FW: AC_SUBST_FILE

[wolfgang haefelinger]

Hi,

my Makefile.in looks like this:

=== Makefile.in ==============

## ...

@stdrules@
===============================

where "stdrules" is a AC_SUBST_FILE() variable, ie. my
configure.in (yes, should be configure.ac - I know)
contains something like

 stdrules="scripts/stdrules.mk"
 AC_SUBST_FILE(stdrules)

The idea of "stdrules" is to provide all standard rules
in a automatic way, ie. I don't  want  to touch N Make-
files just because I made a small typo.

So for example, "scripts/stdrules.mk" shall contain a
rule to regenerate "Makefile" if corresponding 
Makefile.in has changed:

========= scripts/stdrules.mk ======================
#subdir = @????@

Makefile : @abs_top_srcdir@/$(subdir)/Makefile.in
  @echo "*** remake $(subdir)/Makefile"
  @cd @abs_builddir@ && /bin/sh ./config.status $(subdir)/Makefile
====================================================

To get this to work, substitution of  autoconf  vars
need to take place (and as a side issue here - I like
config.status to compute the value for $(subdir) ).

And I like to apologize that I'm not using "automake"

The solution I found then later myself is the same as
Olaf (thanks) sent me today:

--- autoconfig.ac ---
common_mk=common.mk
AC_SUBST_FILE(common_mk)
AC_CONFIG_FILES([common.mk:common.in])
-----------------------

This works fine but one has to make sure that "common.mk"
gets  generated   before   any  other  (Make)files  using 
"common.mk". In other words, "common.mk"  needs to be the
very first listed in AC_CONFIG_FILES ( or  the  last  if
autoconf one day change creation order).

So this solution works but appears a bit hackish..

Another solution is to create "common.mk" on the fly 
while running "configure". For instance:

--- autoconfig.ac ---
common_mk=common.mk
AC_SUBST_FILE(common_mk)

cat >> common.mk << EOF
 common rules go here ..
EOF
-----------------------

This works also but unfortunatly lot's of interesting
values are not fully computed. 

For example, the value of $exec_prefix appears to be always
NONE. No  idear  why  it has  not it's final value while my 
code gets executed.

Regards,
Wolfgang.