Re: HJ's patches break 3.3 install?

[Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>]

On Mon, 10 Nov 2003, Richard Guenther wrote:

> On Mon, 10 Nov 2003, Richard Guenther wrote:
>
> > Hi!
> >
> > After todays upgrade of my 3.3 tree, I notice that the usual make -k
> > install anfter bootstrapping does not leave me with a working install, but
> > rather gives strange error-messages like
> >
> > In file included from
> > /home/rguenth/ix86/pooma/tat-serial/pooma/linux/src/Domain/DomainBase.h:63,
> >                  from
> > /home/rguenth/ix86/pooma/tat-serial/pooma/linux/src/Domain/Domain.h:65,
> >                  from
> > /home/rguenth/ix86/pooma/tat-serial/pooma/linux/src/Domain/Loc.h:58,
> >                  from
> > /home/rguenth/ix86/pooma/tat-serial/pooma/linux/src/Pooma/Domains.h:45,
> >                  from apressure.cpp:5:
> > /home/rguenth/ix86/pooma/tat-serial/pooma/linux/src/Domain/DomainIterator.h:50:33:
> > stddef.h: No such file or directory
> > /home/rguenth/ix86/pooma/tat-serial/pooma/linux/src/Domain/DomainIterator.h:51:20:
> > iterator: No such file or directory
> >
> > etc., while in fact, the cited includes are installed. Just doing make
> > install (dont know why I do -k all the time) stops at
>
> Even with the texinfo problem fixed and make install completing
> successfully the problem with the missing includes persists. For the
> record, I'm bootstrapping with
>
> installdir="/home/rguenth/ix86/$1-`date +%d%m%y`"
> export BOOT_CFLAGS="-O2 -g -march=athlon"
> ~/src/gcc/$1/configure --prefix=$installdir --enable-languages="c,c++,f77"
> --enable-threads=posix --enable-__cxa_atexit --disable-checking && make
> bootstrap && make install
>
> $1 being gcc3.3

It _seems_ the new gcc is confused by being called through a series of
symlinks:

 ~/bin/g++-3.3  ->  ~/bin/ix86/gcc3.3/bin/g++

and ~/bin/ix86/gcc3.3 -> ~/bin/ix86/gcc3.3-101103/

if calling ~/bin/ix86/gcc3.3/bin/g++ everything is fine, but if calling
~/bin/g++-3.3 or g++-3.3 (~/bin is in the PATH) I get the above failures.
This never happened before with 3.3 or CVS HEAD, so something must have
been changed here.

I see some paths encoded in cc1:

strings cc1 | grep rguenth

/home/rguenth/ix86/gcc3.3-101103/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/include
/home/rguenth/ix86/gcc3.3-101103/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/../../../../include/c++/3.3.3
/home/rguenth/ix86/gcc3.3-101103/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/../../../../include/c++/3.3.3/i686-pc-linux-gnu
/home/rguenth/ix86/gcc3.3-101103/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/../../../../include/c++/3.3.3/backward
/home/rguenth/ix86/gcc3.3-101103/include
/home/rguenth/ix86/gcc3.3-101103/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/include
/home/rguenth/ix86/gcc3.3-101103/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/../../../../i686-pc-linux-gnu/include
                                                             ^^^^^^^^^^^^^^^^^^^^^^^^

the last one doesnt exist. But maybe this is not the problem, as the same
is true for the working copy from Oct09.

Richard.

--
Richard Guenther <richard dot guenther at uni-tuebingen dot de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/