be _my_ builds that
break, but I don't wish that aggravation on anyone.

> 
> While I'm at it, should we remove *.la files in the libraries:
> 
> find /usr/lib -name \*.la -delete
> 
> We can add that to Section 6.64 - Stripping Again.  What I've found is 
> that I get a lot of warning messages and sometimes failures when 
> packages try to use the .la files, but just removing them seems to fix 
> things up without causing other problems.
> 

 On BLFS, I've several times summarised my results from removing .la
files as "on x86_64 (pure64), ppc, ppc64 things mostly built ok, but
when I tried to upgrade packages a few months later, to fix new
vulnerabilities, I needed .la files and sometimes wrote-off systems
I had intended to keep as viable because too many packages needed to
be rebuilt".

 This time, I've looked at *all* my notes.  Some are now so old that
I cannot determine whether a system was x86_64 or i686, and some of
those x86_64 systems were  from clfs because LFS didn't support
x86_64 in those days.  My attempts to remove the .la files started
after December 2008 and ceased by September 2009.  The following
points (referencing BLFS - I don't *think* any of the LFS .la libs
were needed, but I'm not 100% sure) summarise what I found.  Perhaps
things are different with the current verison of libtool - I have no
interest in retesting.  I know that Armin has reported no problems,
and that Gilles seems happy, but I'm not willing to take the risk
again :

1. Perhaps even on i686, both mpeg123 and ImageMagick needed .la
files at runtime to be able to work correctly - I see that mpg123
couldn't play anything (noted in July 2010 on ppc64 and x86_64),
and I still remember that ImageMagick programs were unable to open
the delegates for various image files.

2. Several gnome2 libs on x86_64 needed .la files to build.

3. Various gstreamer-0.10 build problems - some were on ppc64, but
it looks as if the bad plugins on x86_64 were also affected.

4. When I upgraded to firefox-3.0.8 on *one* of my systems (probably
x86_64, but at least one other such system was fine) I had problems
with gnash needing libxml2.la and then audiofile.la - it's always
best to rebuild plugins if upgrading firefox, something I had
forgotten until it bit me on an LFS-6.6 system last month.

5. In April 2009 I discontinued a couple of systems after hitting
problems trying to upgrade poppler - variously requiring some or all
of libX11.la, libreetype.la, libexpat.la - at that point I gave up :
note that the initial build had deleted all of those .la files as
each package was installed, it was only the later upgrade which
failed.

 Summary - if deleting .la files works for you (particularly on
i686), be happy.  But don't expect me to endorse that approach, or
to offer help if you are unable to rebuild some of the packages (to
fix vulnerabilities) in the future.  And always remember to test
that your BLFS packages work (e.g. mpg123, ImageMagick, but
basically *everything") - particularly when you change to newer
versions - otherwise *you* will be the one creating the Farce in my
.sig.

ĸen
-- 
das eine Mal als Tragödie, das andere Mal als Farce
-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to