On date Thursday 2007-03-01 23:15:30 +0100, Andreas Schwab configured this message: > Stefano Sabatini <stefano.sabatini-lala@xxxxxxxx> writes: > [CUT] > These substitutions are designed to be used in a makefile where the > variables are recursively expanded. > > > Unfortunately it's rather ad-hoc solution and not very clear from the > > point of view of the script.in reader, and it's not applicable to > > other scripting languages that don't support that particular > > shell-like syntax (e.g. lisp). > > For other languages besided makefiles it is recommended to do the > substitution not during configuration, but at build time using the > recursively expanded values in the makefile. > > [CUT] > > The problem with this approach is that the value is not recursively > expanded. The second level expansion could still refer to other > variables. For example, a recent addition to autoconf is the datarootdir > variable, and datadir is now defined in terms of datarootdir, which in > turn is defined in terms of prefix. A two level expansion of datadir will > no loger lead to a fully expanded value. I finally ended up in this way, with a rule in the Makefile.in that says: script: script.in sed -e 's|__SYSCONFDIR__|$(sysconfdir)|' \ -e 's|__PACKAGE_NAME__|@PACKAGE_NAME@|' \ -e 's|__PACKAGE_VERSION__|@PACKAGE_VERSION@|' script.in > script where I designed a special notation for all the variables in the script that will be expanded at build time, to avoid conflicts with the original content of the script. This definitively resolves the N-level expansion problem (expansion is performed in the Makefile), and maybe it's more straigthforward than the other approaches. Thanks all for the replies. Cheers -- Stefano Sabatini Linux user number 337176 (see http://counter.li.org) _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
