default installation paths to be inside ~/pkg and define UNPRIVILEGED=yes
in the generated mk.conf. This lets regular users to simply bootstrap by
doing './bootstrap --ignore-user-check'.
parsing code. For maximum portability it uses the expr(1) command
instead of sed(1), the same way as it is done in the core of the latest
GNU configure scripts.
(1) rework how command-line arguments are parsed:
instead of --command=<arg>, use --command <arg>
This allows us to not rely on certain commands for which we first need
to figure out where they are to parse the arguments, which in turn
allows us to
(2) add the command-line option
--preserve-path
to prevent bootstrap from munging the PATH (as it does on some platforms)
and look in places that are not currently in the PATH
Finally,
(3) add a check to see if we're using gcc, and set and add the
PKGSRC_COMPILER=<compiler>
flag to the sample mk.conf. This is particularly useful (and actually
necessary) under IRIX.
Bump BOOTSTRAP_VERSION.
PKGSRC_COMPILER=mipspro
if this compiler is used. Otherwise, wrong CFLAGS might be passed.
XXX: we probably want to add a check for the proper compiler into the
bootstrap process and add this line automatically, if necessary.
Fix the override logic for $opsys.bsd.{lib,man}.mk to install the files
in the correct place -- this was previously all kinds of b0rken. Now it's
possible to build shlibs properly on Interix using <bsd.lib.mk>, and may
be possible on Darwin as well.
what files exist in the fs.
This works around a problem that crops up when using a nfs-mounted pkgsrc
repository under Interix. (When going to "su" during the install phase,
bmake sees "makefile" for a moment in lieu of "Makefile", and all hell
breaks loose.)
Remove some code which makes file lookup rely on the fact that
the first two directory entries are "." and "..".
This behaviour is not required by applicable standards, and
actually not provided by "coda".
Now we get the "." and ".." into the per-directiry hash tables,
but this should not hurt.
fixes bmake build on Fedora Core 2, PR pkg/26140 from Shoichi Miyake.
There's one place where you absolutely *must* use bmake: when building
pkg_install. Otherwise its Makefiles will attempt to get $(MACHINE_ARCH)
from the system make, which is not likely to be correct on several
platforms.
