[commit/sim] Force generation of configure and other stuff; Was: [commit/sim] switch autoconf 2.59

[Andrew Cagney <cagney@gnu.org>]

Hans-Peter Nilsson wrote:
> On Tue, 11 Jan 2005, Andrew Cagney wrote:
> 
>>Hans-Peter Nilsson wrote:
>>
>>>On Tue, 11 Jan 2005, Andrew Cagney wrote:
>>>
>>>>Before commiting this I did a full cross build (using gdb_mbuild.sh).
>>>
>>>Looks like that only builds gdb.
>>
>>Why do you say that (yes I can read the subject)?  It builds the GDB the
>>source tree (which includes the sim) and typically GDB won't link if the
>>sim build fails?
> 
> 
> Ah, well, then I'm fresh out of wild guesses why it wasn't
> noticed that the frv sim (and presumably all CGEN targets and
> SIM_AC_OUTPUT users) fails building.  To make sure nothing had
> crept in locally, I just rechecked; checked out a brand new
> "gdb" and did "$srcdir/configure --target=frv-elf && make all"
> elsewhere.  Host is i686-pc-linux-gnu (FC2), but I see this on
> Debian Woody too:

I've still a few ideas.

> ... (from $build/sim)
> configure: configuring in frv
> configure: running /bin/sh '/home/hp/newgdb/src/sim/frv/configure' --prefix=/usr/local
> '--cache-file=./config.cache' '--build=i686-pc-linux-gnu' '--host=i686-pc-linux-gnu' '--target=frv-elf' '--program-transform-name=s,^,frv-elf-,; ' '--srcdir=/home/hp/newgdb/src/sim' 'CC=gcc' 'CFLAGS=-g -O2' 'LDFLAGS=' 'build_alias=i686-pc-linux-gnu' 'host_alias=i686-pc-linux-gnu' 'target_alias=frv-elf' --cache-file=.././config.cache --srcdir=/home/hp/newgdb/src/sim/frv
> configure: loading cache .././config.cache
> configure: error: cannot run
> configure: error: /bin/sh
> '/home/hp/newgdb/src/sim/frv/configure' failed for frv
> 
> Re-doing the configure and make step, changing the "all" to
> "all-gdb" (as in gdb_mbuild.sh), causes the same failure.

That config failure reeks of the problem I fixed.  Seems that you need 
to really want to run the new autoconf (unlike the old) - even though I 
ran autoconf in all those directories it didn't do as I ask.  I've 
re-run it with --force and that's had an effect.

The next problem is somewhere in sim/, try running make all-sim; make 
all-sim.  The first run barfs with no Makefile created, while the second 
prints a fail but makefile still manages to exit with zero status.  A 
guess is that GDB then managed to stumble along and build.

This also rewrites SIM_AC_OUTPUT to be 2.59 ``compliant'' (well at least 
it now generates makefiles); and zaps any other stray AC_AUX_DIRECTORY 
macros.

With that committed, I'm seeing another autoconf bug.  It's using 
ac_objext before it's been defined and that is leading to autoconf test 
failures (it looks for "conftest." instead of "conftest.o").

:-/

Andrew