This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Makefile bug?
- To: egcs at cygnus dot com
- Subject: Makefile bug?
- From: Aaron Jackson <jackson at negril dot msrce dot howard dot edu>
- Date: Sun, 28 Sep 1997 00:05:15 -0400
Tracing through the Makefiles I see from the top egcs-970924 directory
that the ranlib enviroment variable is passed to the gcc Makefile through
RANLIB_FOR_TARGET and RANLIB in both BASE_FLAGS_TO_PASS and EXTRA_GCC_FLAGS.
I then edited the gcc Makefile to echo RANLIB and RANLIB_FOR_TARGET
before the make bootstrap build takes place; both were set to 'ranlib'. After
looking through the gcc Makefile it would seem that the FORTRAN target, f771,
gets the RAINLAB_FOR_TARGET from FLAGS_TO_PASS and F77_FLAGS_TO_PASS. I
didn't see any pass of a RANLIB variable.
RANLIB_FOR_TARGET=$(TARGET_TOOLPREFIX)ranlib in the gcc
Makefile, which is /usr/local/alpha-dec-osf2.0/bin/ranlib for my system.
It would also seem that RAINLAB_FOR_TARGET is used for a cros compiler.
I then added a RANLIB definition to FLAGS_TO_PASS and F77_FLAGS_TO_PASS
in the gcc Makefile. To save space I also added an 'rm -rf stage1' after
the 'make stage2' command and a 'rm -rf stage2' in the compare target if
there is no .bad_compare file instead of just a 'true' statement.
I then went to the top egcs directory and did a 'make bootstrap'.
Except for quite a few warnings everyting built fine. So my questions
are:1) Should there be a RANLIB pass for the f77 target? This seems to
have fixed the build error I kept on getting. 2) How about a make
bootstrap-lean that has the three rm's mentioned above?
Aaron Jackson jackson@msrce.howard.edu