- changes:
* new DEVELOPER_MODE option: it will fail instead of just printing warnings
* package creation is now reproductible, the same package built twice will
have the same checksum
* pkg install, upgrade and fetch now automatically performs the update of the
remote metadata if necessary
* portmaster patch: use_pkgng is not nessecary anymore
* portmaster patch: new format you can just copy it in the
ports-mgmt/portmaster/files
* portmaster patch: support for portmaster -o
* portmaster patch: Enable --check-depends, aliasing pkg check -adv
* pkg set -A to mark a package as automatically installed (-a now matches all:
better consistency with other commands)
* new -D option for pkg info (same as pkg_info -D)
* pkg set -o now change origin for a port and everything that depends on it
* new -L option for pkg {fetch,install,upgrade} to skip updating metadata.
* when a package isn't valid, print human information about what is missing
* some fixes in ordering (revealed by p5-XML-Sax mess)
Thanks to all contributors
* Add a check for over-specific shared lib dependencies [1]
* Add support for TEST_DEPENDS [2]
* Make the error for NO_CHECKSUM to be scarier [3]
PR: 165851 [1]
166173 [2]
167923 [3]
Submitted by: matthew [1]
sunpoet [2]
"Bryan Drewery" <bryan@shatow.net> [3]
- changes:
* Sanity check, clean all the old package not only the one that are concern by the run
* No more recursive cleanup in case of failure
* Code cleanup and consistency improved
Changes:
Fixe regression which causes pkgdb -u to take a long time without pkgng support.
PR: ports/167823
Submitted by: "Bryan Drewery" <bryan@shatow.net> (maintainer)
Changes:
- Fix portsdb -U crash when building INDEX
- All libs are now installed into the pkgtools/ namespace
- Experimental PKGNG support.
portupgrade -o, pkgdb -F and package upgrades do not work yet.
Enable with WITH_PKGNG=yes in /etc/make.conf
PR: ports/167811
Submitted by: "Bryan Drewery" <bryan@shatow.net> (maintainer)
This version adds support for a top-level Makefile.local. Also
restored the priority-setting code, but without making it mandatory
(if the call to rtprio fails, the program will still runs, rather than
aborting as it formerly did, the last time this code was included).
PR: ports/167592
Submitted by: maintainer, conrads@cox.net
- changes:
* -a (all) is now befault behaviour of check, query, rquery
* speed up remote repository query
* fix excessive CPU usage on fetching
* improve man pages documentation
* only display filenames on fetching not full url
* fix an important bug on jobs ordering calculation
* lots of bugs fixes
- changes:
* jails are now real jail(8)
* only have network access during make fetch
* poudriere jail -i -j <yourjail> now shows the activity of the jail
* csup support for poudriere ports
* new poudriere cron and poudriere queue command
Changes:
Fixes regression introduced 2.4.9.4 release, causing failed upgrades to
be seen as successes. Reported by: Matthew D. Fuller
<fullermd@over-yonder.net>
PR: ports/167512
Submitted by: "Bryan Drewery" <bryan@shatow.net> (maintainer)
Changes:
Fixes regression introduced 2.4.9.4 release, causing failed upgrades to
be seen as successes. Reported by: Matthew D. Fuller
PR: ports/167512
Submitted by: Bryan Drewery <bryan@shatow.net> (maintainer)
- While we are here:
- Make the affected sed pattern more flexible with regard to the
repository path from which the port was checked out.
- When handling -d <dir>, if the port's directory does not exist,
fix the path used in the first check for an alternative path.
- Bump PORTREVISION.
PR: ports/166593
Submitted by: Johannes Joemann <joemann@beefree.free.de> (maintainer)
0.1.4 - 2012 April 22:
* Support pkgng as an alternative package manager
* Support installing packages from a repository
* Support port pre-cleaning
* Support dynamic defaults from /etc/make.conf (i.e. DEPENDS_(PRE)CLEAN)
* Major code cleanup and refactor effort
* Switch to a revised milestone schedule
* Fixes:
* Fix assertion when using PKGINSTALL builder
* If kevent fails with EINTR then retry command again
PR: ports/167210
Submitted by: maintainer, naylor.b.david@gmail.com
- changes:
* Fix to be downloaded calculation
* Respected rc.conf with handle_rc_scripts
* Improve man pages
* Fix pkg set -o
* new pkg rquery subcommand, do the same as pkg query but for remote
repositories
* Allow forcing installation of a package even on bad arches
* Add indexes to local database, speed up local queries
* Speed up file loading from database
* lock the database while executing a job
* improvements on new pkg version detection
- avoids issue with not having SSL support
- Adds LICENSE
- Uses proper RUBY_VER
- on request from myself, pkg-message was generalized by maintainer
PR: ports/167096
Submitted by: maintainer, bryan@shatow.net
- addresses not having SSL support in base [2]
- add LICENSE [2]
- some portlint findings [2]
PR: ports/167084
Submitted by: ksmakoto@dd.iij4u.or.jp [1], maintainer bryan@shatow.net [2]
- changes:
* lot's of improvement in man pages documentations
* pkg info and pkg search are now aligned like pkg_info
* pkg search -x is now default
* new pkg shell subcommand, this fires up a sqlite3 shell use with caution
* pkg which gains 2 new options: -q (quiet) -o (origin)
* fix pkg repo incremental
* pkg version can now have -egXx <pattern> options
* when is port is registred this is now notifified as an installation (means syslog informations :))
* force umask 022
- changes:
* Fix memory corruption
* Fix handle_rc_script when upgrading services are now really restarted
* Report Already installed/Not found when installing packages if any
* when trying to pkg add something that doesn't exists, recommend the user to use pkg install
* Fix abi detection on recent current
* pkg -v report the version, -vv more informaton, -vvv even more :)
* new pkg fetch command (not yet documented)
* code factorisation
* pkg repo is now incremental
* install/upgrade now first forces installation/upgrade of pkg before anything
Feature safe: yes
- changes:
* query -f has been replaced by query -F when querying a package (file) for consistency with pkg info
* fix autoremove recursion
* pkg set -o oldorigin:neworigin allow the user to modify the origin of a packages (useful for MOVED)
* several fixes on powerpc
* pkg query evaluation expression now accept !=
* pkg2ng is now able to migrate a local database without needed the ports tree
* speed up pkg check by only loading the informations needed for the operation it will do
* code cleanup in the library
* fix uname/gname lookup when creating packages
* new periodic script for pkg audit
* new keyword shlib, pkgng is able to load information about the needed shlibs for a package (disable by default)
* pkg info act as pkg info -a by default
* new install timestamp in the package database, available using pkg query "%t"
* pkg updating can now take a port name as argument to only show related UPDATING informations)
* -q (quiet) for install, delete, upgrade, update and autoremove
* new ABI detection string
Concerning the new ABI, to be able to determine which package can be installed on which destination, pkgng couldn't rely on the simple:
os version/machine, because some arches have incompatible ABIs. Can compute on its own the ABI, the string is the followin:
OSNAME:OSMAJORVERSION:OSARCH:OSCLASS(:...)
... means the arch specific extensions.
which means on:
* amd64: freebsd:9:x86:64
* i386: freebsd:9:x86
* arm: freebsd:9:arm:eb:oabi:softvp
etc
Be careful all the packages (.txz) files have to be rebuilt!!!
When binary upgradeing from pkg beta8, on amd64 do the following:
UNAME_m=freebsd:9:x86:64 pkg install -f pkg
on i386
UNAME_m=freebsd:9:x86:32 pkg install -f pkg
Bonus when run on a chroot/jail where ABI is different from the host, pkgng will dynamically determine the destination ABI.
Feature safe: yes
PS. I removed attribution in pkg-descr, that was in original patch, because
this feature is deprecated.
PR: 166063
Submitted by: Mamoru Sakaue <sakaue.mamoru at samurai.mwghennn dot net> (maintainer)
Feature safe: yes
on WITH_PKGNG being defined.
No change to the default package, so no version number or portrevision
bump.
Approved by: stefan (maintainer) shaun (mentor)
Feature safe: yes
Fix remote code execution which can occur with a specially crafted
audit file. The attacker would need to get the portaudit(1) to
download the bad audit database, e.g. by performing a man in the
middle attack.
Add signature verification of the portaudit database. The public key
is for the database generated for portaudit.FreeBSD.org is included
in the distribution.
(This parts add the portaudit public key missed in initial commit.)
Submitted by: Michael Gmelin <freebsd@grem.de>
Reported by: Michael Gmelin <freebsd@grem.de>, Joerg Scheinert
Security: Remote code execution
Security: http://vuxml.FreeBSD.org/6d329b64-6bbb-11e1-9166-001e4f0fb9b1.html
Feature safe: yes
With hat: so
Fix remote code execution which can occur with a specially crafted
audit file. The attacker would need to get the portaudit(1) to
download the bad audit database, e.g. by performing a man in the
middle attack.
Add signature verification of the portaudit database. The public key
is for the database generated for portaudit.FreeBSD.org is included
in the distribution.
Submitted by: Michael Gmelin <freebsd@grem.de>
Reported by: Michael Gmelin <freebsd@grem.de>, Joerg Scheinert
Security: Remote code execution
Security: http://vuxml.FreeBSD.org/6d329b64-6bbb-11e1-9166-001e4f0fb9b1.html
Feature safe: yes
With hat: so
- Remove conditionals for PERL_LEVEL < 501200
- Remove regression-test targets b/c this will be centralized in Mk/bsd.perl.mk
- Other minor cleanups
RUN_DEPENDS = ${BUILD_DEPENDS} -> RUN_DEPENDS:= ${BUILD_DEPENDS}
PR: ports/165605
Submitted by: pgollucci (myself)
Approved by: portmgr (linimon)
Exp Run by: linimon
Tested by: make index
- be more strict when checking nonstandart files
- add more whitelists
- speedup misspelled(wrong) USE_* whitelist validation, turn on it by default
Approved by: eadler (mentor)
* Include LICENSE_FILE and LICENSE_GROUPS in the regexp check. [1]
* Add LICENSE_DISTFILES. [1]
* Add a check for the reassignment of ECHO_MSG [2]
* Remove old X.Org checks [3]
* Add PROJECTHOST to the PORTNAME section [4]
Reminded by: Jan Beich <jbeich@tormail.net> [1]
Requested by: jgh [2]
Submitted by: eadler [3]
Alex Kozlov <spam@rm-rf.kiev.ua> [3]
Change initialize() routine to call getuid() to ensure the program is being
run as root before trying to call rtprio(). This allows the program to
be run by a non-privileged user in a jail(8) environment.
PR: ports/165681
Submitted by: Conrad J. Sabatier <conrads@cox.net> (maintainer)
Submitted by: Marco Steinbach <coco@executive-computing.de>
- If applicable, use CVS/Repository as the source of the port's
directory name,
- else try to fall back to the basename of the working directory
if the PORTNAME heuristic fails.
- The patches also fix
<http://sourceforge.net/support/tracker.php?aid=1969774>.
- Add CVSauto as a new diff mode:
- Handle added/deleted files without having to cvs add/delete them.
- Works with :pserver:anoncvs@anoncvs.fr.freebsd.org.
- When submitting you're asked to use "port submit -L" if files are
added or deleted, because that's FATAL for portlint.
- Don't use the full path of the working directory in diffs.
- Include a port's PORTEPOCH in update mode in the generated PR.
- To aid in debugging porttools include this information in the PR:
- porttools' PORTREVISION and PORTEPOCH (if defined),
- submit and diff mode used by the submitter.
- Bump PORTREVISION.
- Pass maintainership to submitter
PR: ports/139872
Submitted by: Johannes 5 Joemann <joemann@beefree.free.de>
to the standard "make readmes" for building the README.html files for the
FreeBSD Ports Collection, building the files in a fraction of the time of the
standard method.
PR: ports/165631
Submitted by: Conrad J. Sabatier <conrads@cox.net>
- changes:
* pkg info -R now show a full yaml (script, options, users and groups were missing in previous versions)
* new pkg query -e <evalutaion> <query-format> to be able to query packages base on boolean expression matching content of the package
* pkg check now works like other subcommand: -a for all, -gxX for (glob, regex, eregex)
* pkg check is now able to check the file checksum and print what has been modified since package installation (-s option)
* pkg check is now able to recompute the flatsize (-r) this can fix the flatsize of packages installed with version prior to beta7
* new pkg set currently can only be used to manually mark package as automatically installed or not
- Radical changes to the cache: now stores a record for every
file that could affect the generated INDEX, which ports
those files belong to and the timestamp of the file.
- Default update mechanism is now to compare timestamps with
what is in the cache. Doesn't need to parse the output of
any update mechanism (although you can still do that.)
- Added new make-readmes command to create a tree of README.html
files.
More details: http://www.infracaninophile.co.uk/portindex/index.xhtml
Approved by: shaun (mentor)
- changes:
* move commands man pages from section 1 to 8
* pkg check can now be run as a user (will print the missing dependencies if any)
* better handling of database creation error
* new option -f to install subcommand to force reinstallation of a package
* new option -f to upgrade subcommand to force reinstallation of the whole set of packages
* pkg audit can now take a package: name-version in argument
- changes:
* pkg register: now run inside a transaction, this speeds up things a lot (thank you avg)
* pkg version: -v now print name-version to match behaviour of legacy pkg_version
* pkg register: fix flatsize calculation
* new pkg-updating(1) man page (thanks beat)
GHC in the ports tree has been updated to version 7.0.4 and the port revision
for all the Haskell ports without version changes are bumped. Other per-port
updates are coming soon.
This update also incorporates some improvements for bsd.cabal.mk that makes
working with Haskell ports even easier.
Thanks ashish@ for the help!
Obtained from: FreeBSD Haskell
- changes:
* move all the private headers into a private subdirectory to prevent the temptation of using them in the frontend
* fix a miss behaviour in pkg query printing things it shouldn't in case of mistakes
- changes:
* do not assert in case of unfinished buffer but finish it instead
* more userfriendly error message for pkg audit
* print not found patterns for pkg info and pkg query
- changes:
* fix pkg add (it is been broken during beta2 sorry about that)
* fix manifest emitter: now really emit the pkg-message in message no more description
* add %M to pkg query to show messages
* concat all pkg-messages and show it once at the end of operations
* print messages about missing dependency during pkg add
- add a message to warn user that they need to migrate using pkg2ng
- add a mirror to download
- changes:
* update bundled sqlite to 3.7.10
* add -n to pkg create to avoid packing already packed packages
* Fix rc script handling
* safer sbuf usage (this fix ia64 support)
* report a list of failing packages if any with pkg add
* allow "<category>/<port>" pattern in pkg search
* add support for syslog (through a pkg.conf option)
* the portmaster patch in the git is now up to date with 3.11 and feature complete (thanks avilla)
* add -R to pkg info to print out the raw manifest of a given package
* fix pkg query "%a"
* more robust to errors when creating packages
Submit system and ports statistics to fbsdmon.org.
Statistics can be sent from command line, from rc script
or from periodic script.
WWW: http://fbsdmon.org/
PR: ports/164608
Submitted by: Grzegorz Blach <magik at roorback.net>
Approved by: rene (mentor)
Changes:
- add ARCH=i386 in make.conf on i386 jail for amd64 host [1]
- testport and bulk now uses the same package dir: ${POUDRIERE_DATA}/packages/<jailname>-<portstreename> [2]
- fix testport with pkgng
- remove an extras .txz when creating a jail > 9.0
Submitted by: martymac[1]
Suggested by: avilla[2]
FreeBSD. However, most people will find it useful to bulk build ports
for FreeBSD.
WWW: http://git.etoilebsd.net/poudriere/
Of course this is pkgng aware and it support incremental upgrade of binary bulks :)
if you want to use it please define WITH_PKGNG=yes in your /etc/make.conf
For more information read the announcement on the mailing list
With hat: portmgr
Non-Regression test: linimon
ports use BUILD_DEPENDS:= ${RUN_DEPENDS}. This patch fixes ports that are
currently broken. This is a temporary measure until we organically stop using
:= or someone(s) spend a lot of time changing all the ports over.
Explicit duplication > := > = and this just moves ports one step to the left
Approved by: portmgr
literal name_enable wherever possible, and ${name}_enable
when it's not, to prepare for the demise of set_rcvar().
In cases where I had to hand-edit unusual instances also
modify formatting slightly to be more uniform (and in
some cases, correct). This includes adding some $FreeBSD$
tags, and most importantly moving rcvar= to right after
name= so it's clear that one is derived from the other.
