from archivers/pax and net/tnftp, respectively. In the past, the
pkgtools version of these packages installed into ${PKG_TOOLS_BIN},
but this was changed in:
pkgtools/pax/Makefile:1.15
pkgtools/tnftp/Makefile:1.3
+ Get rid of archivers/pax/Makefile.common and net/tnftp/Makefile.common
by merging them into their respective Makefiles. The Makefile.common
files existed solely for inclusion by the pkgtools versions of these
packages, but with the removal of those packages, these files are
now unnecessary.
+ Add full DESTDIR support to archivers/pax and net/tnftp.
+ Modify the bootstrap to build archivers/pax and net/tnftp instead of
the pkgtools versions of these packages.
- /var as varbase is only used for privileged builds without prefix
or when prefix is explicitly set to /usr/pkg
- if prefix is set to a non-standard value, derive varbase from it
- derive pkgdbdir from varbase
Based on the discussion in PR 37796 and with jlam@.
and nothing else. This prevents bootstrap from exiting just because
there is a subdirectory named "awk" (or another tool) in one of the PATH
directories.
Fixes PR 37806.
code should have been here since the very beginning of bootstrap.
Additionally, the paths are checked that they only contain characters
from the "Portable Filename Character Set" (IEEE 2003.1, definition 3.276).
Motivated by http://mail-index.netbsd.org/pkgsrc-users/2007/10/17/0000.html
expect full path names. Use the default locations in /bin unless
overriden explicitly. Bump revision of bmake.
On Solaris, use /usr/xpg4/bin/sh if it exists.
bootstrap time these variables are already set as shell variables. As a
result, they were doubled for bootstrap compiling.
An alternative solution to that issue would be to explicitly empty those
shell variables once they have been added to BOOTSTRAP_MKCONF, but that
approach would use more lines of shell code :-)
- phase one builds the essential tools in the bare minimal version
needed by the infrastructure to run "make install".
- phase two runs "make install" for all the bootstrap packages.
Set WRKOBJDIR for the second phase, we never want to leave garbage
around.
This increases the time for running bootstrap, but gives more
deterministic results. It also means that e.g. configuration files
in pkg_install can be handled normally. It is a prerequirement to
sanely allow pkg_install some more extended work like building
its own libarchive without having to worry too much about limitations
on some platforms.
This fixes the expansion of @gzcat@ in the download-vulnerability-list
script.
Tested by tnn@ on Interix and myself on DragonFly.
even if standard input is the empty string. Install a wrapper script
bundled with pkgtools/bootstrap-extras to deal with this. This is an
attempt at a permanent workaround for the problem described in PR pkg/25777
which has regressed since it was initially fixed.
We can now bootstrap again on Interix 3.5.
search order means that the system libedit gets found first. This
hasn't been much of a problem until now, but the system libedit
lacks support for Apple's 64-bit architectures, which breaks the
Universal build. Therefore, force linking with tnftp's internal
libedit by replacing "-ledit" in ${LIBS} with "../libedit/libedit.a".
Workaround suggested by tls.
with /Developer/Applications/Utilities/PackageMaker.app, then
converted from blob to the equivalent XML with:
$ plutil -convert xml1 macpkg.pmproj
mkbinarykit will run this file through a bit of sed, then use the
result to generate a double-clickable Mac package.
use inplace libnbcompat like the direct build from bootstrap
does. Fix a long-standing bug in mkbinarykit. The default work
directory was "work" (relative) and therefore MAKECONF=work/...
was passed down, effectively making it a nop. That resulted in
digest being incorrectly installed during the bootstrap.
Tested on DragonFly by myself and on OSF by tnn@, discussed with jlam@
various packages, pass down PKG_PRESERVE. Remove the default setting of
PKG_PRESERVE, it makes perfect sense to install e.g. pkgtools/mtree
and be able to remove it again, if it is not an essientiel tool.
OK jlam@
