Re: Fortran bootstrap failure on freebsd2.2.6

[Craig Burley <burley at gnu dot org>]

>With the current CVS snapshot under i386-unknown-freebsd2.2.6 I get
>
>---- snip ----
>gmake[2]: Entering directory
>`/files/pfeifer/OBJ-1506-22:27/libraries/libf2c'
>rm -f stamp-libf77
>cd libF77; gmake "CC=/files/pfeifer/OBJ-1506-22:27/gcc/xgcc 
>  -B/files/pfeifer/OBJ-1506-22:27/gcc/" "CFLAGS=-g -O2"
>  "F2C_H_DIR=/files/pfeifer/egcs/libf2c"
>  "G2C_H_DIR=/files/pfeifer/OBJ-1506-22:27/libraries/libf2c"
>  "GCC_H_DIR=/files/pfeifer/egcs/include" "LIBCFLAGS=-g -O2" "PICFLAG="
>  "RUNTESTFLAGS=" all
>gmake[3]: Entering directory `/files/pfeifer/OBJ-1506-22:27/libraries/libf2c/libF77'
>Makefile:85: *** multiple target patterns.  Stop.
>gmake[3]: Leaving directory
>`/files/pfeifer/OBJ-1506-22:27/libraries/libf2c/libF77'
>gmake[2]: *** [stamp-libf77] Error 2
>gmake[2]: Leaving directory
>`/files/pfeifer/OBJ-1506-22:27/libraries/libf2c'
>gmake[1]: *** [all-target-libf2c] Error 2
>gmake[1]: Leaving directory `/files/pfeifer/OBJ-1506-22:27'
>gmake: *** [bootstrap] Error 2
>---- snip ----
>
>when doing a vanilla
>  /usr/bin/nohup gmake LANGUAGES="c c++" bootstrap &

Without the contents of the relevant Makefile, it's hard to guess,
but I haven't tried the latest snapshot myself.  If you could
look at /files/pfeifer/OBJ-1506-22:27/libraries/libf2c/libF77/Makefile
(maybe send it along), perhaps the problem would make itself
clear.  This area of egcs certainly has changed recently; it's even
possible that the CVS operations to install my latest patches to
this stuff were only partway completed when the snapshot itself
was built, leading to an out-of-phase snapshot.

>And why the h* do I get an error in the _Fortran_ frontend, when I
>explicitely asked for C and C++ only?!? 

Actually, egcs/libf2c isn't the Fortran *front end*, it's the
Fortran run-time library.  I don't know how/whether LANGUAGES
is expected to affect whether run-time libraries get built.  I can't
find any references to LANGUAGES in the Makefile.in files for
either egcs/libstdc++ or egcs/libio, FWIW.

It's easy enough to conditionalize all of libf2c building activity
on whether "[fF]77" is present in $(LANGUAGES), if people who
understand the architecture of egcs' directory layout and
configure/build process tell me it should.

In any case, I'd really like to get all this nailed down nice and
tight for egcs 1.1, at least in the g77 area.  My main problem is
not being sure I understand the whole vision of the directory
layout, plus I know I don't understand how cross-compilation works
these days (it's been maybe 18 months since I tried that), so I'm
hesitant to just assume I know how to "fix" the g77 portions of
the egcs build process, and have probably already broken things
that worked in the previuos snapshot.

        tq vm, (burley)

P.S. Is <egcs-bugs@cygnus.com> really the right list for reporting
bugs in *snapshots*?