Re: bootstrap/8657: No rule to make target `bootstrap' in directory 'gcc'

[Andris Pavenis <pavenis at latnet dot lv>]

On Thursday 21 November 2002 08:50, Neil Booth wrote:
> Zack Weinberg wrote:-
>
> > In my experience, if objdir is a subdirectory of srcdir AND configure
> > is invoked by a relative pathname, it doesn't work.  But if configure
> > is invoked by an absolute pathname, objdir can happily be a
> > subdirectory of srcdir; and if objdir is not a subdirectory of srcdir,
> > relative paths work fine.
> >
> > I don't remember the exact failure mode, but it was clearly a case of
> > some shell script fragment somewhere getting mixed up about how many
> > ../ components it needed to put in a pathname.
>
> For nearly 3 years I have used
>
> ../configure --enable-languages=whatever --prefix=whatever
>
> and it works fine.  I think a '~' in prefix used to always work, but
> that broke recently in libstdc++.

I also up to recently used similar directory layout for building GCC for 
target i586-pc-msdosdjgpp (build directory was subdirectory of source one
and I run configure using relative path). With GCC-3.2 I found that
I'm getting trouble with 'make -C gcc gnatlib_and_tools'. Moving build
directory out of source one fixed problem and I'm using relative path to 
configure (like ../gcc-3.2/configure ....).

So my experience shows that simple moving build directory from
build.djg to ../build.gcc relative to source directory fixed build failure
with building gnatlib and Ada tools. I didn't study this problem more 
detailed though. I remember trying something similar under Linux and
getting the same problem, but I may be wrong.

This was really only problem with build directory being subdirectory of
source directory I have noticed up to this time. 

Andris