Bruno Haible wrote:
Sam Steingold wrote:
gnulib/gnulib/m4/stdint offers a multi-line sed rule for generating
stdint.h from stdint_.h,
why do I need to maintain the sed command by hand?
why can't this be done by config.status?
Two answers:
Regarding the integration in GNU clisp: Since 2004, when gnulib took stdint_.h
from clisp, gnulib has evolved a bit. Nowadays nearly all users of gnulib use
gnulib-tool to assemble the pieces from various modules into their source
code hierarchy. The old "merge diffs by hand" approach become too burdensome,
even for die-hards like Paul and me. Therefore what we recommend now is to
use gnulib-tool. In the case of clisp it can be like this:
gnulib-tool --import --source-base=src/gllib --m4-base=src/glm4 \
--aux-dir=src/build-aux --no-changelog \
stdint stdbool ...
this printed
gnulib-tool: module config.guess doesn't exist
gnulib-tool: module config.rpath doesn't exist
gnulib-tool: module config.sub doesn't exist
gnulib-tool: module depcomp doesn't exist
gnulib-tool: module install-sh doesn't exist
gnulib-tool: module missing doesn't exist
even though I copied these files into src/build-aux
this also printed some suggestions that I followed.
now, configure.in starts with
AC_CONFIG_SRCDIR(lispbibl.d)
AC_CONFIG_AUX_DIR(build-aux)
AM_INIT_AUTOMAKE
and has
AC_CONFIG_FILES([po/Makefile.in gllib/Makefile])
and it fails with
config.status: creating gllib/Makefile
config.status: error: cannot find input file: gllib/Makefile.in
I hoped that automake would generate gllib/Makefile.in out of
gllib/Makefile.am automatically, but this does not happen (why?)
So I had to run "automake gllib/Makefile" by hand. is this right?
do I now commit the generated gllib/Makefile.in to the CVS?
thanks.
Sam.
