building rtems-cvs failed

[Ralf Corsepius]

On Tue, 2005-05-10 at 19:06 +0200, Philippe Simons wrote:
> On Tue, 10 May 2005 07:03:19 +0200, "Ralf Corsepius"
> <ralf.corsepius at rtems.org> said:
> > On Mon, 2005-05-09 at 12:41 -0500, Joel Sherrill  wrote:
> > > 
> > > Philippe Simons wrote:
> > > > this is for gcc
> > > > ../gcc-4.0.0/configure --target=arm-rtems --with-gnu-as --with-gnu-ld
> > > > --with-newlib --verbose --enable-threads --enable-languages="c,c++"
> > > > --prefix=/opt/rtems-4.7 --disable-multilib --disable-interwork
> > > > 
> > > > and this is for rtems
> > > > ../rtems/configure --target=arm-rtems --disable-posix
> > > > --disable-networking --disable-cxx --disable-itron --disable-rdbg
> > > > --enable-rtemsbsp=gp32 --prefix=/opt/rtems-4.7
> > > 
> > > Ralf, is this some odd type of configure/Makefile bug that is only 
> > > manifested when some timestamps trick rpcgen into running?
> > I guess so.
> > 
> > > I can't duplicate it.
> > Neither can I.
> > 
> > ATM, I suspect this to be the result of several bugs interacting ...
> > 
> > 1. He reports not to be using --enable-rpcgen. Without it, the rpcgen
> > generated programs are not supposed to be regenerated.
> > 
> > To be investigated, probably a bug in c/src/configure.ac.
> > 
> > 2. The error message indicates "cpp" complaining on "-B" options as part
> > of the call to rpcgen, despite the fact that rpcgen does not pass these
> > options to rpcgen.
> > 
> > This indicates the rpcgen Philippe is using is could be pulling these
> > flags "from the air". The question is why and where (environment,
> > Makefile, wrapper scripts?)
> > 
> > The only place "-B's" are used inside of the Makefile is as part of "CC"
> > and "CPP", which would indicate Philippe's rpcgen being sensitive to one
> > or more of these environment variables.
> > 
> > However, this would be non-documented behavior and also doesn't make
> > sense, because rpcgen is supposed to be using nothing but /lib/cpp
> > underneath.
> > 
> > Philippe, which OS are you using? Do you have CPP or CC explicitly set
> > in your environment? Do you have any GCC variable set in your
> > environment? Do you use any wrapper scripts to GCC?
> > 
> My box is a Gentoo 2005.0
Well, this might explain a lot ;-)

>  running on amd64, here is my env
> 
> INFODIR=/usr/share/info
> HOSTNAME=omega3000
> SHELL=/bin/bash
> TERM=xterm
> ANT_HOME=/usr/share/ant-core
> USER=root
> PRELINK_PATH_MASK=/usr/lib/gstreamer-0.8
> GDK_USE_XFT=1
> PAGER=/usr/bin/less
> CONFIG_PROTECT_MASK=/etc/gconf /etc/terminfo
> XINITRC=/etc/X11/xinit/xinitrc
> PATH=/opt/rtems-4.7/bin:/bin:/sbin:/usr/bin:/usr/sbin:/usr/local/bin:/opt/bin:/usr/x86_64-pc-linux-gnu/gcc-bin/3.4.3:/opt/ati/bin:/opt/blackdown-jdk-1.4.2.01/bin:/opt/blackdown-jdk-1.4.2.01/jre/bin
> INPUTRC=/etc/inputrc
> PWD=/root
> JAVA_HOME=/opt/blackdown-jdk-1.4.2.01
> EDITOR=/bin/nano
> JAVAC=/opt/blackdown-jdk-1.4.2.01/bin/javac
> PS1=\[\033[01;31m\]\h \[\033[01;34m\]\W \$ \[\033[00m\]
> JDK_HOME=/opt/blackdown-jdk-1.4.2.01
> SHLVL=1
> HOME=/root
> LESS=-R
> LOGNAME=root
> CVS_RSH=ssh
> GCC_SPECS=
> CLASSPATH=.
> LESSOPEN=|lesspipe.sh %s
> INFOPATH=/usr/share/info:/usr/share/binutils-data/x86_64-pc-linux-gnu/2.15.92.0.2/info:/usr/share/gcc-data/x86_64-pc-linux-gnu/3.4.3/info
> DISPLAY=:0.0
> LADSPA_PATH=/usr/lib64/ladspa
> OPENGL_PROFILE=ati
> CONFIG_PROTECT=/usr/lib/X11/xkb
> G_BROKEN_FILENAMES=1
> XAUTHORITY=/root/.xauthRunSJa
> _=/bin/env
> 
> what do you mean by a wrapper script?...
A script around gcc to hard-code certain options to it. Some people use
something of this kind to avoid having to pass options to gcc to it.


And what does 
"ls -l /lib/cpp"
say?

> I reconfigured rtems with --disable-rpcgen option, and now build fails
> with
> Making all in librdbg
> gmake[3]: Entering directory `/tmp/b-rtems/arm-rtems/c/gp32/librdbg'
> gmake[3]: *** No rule to make target
> `src/powerpc/new_exception_processing/remdeb.h', needed by `all'.  Stop.

This file is part of the sources, if it is missing, your cvs checkout is
incomplete.

> > > > and here the full error
> > > > gmake[3]: Entering directory `/tmp/b-rtems/arm-rtems/c/gp32/librdbg'
> > > > rm -f
> > > > ../../../../../rtems/c/src/librdbg/src/powerpc/new_exception_processing/remdeb.h;
> > > > ( cd ../../../../../rtems/c/src/librdbg && \
> > > > rpcgen -h -DFRONTEND=\"powerpc/new_exception_processing/remdeb_f.x\" \
> > > >   -o src/powerpc/new_exception_processing/remdeb.h src/remdeb.x )
All the code above should only be run with --enable-rpcgen enabled.

> > > > cpp: -B../../../lib/ -B../../../gp32/lib/: No such file or directory
> > > > rpcgen: C preprocessor failed with exit code 1
We still don't know why your rpcgen seems to use CC or CPP. 

Under linux, rpcgen is part of glibc2's sources. I had a look into
glibc2's sources in its CVS and could not find trace of rpcgen being
sensitive to CC or CPP. The same applies to FreeBSD, Cygwin and the
original Sun rpcgen. AFAIS, none of them is using CC nor CPP nor any
*FLAGS - If Gentoo's rpcgen does, it is broken.

Please send me your 
arm-rtems4.7/c/gp32/config.status
arm-rtems4.7/c/gp32/config.log
arm-rtems4.7/c/gp32/librdbg/Makefile

which should exist in your build tree after configure had been run on
PM.

Ralf