>>> "RW" == Ralf Wildenhues <[EMAIL PROTECTED]> writes:
[...] RW> While I understand this, I do think your change is unfortunate: RW> First, it require Autoconf to always use `mkdir' with an absolute RW> directory. This is neither documented nor is it a desirable situation: RW> http://lists.gnu.org/archive/html/autoconf-patches/2006-05/msg00052.html RW> but it ended up being part of Autoconf for simplicity of the code (or RW> maybe just by chance?): RW> http://lists.gnu.org/archive/html/autoconf-patches/2006-05/msg00142.html We could change Automake's case $mkdir_p in [\\/$]* | ?:[\\/]*) ;; */*) mkdir_p="\$(top_builddir)/$mkdir_p" ;; esac into case $mkdir_p in [\\/$]* | ?:[\\/]*) ;; */install-sh*) mkdir_p="\$(top_builddir)/$mkdir_p" ;; esac RW> In hindsight, it may have been easier to just have AC_PROG_MKDIR_P use RW> the mkdir_p variable; but back then we weren't sure whether Automake RW> would adopt it quickly, or incompatibilities could develop: RW> http://lists.gnu.org/archive/html/autoconf-patches/2006-05/msg00034.html RW> Oh well. A bit of a mess. I'm starting to wonder why we switched to this macro, and whether it wouldn't be wiser to revert all these changes back to the original AM_PROG_MKDIR_P. It seems the switch to AC_PROG_MKDIR_P only brought us trouble, and no improvement :( -- Alexandre Duret-Lutz Shared books are happy books. http://www.bookcrossing.com/friend/gadl