[libstdc++] patch for "make install" problems

[Uros Bizjak]

On Sat, 6 Jul 2002, Phil Edwards wrote:

> On Thu, Jul 04, 2002 at 11:30:46AM +0200, Uros Bizjak wrote:
> >   This is a patch for "make install" problems, described in 
> > http://gcc.gnu.org/ml/gcc/2002-07/msg00121.html
> 
> > -	"INSTALL=$(INSTALL)" \
> > -	"INSTALL_DATA=$(INSTALL_DATA)" \
> > -	"INSTALL_PROGRAM=$(INSTALL_PROGRAM)" \
> > -	"INSTALL_SCRIPT=$(INSTALL_SCRIPT)" \
> 
> This is probably incorrect.
> 
> 1)  You mentioned in the previous message about command-line overrides.  Are
>     you trying to override these variables on the "make install" command?

  No, I'm just doing "./configure", "make" and "make install" from
toplevel (gcc) directory. INSTALL_* values are sent from Makefile in
libstdc++-v3 directory to Makefiles in libstdc++-v3 subdirectories as
command line parameters in AM_MAKEFLAGS variable. In subdirectories,
Makefiles re-define INSTALL_* variables to correct values, but make
takes incorrect INSTALL_* values from toplevel Makefile.

> 2)  Are you configuring in the source directory?  If so, try a build in a
>     separate objdir as recommended by the GCC install instructions, and if
>     that fixes it, then please submit a bug report about in-srcdir builds.

  Yes, I was doing ./configure in toplevel directory. And ./configure
worked OK, "make" worked OK, but "make install" failed in libstdc++-v3.
Funny thig is, that sun's "make install" worked, and gcc was
instaled and worked OK with posted patch. I will try with recommended way
of building gcc.

  BTW: I would suggest some kind of warning in ./configure stage if
objdir==srcdir. People (like me), who are not familiar with gcc build
process will try to ./configure gcc from toplevel source dir (where most
other programs are configured before compilation).

  Thanks for your answer!

	Uros.