This is part 2 to support additional platforms with the RC scripts provided in
pkgsrc, in privileged as well as in unprivileged mode, including on NetBSD.
These changes:
- install rc.subr to PKG_SYSCONFDIR if SYSCONFBASE is not /etc
- let RC scripts source rc.subr in PKG_SYSCONFDIR
- effectively allow unprivileged setups to run RC scripts easily
- in addition, give a chance for more platforms to run RC scripts
Tested in privileged and unprivileged modes on NetBSD/amd64, and unprivileged
mode on Darwin/amd64; submitted for review on tech-pkg@.
Changes since 21.1.0:
Added check for packages that have been migrated to Meson but still
refer to their previous build system, most often involving GNU make and
GNU autotools. Suggested by nia.
Changes since 20.4.1:
Error out on package directories that differ only in case. This ensures
that pkgsrc can be used on case-insensitive file systems as well, such
as on macOS or Windows.
Changes since 20.4.0:
When a package has MAINTAINER set and there are modified local files,
only remind the user once per package that changes should likely be
approved by the maintainer. Previously there was a reminder once per
file, which was just too much for a package that has Makefile and PLIST
and maybe some patches changed.
Small performance improvement for loading the doc/CHANGES files, since
that costs a few tenths of seconds at every startup.
Change the /dev/shm handling on Linux to account for systemd being a complete
nightmare and changing mount types behind your back automatically.
The previous fixed works fine, until it magically changes from a tmpfs with
1777 permissions to a devtmpfs with 0755 permissions, with obvious failures
resulting when building as non-root.
Tested on el6 and el7, with the latter now reliably able to build Python.
When using the -u option to find mismatched packages, mark each affected
package with 'mismatch=YES'. Otherwise, consider such packages as
targets to rebuild.
ok gdt@.
Whatever is in $prefix shouldn't affect the result -- both scanning
and building happen in the workers -- and it is a bit of a nasty
surprise for merely issuing bulkbuild-restart to nuke all the host's
installed packages.
ok joerg
## Version 20.12.1 (2020-12-30)
* Fix version selection of packages and dependencies for install. Previously
the first returned match would win, and preferred.conf was not considered
for dependencies.
- Support continuing to install to /var/db/pkg if it exists and the
new pkgdb doesn't.
In the future, we can warn about this once we have tested advice that
we can give to users who want to move the location of pkgdb.
- Don't do anything about /var/db/pkg on non-NetBSD-base.
This creates conflicts with other package managers that also install to
/var/db/pkg.
Changes since 4.95:
Do not complain about "Bogus: ${_SVN_PKGREVISION:...}" that are
generated by wip/mk/*-package.mk. Using PKGREVISION to track the
current date of the package is the wrong approach; PKGVERSION_NOREV
should be used instead. Anyway, this is nothing the average pkgsrc user
should be worried about, so shut up the warning for now.
https://mail-index.netbsd.org/pkgsrc-users/2020/12/16/msg032884.html
When set to "no", pkg_add will not issue a warning if the host OS version does
not exactly match the OS version the package was built on. This can be useful
on many OS where the kernel release version has nothing to do with userland
compatibility, or where it may differ without being ABI incompatible.
Ultimately it would be ideal if the version checks were smart enough across all
our supported OS to not need this, but until then this is useful for users who
know what they're doing.
Bump version to 20201211.
