rtems version information/errata

[lange92 at 2067.resnet.uni.edu]

Hi all,
  I'm trying to install RTEMS-4.5.0 and gnat3.13 on a linux
system. However, I'm having a good bit of difficulty trying to figure out
what exactly is going on with the bit_ada script. I was unable to find an
ada_build_scripts.tar.gz in the 4.5.0 release, and decided that my best
bet was to use the one from the 4.0 release instead.
  I am now compiling on a linux-from-scratch machine, using gcc-2.8.1 and
gnat-3.13 to compile. Installed in my root filesystem are binutils-2.10,
glibc-2.1.3, libtool-1.3.5, bin86-0.15.1, and ld.so-1.9.9. gcc-2.8.1 and
gnat-3.13p are in the default locations, as per the instructions.
  I tried installing on a Debian system, but gave up on that because even
though the system passed the `gnatmake hello.adb` test on the web page,
the compile failed with dozens of errors, indicating that gcc didn't
support the various "--gnat" flags.
  So anyway, what are the correct files to compile together? I currently
have the following files in my oar/archive directory:

ada_build_scripts-4.0.0.tgz
binutils-2.9.1-rtems-gnat-3.12p-20000429.diff
binutils-2.9.1-rtems-gnat-3.12p-20000429.diff.gz
binutils-2.9.1-rtems-gnat-3.13p-20000429.diff
binutils-2.9.1.tar.gz
c_build_scripts-4.5.0.tgz
gcc-2.8.1-rtems-gnat-3.12p-20000429.diff.gz
gcc-2.8.1-rtems-gnat-3.13p-20000429.diff
gcc-2.8.1.tar.gz
gnat-3.13p-rtems-20000829.diff
gnat-3.13p-src.tar.gz
hello_world_c-4.5.0.tgz
newlib-1.8.2-rtems-20000606.diff.gz
newlib-1.8.2-rtems-gnat-3.13p-20000907.diff
newlib-1.8.2.tar.gz
rtems-4.5.0-beta3a-4.5.0.diff
rtems-4.5.0-beta3a-4.5.0.diff.gz
rtems-4.5.0.tgz

  I know there are many files in there relating to the gnat-3.12p
installation, but I was careful to prepare for the bit_ada script using
only the 3.13 files. Everything patched fine, and I don't have any .rej
files anywhere in the build directories. However, compilation fails with
the following output:

../varasm.o: In function `initializer_constant_valid_p':
../varasm.o(.text+0x5d0c): multiple definition of `initializer_constant_valid_p'
typeck2.o(.text+0x4d4): first defined here
ld: Warning: size of symbol `initializer_constant_valid_p' changed from 760 to 792 in ../varasm.o
make[2]: *** [../cc1plus] Error 1
make[2]: Leaving directory `/mnt/home/lange92/oar/tools/build-i386-elf-tools/gcc/cp'
make[1]: *** [cc1plus] Error 2
make[1]: Leaving directory `/mnt/home/lange92/oar/tools/build-i386-elf-tools/gcc'
make: *** [all-gcc] Error 2
ERROR: **** FAILED **** base build of C and C++ (step 1)

Started:   Fri Jan 19 02:24:38 CST 2001
Finished:  Fri Jan 19 02:45:43 CST 2001

  I did notice that the gnat/rtems documentation has not been updated
since 4.5.0-beta3 was released as 4.5.0, and there were some interesting
challenges locating some of the files, but other than that, things looked
okay until the compile failed. I also noticed that the gcc-281.dif file
that comes with gnat-3.13 patched the varasm.c file that wouldn't
satisfactorily assemble.

  What are the proper files and versions to build with? Should I try to go
with the 4.0.0 version or the 4.5.0 version? And should I use gnat-3.12 or
gnat-3.13?

  On a more personal note, I would like to get RTEMS to successfully
compile once, because I'm hoping to use RTEMS for the project we're
working on in class. My professor is developing a computer-controlled
train set, and I would prefer to build our controlling programs using
RTEMS, because it appears to be a considerably more mature project than
the MaRTE executive that is otherwise planned for use in class. Plus, the
documentation suggests that RTEMS has networking support, which could
prove very useful in the train project. (I personally would like to build
a remote-control for the train set, so it could be remotely and viewed
using a webcam, for example.) The target I'm trying to build for is an
i386 machine...

  I see several files for sparc and ppc targets, in at least .rpm
format. Does anyone have a working, precompiled, version of RTEMS for the
i386 (preferably i386-elf) platform, that could be easily installed
without hassling with the compilation?

  Thanks for the input.


Daniel Lange