The problem I was trying to solve is with "R" *** caught bus error ***
address 5, cause 'invalid alignment'. The most recent gcc binary I
could find was 3.4.2 and I suspected getting bug fixes would be hard.
Recompiling R with 4.1.1. fixed the problem, and it runs better on an
old Sparc than on a newish AMD64!
However, building 4.1.1. on Solaris 8 wasn't problem free, so I thought
I'd share the problems and solutions and maybe someone could pass this
onto the relevant developers The problems were few and rather minor
and don't reflect on the exceptionally high quality of this release. I
did the build in /usr/local/gcc.
---
../gcc-4.1.1/libjava/classpath/configure:11740
QT_INCLUDE_DIR=$($PKG_CONFIG --variable=includedir QtGui)
syntax error: `QT_INCLUDE_DIR=$(' unexpected
fixed by editing ../gcc-4.1.1/libjava/classpath/configure
and changing #!/bin/sh to #!/bin/bash
----
missing makeinfo
fixed by: ln -s /opt/csw/bin/makeinfo /usr/local/bin/makeinfo
(it looks like /usr/local/bin/makeinfo must be hard wired someplace
because /opt/csw/bin was first in PATH)
---
xgcc: error trying to exec '/usr/local/gcc/./gcc/as': execv: No such
file or directory
fixed by editing /usr/local/gcc/./gcc/as and /usr/local/gcc/./gcc/collect-ld
to change #!bash to #!/bin/bash
Now this last problem might have been because CONFIG_SHELL=bash and maybe
CONFIG_SHELL should have been /bin/bash, in which case perhaps someone
could fix the documentation to say that CONFIG_SHELL should be the path
to the shell, and not the shell itself. I didn't try this fix...
---
Thanks for gcc 4.1.1 and I hope these notes are useful
Regards -- Frank Middleton
