- The conditional that checked what kind of ld.so,conf has to be generated
was using $slibdir = lib64 in postmake to check if it is a 64bit build,
and because multilib support was hardcoded for 64 it injected a corresponding
ld.so.conf.
Now that we implemented multilib support by using buildloops $slibdir gets
altered to lib32 in the second loop while building 32bit libs.
This commit will change the test checking if multilib is enabled for the
target instead of using $slibdir. It also adopts the new filesystem layout
to the ld.so.conf.
- SCRIPTDIR was hardcoded to $(prefix)/lib which is not very flexible
in case we build multilib targets.
New behavior is to use the libdir provided by configure so we will
be able to handle lib, lib64 and lib32.
We have to use an 'objdir' directory for building glibc and make its
name unique by appending $arch_target to distinguish while building
a multilib capable glibc using buildloops.
- Since we build 32bit libs of glibc by using a second buildloop
if we build a multilib target, we have to assure that we build
64 and 32bit libs in separate directories.
- This patch has to be applied if we want to build "pure" 64bit builds,
that means no multilib support!
It fixes the ldd-rewrite sed rules according to the pure64 patch for
gcc.
Note:
Currently tested for x86-64 only!
With this fix stone will be able to detect drives properly, while using
one of the following RAID controllers:
- Mylex DAC960/AcceleRAID/eXtremeRAID PCI RAID Controllers
Driver: DAC960
Device: /dev/rd/c<controller>d<logical drive>p<partition>
- Compaq Intelligent Drive Array
Driver: cpqarray
Device: /dev/ida/c<controller>d<logical drive>p<partition>
- Compaq Next Generation Drive Array
Driver: cciss
Device: /dev/cciss/c<controller>d<logical drive>p<partition>
The drives were not properly detected by the stone installer because those
controllers are using a non-common way of representing devices (see above).
Alejandro Mery
Christian Wiese
Nagy Karoly Gabriel