Hi,

We used to have a some configure hacks to get libunwind configured and
build for the different architectures we support. This patch adds some
different configure magic so we can use "normal" recursive configure.
Which has as advantage that you can just pass top level configure
options (--enable-debug is really handy when hacking on libunwind
stuff). And it makes sure that we use the same options for all target
builds (previously the 'native' target would be build without
UNW_REMOTE_ONLY, but all 'secondary' targets would be build with it.)

It works because git allows symlinks in the repository. So no more
tricks to get the right directories created, .keep_me files.

frysk-imports/ChangeLog
2007-12-20  Mark Wielaard  <mwielaard@redhat.com>

       * .gitignore: Remove libunwind-*.
       * Makefile.am (LIBUNWIND_DIR): Set to either empty or canonical
       libunwind dir.
       (LIBUNWIND_ARCH_DIRS): Define.
       (LIBUNWIND_OTHER_DIRS): Removed.
       (CHECK_SUBDIRS): Include LIBUNWIND_DIR (if defined).
       (SUBDIRS): Include LIBUNWIND_ARCH_DIRS.
       * bootstrap.sh: Don't create libunwind-* dirs.
       * configure.ac: Don't create and configure libunwind-* dirs
       explicitly.
       (AC_CONFIG_SUBDIRS): Add libunwind-* dirs.

frysk-imports/libunwind/ChangeLog
2007-12-20  Mark Wielaard  <mwielaard@redhat.com>

       * configure.in (is_local_frysk_build): New function.
       (frysk_local_target_cpu): New function.
       Define target_cpu, target_alias and UNW_REMOTE_ONLY based on new
       functions.

Tested with several builds on x86_64, x86, with and without
--enable-debug and --enable-check-libunwind and TestRunner -arch 32 on
the 64bit system.

Cheers,

Mark