overall:
- Use sudo in installing (copying) a file only if once it failed as
the current user.
- Speed up string substitution.
portupgrade(1):
- Use sudo as necessary in saving log files also.
pkgdb(1):
- Look up ports/MOVED to fix stale origins (both ORIGIN and
DEPORIGIN).
- Skip the check for origin duplicates when in automatic mode. (This
speeds up autofix)
- When pointing out a missing origin, show the name of the package in
question.
Suggested by: Genesys on EFnet
portupgrade(1):
- Check return status of hardly failing commands also.
pkgdb(1):
- Add -Q/--quiet and -QQ for internal use.
portsclean(1):
- Ditto.
overall:
- Reduce /bin/sh process calls by replacing shell redirection to
/dev/null with pkgdb -Q/portsclean -Q/fetch -q. (I know I can do
fork, dup and exec if necessary)
- Eliminate zombies during running. Formerly waitpid's against
terminated processes are delayed until Ruby gets short of resource
and thus runs a GC, but now portupgrade(1), pkgdb(1) and
portsclean(1) explicitly kick a waitpid immediately after each
popen() is done.
PR: ports/38075
Submitted by: Chris J. Mutter <cjm@s2.enemy.org>
Reported by: several others
overall:
- Allow a regular expression or glob pattern against origins as
pkgname glob pattern. For example, the following commands work as
expected now:
pkg_glob 'lang/*'
portupgrade lang/ruby
pkg_info ':(japanese|chinese|korean)/'
- Use install(1) instead of cp(1) and chmod(1) to install modified
data files such as +CONTENTS.
pkgdb(1):
- Add a new option --autofix, which is a shorthand of --auto --fix
(-aF).
portupgrade(1):
- Allow specifying a directory after -L.
- Auto-fix dependencies before creating a package.
PR: ports/41383
Submitted by: Vivek Khera <khera@kciLink.com>
pkgtools.conf(5):
- Add a new configuration variable ALT_PKGDEP, which defines
alternative dependencies, used by pkgdb -F/-aF. This makes yet
another step forward to eliminate the need for manual pkgdb -F's.
overall:
- Make these work with ruby 1.7.
portupgrade(1):
- Get portupgrade -s to work again.
- Synchronize the failure reason guesser with the latest "processlogs"
script.
- Rename -g/--go-on to -k/--keep-going, on the (vague) analogy of
make(1).
possibly be a couple of bugs. Make sure to do your backups as always,
and have fun. :)
overall, pkgdb(3):
- The DEPORIGIN information is fully supported. Most stale
dependencies are automatically fixed based on that.
Hinted by: sobomax
- Disregard a package if it has a file named "+IGNOREME" in the
package directory.
PR: ports/41237
Hinted by: Jerry Murdock <j@jerrymurdock.com>, and some others
pkgdb(1):
- Add a new option -a/auto, used with -F/--fix, which turns on
automatic mode, in which pkgdb(1) only fixes discrepancies that can
be fixed 100% obviously and securely, and leaves the others.
- Make pkgdb(1) less verbose when -v is not specified.
portupgrade(1):
- Time the whole session also.
Requested by: nork
- Fix a bug in shellwords() where it wrongly interprets \ in '...' as
meta-character. This bug or maybe feature was inherited from Perl's
shellwords.pl / Text::ParseWords.
Reported by: Gabriel Ambuehl <gabriel_ambuehl@buz.ch>
- The DEPORIGIN information is fully supported. It fixes DEPORIGIN
when a package is upgraded with a different port using -o/--origin,
and "pkgdb -aF" is automatically called at closing. This makes
portupgrade a tad slower but greatly reduces the "pkgdb -F"
annoyance.
- Fix a race where it rarely failed in updating the pkgdb after
installation and the user had to run "pkgdb -fu".
- Do not redirect stderr to /dev/null when calling portsclean -L. Now
the pkgdb updater's indicator is displayed properly.
- Synchronize the build failure checker with the latest version of
ports/Tools/portbuild/scripts/processlogs.
- Do not bother with +REQUIRED_BY files when they are mechanically
generated from @pkgdep information by pkgdb(1).
pkg_fetch(1):
- Try .tgz first on FreeBSD 4.x and prior. Although .tbz is smaller,
there are few sites that prepare .tbz packages for FreeBSD 4.x.
pkgtools.conf(5)
- Prepare a new constant OS_PKGBRANCH, which is set to a
release/branch name that is suitable for specifying a FTP packages
directory, and make pkg_site_mirror() use it.
PR: ports/43265
Submitted by: Andriy Gapon <avg@icyb.net.ua>
portsclean(1):
- Suggest using libchk(1) to help manually clean out lib/compat/pkg.
- Improve messages. Use the word "shadow" to say that a library hides
another one with the same name.
- Fix a bug where portupgrade creates packages in a wrong directory.
This was introduced in 20020921.
Submitted by: "Dmitry A. Panov" <redfox@Fluffy.Khv.RU>
- Make portcvsweb <pkgname> work when it is an obsoleted package.
- Update documents.
- Respect PKG_SUFX, which could be defined in /etc/make.conf or as an
environment variable.
- Drop the pkg_tarup(1) dependency in favor of pkg_create(1) -b, and
mark IGNORE for systems which pkg_create(1) does not support -b.
On those systems such as 4.4-STABLE and prior, users can either
a) update pkg_* tools or b) stick with portupgrade-20020920 +
pkg_tarup-1.2_3.
- Remove the workaround against somewhat old 5-CURRENT's buggy
bsd.*.mk.
- Improve zsh/bash/tcsh completion definitions. (not much for tcsh,
though..)
portupgrade(1):
- Time each build, install, upgrade, etc. and show times
when -v/--verbose is specified.
[Requested by: people on freenode:#freebsd]
pkgdb(1):
- Extend pkgdb(1)'s -o so it can also modify a port's origin.
[Requested by: dcs]
- Raise an error when an empty pkgname is given with -s.
pkgtools.conf(5):
- Make MAKE_ARGS accept a proc.
[Suggested by: alane]
miscellaneous:
- Now all the tools including portupgrade(1) should work on
lang/ruby_static as well as lang/ruby_static-devel again.
- Completion definitions for zsh and bash are even slightly
improved.
- Fix the regexp for uname output so it can parse such as
"5.0-CURRENT-YYYYMMDD-JPSNAP".
Spotted by: will@FreeBSD.org
- portupgrade(1): Add more command failure checks in case sudo fails.
- portsclean(1): Add a cross reference to libchk(1) to the manpage.
portupgrade(1):
- Add more command failure checks in case sudo fails.
- Always build backup packages as root.
Spotted by: Stijn Hoop <stijn@win.tue.nl>
- Let NO_IGNORE negate IGNORE.
FreeBSD PR: ports/41467
Suggested by: Alan Eldridge <ports@geeksrus.net>
portsclean(1):
- Do not mix stderr output from make(1) with stdout.
Submitted by: Jos Backus <jos@catnook.com>
pkgdb(1):
- Do not fail in error when a non-installed package is specified
with -o.
pkg_fetch(1):
- Support bzip2'd packages (*.tbz).
- Slightly fix messages.
- Introduce OS_PATCHLEVEL and exclude `-p#' from OS_RELEASE.
FreeBSD PR: ports/40375
Submitted by: Andriy Gapon <agapon@excite.com> (partly)
portsclean(1):
- -P/--pkgclean: Sweep away bzip2'd packages (*.tbz) too.
portupgrade(1):
- Support bzip2'd packages (*.tbz).
pkgtools.conf(5):
- Provide a couple of new utility functions, deorigin(origin) and
rc_file(origin_or_pkgname).
- Fix an example and provide some more examples.
- Introduce OS_PATCHLEVEL and exclude `-p#' from OS_RELEASE.
misc.:
- zsh compdefs: Look for *.tbz too.
- Avoid ruby 1.7 warnings.
- Show better error messages on command failure.
- pkg_deinstall(1): Unbreak -c by removing a leftover variable
reference.
Reported by: Tod McQuillin <devin@spamcop.net>
- pkg_deinstall(1): (manpage): Clarify that -c/--collate is disabled
by -f/--force.
Prompted by: parv <parv@pair.com>
- pkg_sort(1): Warn of each package entry that's not installed.
- pkg_sort(1): Do not print an empty line if no valid entry is given.
- pkgdb(1): (manpage) Fix typos, improve phrases and remove an
incomplete paragraph.
FreeBSD PR: ports/39183
Submitted by: Chris Pepper <pepper@rockefeller.edu>
- portsclean(1): (manpage) Fix a typo.
Submitted by: Martijn Koster <mak@greenhills.co.uk>
- Make portsdb -U respect PORTS_DBDIR even when it is different than
PORTS_DIR.
Reported by: Shigeto TAJIMA <tajima@ics.es.osaka-u.ac.jp>
- Do not override the user-defined value of PKG_PATH.
Reported by: The Anarcat <anarcat@anarcat.dyndns.org>
2002-04-05 21:08 knu
* lib/pkginfo.rb: Show the full pkgname as exception message when
the version part is invalid.
2002-04-05 21:06 knu
* bin/portupgrade: Fix a flaw where dependent ports' build/install
error was not properly detected when -s is specified.
Reported by: Stijn Hoop <stijn@win.tue.nl>
20020327 or later.
portsclean(1):
- Run mv(1) with -f unless -i is given.
This fixes:
PR: ports/36407
Submitted by: Albert Kinderman <albert.kinderman@csun.edu>
If you see portupgrade(1) stop at 'Cleaning out obsolete shared
libraries', hit Ctrl-C (just once) to abort portsclean(1).
portupgrade(1):
- Fix portsclean -L invocation.
- Fix some messages.
- Update the manpage.
o Add another example for -B that runs 'cvs update' for the master
port(s) of a slave port, using 'ports_glob -M'.
Question by: Ollivier Robert <roberto@FreeBSD.org>
o Mention that portupgrade(1) now runs `portsclean -L' at the end of
each upgrade.
o Fix wording and update the descriptions about portsclean -L.
portsclean(1):
- Fix wording and update the descriptions about -L.
portsclean(1):
- Improve -L/--libclean. Ability to delete duplicate libraries is
added.
portupgrade(1):
- Call portsclean -L at the end of each upgrade.
pkg_deinstall(1):
- Show the results more precisely.
- -P: Only preserve real FreeBSD shared libraries, determined by
file(1).
- -P: Move preserved shared libraries to $LOCALBASE/lib/compat/pkg.
portupgrade(1):
- If an error is returned from getcwd(), show a better message and
abort.
FreeBSD PR: ports/35893
Submitted by: Ernst de Haan <znerd@FreeBSD.org>
- -N: If multiple ports match a given pattern, list them.
Suggested by: Will Andrews <will@FreeBSD.org>
portsdb(1):
- On exception, handle it and exit normally.
- Check the file/directory permissions on the INDEX file before
generating a new INDEX.
Suggested by: parv <parv_@yahoo.com>
- Make automatic portsdb directory selection in the case of "portsdb
-fu" too.
ports_glob(1):
- -M: Unbreak.
portsdb(3):
- Fix the EXTRA_CATEGORIES support that was broken in recent updates.
PR: ports/35319
Submitted by: Ryan Davis <ryand-bsd@zenspider.com>
pkgtools.conf(5):
- Add a note that a user must run portsdb -Ufu to reflect values of
EXTRA_CATEGORIES and IGNORE_CATEGORIES.
For the victims of the above bug, please run "portsdb -qu" and then
"portupgrade -q portupgrade".
20020225.
- Clarify that portupgrade -O is disregarded if -r or -R is given.
- Reflect the fact that INDEX.rbo no longer exists.
- Document portsdb(1) better regarding PORTS_DBDIR.
- Merge INDEX.rbo into INDEX.db, and remove INDEX.rbo on portsdb
rebuild if it exists.
- Reduce the size of INDEX.db by half. :)
- Introduce PORTS_DBDRIVER and allow specifying a db driver for the
portsdb (INDEX.db), and switch the default portsdb driver from
db1.85 hash to db1.85 btree.
- Fix a bug where it referenced the obsoleted PortsDB::PORTS_DIR.
- Depend on ruby-bdb1 and make bdb1_btree (db 1.85 btree) the default
pkgdb driver/format. This probably fixes the problem where pkgdb is
unexpectedly broken and rebuilt from scratch.
- Support pkgdb driver fallback. (bdb -> bdb1 -> dbm)
- Show the pkgdb driver (format) on updating/rebuilding.
- Reflect environment variables defined in pkgtools.conf properly.
(Now ENV['PORTS_DBDIR'] is reflected properly)
Reported by: Mori Kouji <mori@tri.asanuma.co.jp>
pkg_sort(1), portupgrade(1):
- Sort packages by build-time dependencies in addition to run-time
dependencies. This slightly slows sorting down but offers faultless
upgrading.
Inspired by: ade
portcvsweb(1):
- Accept a pkgname_glob, a portorigin_glob, and a cvs module alias.
portcvsweb bash-1\*
portcvsweb portupgrade
portcvsweb \*/fd
protcvsweb sys/i386/conf
etc. etc...
- Suppress ident(1)'s error message.
- Add -h, -q and -v.
in the last version. For the victims please manually reinstall the
portupgrade port, as always. ;)
Reported by: many users (Sorry!)
Use some logic and keep pkg_add/pkg_delete from complaining about the
@exec/@unexec return codes.
Submitted by: olgeni
pkgdb(1):
- Set the default answer to no when a score is under 80%.
pkgdb(1), portupgrade(1):
- Add a BUGS entry to each of the two man pages to note that a user
must run pkgdb -fu to rebuild a corrupt pkgdb.
pkgdb(3):
- Add a silly workaround against a possible bug of the dbm module to
keep a pkgdb from getting insane and rebuilt.
- [exp.] Add a knob environment variable PKG_DBDRIVER to specify an
alternative binary database format for pkgdb. The available options
are:
"bdb_btree":
DB 3.x/4.x B+tree; requires databases/ruby-bdb.
"bdb_hash" or just "bdb":
DB 3.x/4.x hash; requires databases/ruby-bdb.
"bdb1_btree" or just "btree":
DB 1.85 B+tree; requires databases/ruby-bdb1.
"bdb1_hash", or just "bdb1" or "hash":
DB 1.85 hash; requires databases/ruby-bdb1. This should be
equivalent to the default except some tuning parameters, but
it is quite possible that ruby-bdb1 is stabler than the
standard dbm module included in ruby.
default:
DB 1.85 hash; uses ruby's standard dbm module.
In principle, a B+tree database is faster to store data, a little
bit slower to extract data, and takes more space than a hash
database.
This is still an experimental/undocumented feature only for those
interested, although I don't believe it brings any unstability.
Note that you can always run pkgdb -fu to rebuild a database.
portcvsweb(1):
- Invoke browsers with system() instead of exec() so that it will
properly try every browser in turn - exec('/bin/sh', '-c', '..') may
cause an error within /bin/sh but it never returns.
Reported by: Jimmy Olgeni <olgeni@FreeBSD.org>
portsclean(1):
- Make portsclean -D remove empty directories (recursively) as well.
pkg_deinstall(1), portupgrade(1):
- Properly print the legend.
pkg_glob(1), portsdb(1):
- Do not print an empty line when the result list is empty.
pkgdb(1):
- Greatly improve the origin fixing procedure. Do the origin
duplicateness check before checking dependencies and avoid redoing
the origin check.
- Show the user CVS history on the spot via the CVSweb.
portcvsweb(1):
- Allow specifying a no longer existent port file/directory.
"portcvsweb net/ruby-uri" now works.
- Change the default BROWSER value from 'lynx' to
'w3m:lynx:links:mozilla:netscape'.
portupgrade(1):
- Fix a bug where portupgrade tried to use a package with a different
origin when the package name is identical to that of what it wants
except for the version part.
pkgdb(3):
- Abolish the [You have no permission to update the pkgdb -- using a
slow method] message. Instead, run 'pkgdb -u' via sudo(8) when
under portupgrade -s, or stop the process immediately with a
message that the user must run the command as root.
pkgtools.conf(5):
- Allow specifying unofficial ports' origin globs in HOLD_PKGS as
well as package globs.
overall/miscellaneous:
- Change --no-config to --noconfig in order to get the -q option
really working. Since --no-config was regarded a negative form of
--config, it didn't work as expected.
- Fix some bugs in the zsh compdefs and add missing descriptions for
some options.
o Use bzip2 instead of gzip for the distribution tarball to save 20KB.
As the portupgrade user base is growing significantly, I'll get more
serious about the download size issue from now on. The next few
releases will be distributed in patches against this version.
pkg_which database. (The first stage was 'portsclean -L'.)
Hopefully these tools will support divided/obsoleted/renamed ports
some way in the not-too-distant future..
pkgdb(1):
- Make pkgdb(1) -F provide a user an option to unregister stale
packages, in addition to deinstalling them.
- Reword: 'remove a package record' -> 'unregister a package', etc. .
pkg_deinstall(1):
- Add a new option:
-c, --collate
Check if any of the installed files of each package have been
overwritten by other packages, and do not deinstall if any.
pkg_which(1):
- Make pkg_which(1) capable of showing multiple file owner packages.
- Add a new option:
-c/--collate PKGNAME
Show files installed by the given packge that have been
overwritten by other packages.
- Strip -c/-o's argument of PKG_DBDIR and a trailing slash.
- Be less verbose unless -v is given.
pkgdb(3):
- Record and handle files that are installed by multiple packages.
- Bump the DB version.
- Squeeze series of slashes seen in the output of 'pkg_info -L'.
- Show the reason for rebuilding a package database.
