Re: glibc 2.2 vs 2.3 issue back again
- From: John Reiser <jreiser@xxxxxxxxxxxx>
- Date: Mon, 02 Jan 2006 12:22:02 -0800
Denes Molnar wrote:
> ... so copying over from Fedora 4 is a problem then. Is there a way
> to avoid installing a new kernel (I fear breaking software suspend).
> Say, by compiling the required libs from source? If yes, is it only
> libpthread, or glibc 2.3 and all others as well?
libc, libpthread, ld-linux, and hundreds more all must match exactly,
except of course for the pieces that you do not use at all.
See a typical list (table) at http://bitwagon.com/rtldi/rtldi.html .
If you don't want to use a matching set that sombody else has
prepared already, then start at http://sourceware.org/glibc/
and proceed to ftp://sources.redhat.com/pub/glibc/releases/
to get your own sources for glibc-2.3 and later. Building takes
a few hours using 2GB of disk, 768MB of RAM, 1GHz CPU. The
advantage is that you select exactly which features get built.
This is also the disadvantage. It'll take a few hours to read
and understand the instructions, and to coordinate environment
definitions with your operating system kernel. It's a good idea
to compare and contrast what you're trying to do, with what a
Linux distribution does (say, rebuilding a Fedora glibc-*.src.rpm.)
> Or could the
> (Mathematica) executable itself require NPTL threads and, therefore,
> there is no way around getting a new kernel that supports that?
It is unlikely that Mathematica requires NPTL. If it does, then
its documentation should say so. It had to run on old libpthread,
and NPTL is only about 2 years old. "Ordinary" use of threads
that does not delve into implementation details runs on either,
and usually better under NPTL.
More hints: Run
readelf --version-info /path/to/mathematica
to see the GLIBC_x.y versions that are required, and
readelf --symbols /path/to/mathematica
for detailed symbol@xxxxxxxxx information. For example,
this will tell you the minimum GLIBC_2.3.z subversion you
require, and which symbol(s) trigger it.
- Prev by Date: Re: compile with libc5
- Next by Date: Re: glibc 2.2 vs 2.3 issue back again
- Previous by thread: Re: glibc 2.2 vs 2.3 issue back again
- Next by thread: Re: glibc 2.2 vs 2.3 issue back again