portupgrade:
- Fix quotes.
[Submitted by: Jimmy Olgeni <olgeni@uli.it>]
- Do not ask whether to install a port twice. (portinstall/-N)
- Complete missing words in the manpage.
pkgdb:
- Comment out deleted pkgdeps properly.
misc.:
- Convert the install script to Makefiles.
- Change the examples/docs subdirectory name from `portupgrade' to
`pkgtools'.
reached the one step higher level. It's *really* powerful. Enjoy! :)
portupgrade:
- Greatly improve the procedure of binary upgrading (-P/-PP):
When a fetched package is not of the latest version:
1) If -PP is given and the package is at least newer than the
current installation, put up with it and do an upgrade.
2) If -PP is given and the package is now newer than the current
installation, give up.
3) Else, fall back to the port.
This way you can now do binary upgrades just as you'd expect.
- Improve the new installation procedure as well.
- Allow using ports glob patterns in the configutation variables
MAKE_ARGS and HOLD_PORTS.
- If a port/package matches multiple entries of MAKE_ARGS, join all
the arguments using the space as separator.
pkg_fetch:
- Introduce a new {environment,configuration} variable PKG_SITES.
You can now specify multiple URI's to fetch packages from.
- When a URI is given, check for the last path component and if it is
not "Latest" or "All", fetch the dependent ports from the same
directory. (-R)
- Do not repeat same error messages.
- Properly reflect the environment variables defined in pkgtools.conf.
pkg_deinstall:
- Add a timestamp hack as well as portupgrade to let the pkgdb engine
properly detect an update of PKG_DBDIR.
pkgdb:
- When the user chooses to delete a dependency, comment the line out
instead of deleting it.
pkgtools.conf:
- Provide some useful predefined constants and functions.
people's help. Thanks so much!
portupgrade:
- D'oh! Fix a stupid bug where portupgrade didn't modify
dependencies when upgrading a package while portupgrade _is_ meant
to do that. ;) But you don't need to worry: `pkgdb -F' can always
handle that situation.
This bug had been there since 2 October, and was finally exposed by
the newly introduced dependency sanity checks. I guess you has been
annoyed by the warnings, but they are gone now. ;)
Reported by: Ollivier Robert <roberto@eurocontrol.fr>
- Make portupgrade download packages from a remote site if -P is
specified. Now one can do binary upgrading in a handy way:
portupgrade -P foo bar
Instead of:
portupgrade -FP foo bar && portupgrade -P foo bar
Although the latter is friendlier to dialup users.
- Add a delay between deinstallation and installation to let pkgdb
properly detect the update of PKG_DBDIR entries.
- Since pkg_add(1) is mute, always show a progress message when
invoking it no matter whether -v is given or not.
- Make -f override "hold" marks set in pkgtools.conf.
pkgdb:
- Since the dbm routines in libc seem to dead lock in some cases,
slightly change the DB format and bump the DB version accordingly.
Reported by: Ollivier Robert <roberto@eurocontrol.fr> and others
- Make sure to check DB version when checking for a DB. (-u)
pkg_fetch, pkgdb, portsclean, portsdb:
- Make them read the configuration file and add the -q/--no-config
option just as other tools.
pkgtools.conf:
- Add examples for PKG_FETCH, PACKAGEROOT and PACKAGESITE.
- Introduce new configuration variables: IGNORE_CATEGORIES and
EXTRA_CATEGORIES.
- Now one can set common environment variables in this file.
- Now all the tools read this file by default.
overall:
- Reword and enhance the manual pages and the usage instructions.
- Clarify the explanation about port/package dependency.
Submitted by: Jos Backus <josb@cncdsl.com> (Thanks a lot!)
portupgrade, portversion, pkg_deinstall, pkg_glob:
Introduce the configuration file `pkgtools.conf' for the pkgtools
suite. Currently portupgrade(1), portversion(1), pkg_deinstall(1)
and pkg_glob(1) use it.
Add a new option -q/--no-config to avoid reading the configuration
file.
Introduce dependency sanity checks. If an inconsistency is
detected, immediately abort a process suggesting the user should run
`pkgdb -F'.
Add a new option -O/--omit-check to omit the sanity checks.
pkgdb:
Fix the origin guesser's minor bug when the origin port directory
does not exist.
portupgrade:
Add a new option -y/--yes, which lets portupgrade assume user
answers yes to all the questions.
Make sure to ask user where it should. (Get -N and -i right)
Fix a bug where portinstall -i didn't work interactively.
[Submitted by: Gregory Neil Shapiro <gshapiro@gshapiro.net>]
portsdb:
Attempt to offer fast INDEX generation. Now it generates an INDEX
file more than a few minutes faster.
overall:
Emit a warning when pkgdb cannot be updated though it is not
up-to-date.
Finally remove -E/--regex. Use the ':RE' glob pattern instead.
portupgrade:
Fix a couple of bugs where `portupgrade -h | more' doesn't show the
usage nicely.
Reported by: Gregory Neil Shapiro <gshapiro@gshapiro.net>
portsclean:
Make good use of portsdb and pkgdb, use lazy (but virtually
sufficient) checks and get distclean much faster. (both -D and -DD)
portsdb:
Do not fail even if some categories are missing in the ports tree.
Reported and tested by: Jos Backus <josb@cncdsl.com>
And some other internal bugfixes that do not affect the user tools.
release again. (Hopefully)
portupgrade:
Delete an unneeded exception snatcher. This fixes a bug where ports
marked as IGNORE are not properly skipped.
pkgdb:
Rebuild pkgdb if it looks like broken, instead of resulting in
failure.
Unbreak regexp match. ;)
Fix a typo of a variable name in an error message.
[Reported by: Stijn Hoop <stijn@win.tue.nl>]
portsdb:
Allow a port dir to be a symlink to a directory.
Add support for the case where some catetories are missing in a
ports tree.
[Submitted by: Jos Backus <josb@cncdsl.com>]
pkgdb:
Fix a type error and unbreak date-based pkg glob.
[Reported by: Ollivier Robert <roberto@keltia.freenix.fr>]
portupgrade:
Silence pkg_info when it does not find a package.
[Submitted by: Jimmy Olgeni <olgeni@uli.it>]
pkgdb:
Make pkgdb fixer's dependency guesser aware of language specific
categories. Now it suggests ja-qt-2.3.1 instead of qt-1.45_1 for
a missing qt-2.3.1.
Do not fail over a package entry that has no +COMMENT file, like an
XFree86 package entry faked by the FreeBSD installer.
[Reported by: Ollivier Robert <roberto@eurocontrol.fr>]
portsdb:
Do not allow specifying a ports glob using a relative path to the
current directory. This prevents the tools from confusing when the
current directory is under PORTS_DIR.
Give a better progress display.
create a new database for the very first time. (Run "pkgdb -fu" if you have
20011006 installed)
Fix MAN1 entries.
Remove old database pkgdb.byfile.db on installation.
(new database is named pkgdb.db)
Beware, there can probably be some bugs left in the new pkgdb code.
In such a case please run "pkgdb -fu" to fix the situation.
overall:
Utilize pkgdb better and optimize.
portupgrade:
Do not scan packages all at once when -P is given, but find one per
request. [Requested by: Hideaki OKADA <hokada@isl.melco.co.jp>]
Create a backup package record in tmpdir instead of pkgdbdir, in
order not to update pkgdbdir's mtime.
Rephrase/add messages.
portinstall:
Commit major changes and improvements over portinstall.
Now it warns and asks the user for confirmation if a port is being
installed when a package of the same origin is already installed.
Upgrade all the required packages before installing a new port.
[Requested by: Hajimu UMEMOTO <ume@mahoroba.org>]
Do not install ports immediately but queue them up, then install
them after upgrading all the specified and required packages.
pkgdb:
Record pkgname <=> origin mapping in the pkgdb and make full use of
it instead of repeatedly calling pkg_info -o which has undesirable
overheads.
Make portupgrade, pkg_deinstall update the pkgdb after they
install/deinstall packages or modify package records.
Integrate pkgdepfix(1) within. (pkgdb -F/--fix)
Add -o/--origin, which looks up an origin of a package in pkgdb.
Detect added/deleted/reinstalled packages more strictly so pkgdb can
keep being sane and consistent.
Do not fail over a small conflict found during updating pkgdb;
portupgrade should not be aborted by such a minor problem.
pkgdepfix:
Obsolete -o and turn -o on by default. Now we can look up an origin
of a package so quickly thanks to the pkgdb.
Integrate into pkgdb(1). (pkgdb -F/--fix)
portsdb:
A date spec is not a valid ports glob; emit a warning if specified.
[Found by: Hajimu UMEMOTO <ume@mahoroba.org>]
and that on www.idaemons.org didn't match. They are identical in the
uncompressed form but I think I compressed one with -9 and another
with -6 somehow. Thereby list the MD5 checksums of both in pkg-plist
as a workaround.
The one on ftp.FreeBSD.org will be overwritten by the one on
www.idaemons.org soon.
Submitted by: John Merryweather Cooper <john_m_cooper@yahoo.com>
overall:
- Implement new package glob patterns: {<|<=|>=|>}{date|pkgname_glob}.
For example, you can rebuild and reinstall all the dependent
packages of png that had been installed prior to png this way:
portupgrade -fr png -x '>=png'
- Introduce the packages database, which is currently a simple hash
that maps file paths to package names.
Inspired by: NetBSD
pkgdb(1): (new)
- A command to create/update the packages database.
pkg_which(1): (new)
- A command to check which package a file came from quickly.
portsclean(1):
- Suppress "can't cd" error messages.
Submitted by: Jos Backus <josb@cncdsl.com>
ports_glob(1): (new)
- Give portsdb(1)'s ports glob expanding feature a suitable name.
overall:
- Work around pkg_info(1)'s "feature" where it gets info from
package files under the packages directory prior to installed
packages.
[Reported by: kuwa@flab.fujitsu.co.jp]
portupgrade:
- Skip ports that are marked as IGNORE.
[Requested by: Jimmy Olgeni <olgeni@uli.it>]
portsdb:
- Stop specifying -j# when making a ports index to avoid outputting
a buggy INDEX.
portsclean:
- Do not assume a distinfo file is in each port's own directory, and
use `make -V MD5_FILE', in order to support master-slave ports
properly.
Now that RUBY_ARCH has been changed not to include a minor version of
the system, you no longer need to have ruby_s in order to cope with
the migration problem with a minor system upgrade like 4.3 -> 4.4.
You will need ruby_static only if you are going to do a major upgrade
of your system like 4.3 -> 5.0.
portupgrade:
- Support the environment variable PKG_PATH which is supported by
pkg_add(1).
- Try to fetch the latest version of a package if the exactly
same version as the port is not found in the package site.
portupgrade:
- Add a new option -b/--backup-packages, which keeps backup packages
of the old versions'.
Requested by: mistral@imasy.or.jp (Yoshihiko SARUMARU)
pkg_fetch:
- Accept a full URI as well.
- Do not try both All and Latest; now one must put an @ at the end
of a package name if one wants to omit the version part.
("cyrus-sasl@", etc.)
- Obsolete the PACKAGEBASE environment variable.
- If -f is specified, remove existing packages if they are corrupt.
- Do not depend on the ports database.
portsclean:
- Fix the command line help regarding -D and -DD.
Submitted by: Jimmy Olgeni <olgeni@uli.it>
compdefs:
- Fix stupid bugs in the zsh compdefs.
overall:
- Allow using regexp in a glob pattern on the spot, by just putting a
regexp preceded by a colon (:).
- Thereby deprecate -E/--regex, which is evil because it changes the
global status. (but the option is still kept for compatibility)
portupgrade:
- Return correct exit status.
[Reported by: Trevin Chow <trevin@mail.com>]
portinstall:
- Read the env variable PORTUPGRADE. (it read PORTINSTALL previously)
pkg_fetch:
- Be even more smart in deciding which to try first, All or Latest.
overall:
- Fix a bug where a glob expansion did not work as expected when in a
port direcotry.
portupgrade:
- Fix a bug where a port not listed in the INDEX failed to install
because of a dependency error.
portversion:
- Work around a feature of the option parser library where -l= is
treated as -l "". Now portversion -l= and portversion -L= will work
just as pkg_version(8) does.
portupgrade:
- Make portupgrade download packages using pkg_fetch(1) when -FP is
given. (also aware of -R, -f and -v)
pkg_fetch:
- Be more intelligent and try All and Latest in the proper order.
- Return a non-zero status value when some are failed.
- Change -r to -R, since it recurses upwards through dependencies.
Overall:
- Fix glob expansion routines not to see a "No such package is
installed" error when it is actually installed. It was broken when
I hacked them to allow rather stupid patterns.
portupgrade:
- Add a new option -M/--make-env, with which a user can specify
arguments to prepend to each make(1) command line. Any wrapper or
environment variable assignment can be prepended.
- Make -RF invoke the `fetch-recursive' target.
Suggested by: Jimmy Olgeni <olgeni@uli.it>
- Optimize the combination of make(1) targets.
- Fix a bug where it causes an error when portsdb detects a DB_VERSION
bump.
Reported by: Ollivier Robert <roberto@eurocontrol.fr>
- Add a new option for portsdb(1): -f/--force, so you can force it to
update database regardless of timestamps.
- Bump DB_VERSION, since PkgVersion was changed a bit.
Reported by: fritz.heinrichmeyer@fernuni-hagen.de
- Add a missing argument for Dir::chdir_do().
Reported by: "Li-lun Wang (Leland Wang)" <llwang@infor.org>
pkg_fetch:
- Yet another new utility, which downloads binary packages from a
remote site. Optionally it can also download the required packages
recursively.
portupgrade:
- Do not regard "no package found" as an error when upgrading packages
with -PP.
pkgdepfix, portsclean -L:
- Fix a problem where the ruby 1.6.4 release fails in error.
Reported by: David Wolfskill <david@catwhisker.org>
portcvsweb:
- Explicitly execute a browser via /bin/sh, as the specification says.
other stuff:
- Improve the manpage and the completion definitions.
- Fix a stupid bug in the last version when generating a pkgname as a
string.
- Fix a stupid sorting bug in portsclean -L.
and some minor corrections that don't really matter.
Reported by: Jimmy Olgeni <olgeni@uli.it>
Overall:
- Support the cases where PORTREVISION == 0 and PORTEPOCH == 0.
portupgrade:
- Add a new option: -l/--log-results, which is used to save the
results as a file. [new]
- Properly exit with a non-zero status when main() does not return a
valid value. [fix]
portsclean:
- Finally implement portsclean -L/--libclean, which cleans the old
shared libraries that are not recorded in the package database. It
moves old and orphan shared libraries to ${PREFIX}/lib/compat/pkg.
[new]
- In concert with this feature, the portupgrade port now digs the
directory and installs a startup script which runs `ldconfig -m
${PREFIX}/lib/compat/pkg'. [new]
- Implement portsclean -DD, which cleans all the distfiles which are
not referenced from any port that is currently installed, which
portsclean -D only cleans the distfiles which are not referenced
from any port in the ports tree. [new]
Note that -DD runs very fast but removes more, whereas -D runs so
slow but removes less.
Idea provided by: Kris Kennaway <kris@obsecurity.org>
portsdb:
- Do not fail in the case when INDEX.db and INDEX.rbo are not in sync
somehow. [fix]
- Implement -r/--recursive. (a bit slow, though) [new]
Completions:
- Improve and update zsh compdefs.
- Add completion definitions for bash. [new]
..and several minor improvements. Enjoy!
Overall:
- Fix commands and libraries to run on a statically linked ruby.
- If a command fails to load dynamic modules with dynamically linked ruby,
restart itself with statically linked ruby.
This is to cope with the 4.3 -> 4.4 upgrade, since ruby's dynamic modules
are under a directory with a name including a FreeBSD version. Now
portupgrade can upgrade itself and ruby even after a major FreeBSD upgrade.
portupgrade:
- Fix a bug where portupgrade -sS 'su root -c %s' did not properly work.
[Submitted by: KOMATSU Shinichiro <koma2@jiro.c.u-tokyo.ac.jp>]
portcvsweb:
- Accept a port path as well.
portsdb:
- Adopt better error handling and a better PORTS_DBDIR selection algorithm.
- Check database errors a bit harder.
Overall:
* Fix a bug where when -R is set the specified package in the first
place is not included somehow.
[Reported by: Jose M. Alcaide <jose@we.lc.ehu.es>]
* Improve the path regularization routines.
portsdb:
* Update the INDEX first, and then the database even if the options
are specified in reversed order. (-uU)
portupgrade:
* Fix a misfeature where "skip" was regarded as failure, when checking
for dependent ports' build/install failure.
[Pointed out by: Jimmy Olgeni <olgeni@uli.it>, Jose M. Alcaide
<jose@we.lc.ehu.es>]
* Do not move a backup package under PKG_DBDIR, but just keep it under
TMPDIR. It should not require a large amount of free space in
PKG_DBDIR.
[Pointed out by: John Merryweather Cooper <jmcoopr@webmail.bmi.net>]
Overall:
- Disregard -r and -R when -a is also specified; there is no
need to recurse when you do with everything.
- Fix a bug in -r and -R routines where it might fail in error
if dependencies of a package were all wrong.
[Caught by: Mamoru Iwaki <iwaki@bsp.bc.niigata-u.ac.jp>]
- Fix the support for the case where PORTSDIR is a symlink.
- Squeeze //'s.
- Run `stty sane' before building each port, on interrupt, and
before exit.
- Accept portorigin_glob and pkgname_glob with paths, except
for -x/--exclude.
portupgrade:
- Fix `portupgrade -Ni' (portinstall -i).
- Skip ports and packages that once failed, and do not try to
upgrade multiple packages of an origin.
- If a port or a package failed to build or install, skip its
dependents. This behavior can be overridden by the new
option, -g/--go-on.
pkgdepfix:
- Backup +CONTENTS files before removing duplicates.
portsclean:
- Start implementing -L/--libclean.
misc/tcsh/complete.sample:
- Add a simple example of tcsh completion definitions.
[Contributed by: kuwa@flab.fujitsu.co.jp]
portupgrade:
Fix -A and -B which were broken when I introduced -L.
[Submitted by: Ollivier Robert <roberto@keltia.freenix.fr>]
portsdb:
Implement -M/--master-recursive, -R/--upward-recursive and
-x/--exclude, and sort the result in dependency order.
portsdb:
Do not clobber the INDEX file until the new one is ready,
so you don't have a zero-sized INDEX while updating it.
pkgdepfix:
Implement an origin duplicates fixer (only invoked when -o is
specified), and when it fixes one, redo the whole session.
Make the cyclic dependencies fixer a bit more intelligent.
portcvsweb:
Conform to the Secure BROWSER (SB) Specification proposed by
David A. Wheeler.
cf. http://www.dwheeler.com/browse/secure_browser.htmlhttp://www.tuxedo.org/~esr/BROWSER/
- Introduce a new environment variable "PORTS_INDEX", to specify an
alternate ports INDEX file location. You can set this to something
like `/usr/ports/INDEX.txt" not to overwrite the one under the
control of CVSup/CVS/CTM.
- Fix portinstall (portupgrade -N) which I broke recently by mistake.
- Change the report format. It shouldn't be too verbose unless you
specify -v.
