lsbsi notes, gcc 3.4

"Wichmann, Mats D" <[email protected]> Tue, 21 Dec 2004 11:51:52 -0800
Newsgroups gmane.linux.lsb.implementation
Message-ID <A06801158AE07847B27A52C1A074BC1D06D9652B@fmsmsx404.amr.corp.intel.com>
I've been playing around some in odd moments (the builds
are slow) with building an lsbsi which uses gcc 3.4.=20
This basically works now, but there are some problems
remaining.

x86_64 builds cleanly

ia32 is not tested, but there's unlikely to be
  anything significant if x86_64 worked

ia64 builds cleanly all the way through phase3,
  but fails when phase 3 is started as a chroot
  to do some final fixups. the problem is that
  libunwind is built, and used, by gcc34, but is
  not installed in phase3.  adding a case to
  install it is easy (and tested, and works, but
  not checked in), but I'm also experimenting
  with some different install targets for phase 3=20
  gcc that will, if they work, eliminate the need
  for arch-specific xml code for the install.

ppc64 did not build bootstrap.  there's an existing
  bug that the bootstrap builds don't configure to
  use lib64 - the problem with this is not in what's
  built, but in what gets used from the native host
  due to bad assumptions (the other time I saw it,
  it was missing a 32-bit-compatibility build setup,
  which I didn't think I was using).  I think this
  bug is going to need fixing, but don't have more
  time to spend on it now.


We're going to need to address a couple of upgrade
questions for LSB 3:
- what to do with nptl, which we don't currently build
  and which, I think, didn't work when tried (Marvin?)
- what to do with the base kernel (which we currently
  only use for headers) which is pegged back at 2.4.24.