| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
|
|
reset later on) and had a typo. Add a dependency on gnupg instead to make
sure a pgp (gpg) binary is around to be used (provision should be made to
be able to use eg netpgp instead .. to be done later, hopefully after
someone else created the 'pick my favourite pgp version' infrastructure :> ).
|
|
Major changes from 2.5.0 to 2.5.1
* Fixed a segfault in imap_connection which could occur when SASL was
used.
* Fixed a segfault in the keyword generation code which was assuming
that an article was nul-terminated. Fixed another segfault in the
keyword generation code when an article already contained a Keywords:
header. Thanks to Nix for the bug reports.
* Owing to the US-CERT vulnerability note VU#238019, Cyrus SASL library
has slightly changed. imap_connection and nnrpd now handle that
change. Otherwise, some answers are too long to be properly computed
during SASL exchanges.
* Fixed a memory allocation problem which caused nnrpd to die when
retrieving via HDR/XHDR/XPAT the contents of an extra overview field
absent from the headers of an article. The NEWNEWS command was also
affected on very rare cases. Thanks to Tim Woodall for the bug
report.
* HDR/XHDR/XPAT answers are now robust when the overview database is
inconsistent. When the overview schema was modified without the
overview database being rebuilt, wrong results could be returned for
extra fields (especially a random portion of some other header). The
desired header name is now explicitly searched for in the overview
information.
* Fixed the source which is logged to the news log file for local
postings when the local server is not listed in incoming.conf. A
wrong name was used, taken amongst known peers. The source is now
logged as "localhost".
* Fixed a bug in the timecaf storage method: only the first 65535
articles could be retrievable in a CAF, though everything was properly
stored. (A Crunched Article File contains all the articles that
arrive to the news server during 256 seconds.)
The storage token now uses 4 bytes to store the article sequence
number for timecaf, instead of only 2 bytes. Thanks to Kamil Jonca
for the bug report and also the patch.
* Fixed a bug in both timecaf and timehash which prevented them from
working on systems where short ints were not 16-bit integers.
* When there is not enough space to write an entire CAF header, the
timecaf storage manager now uses a larger blocksize. On 32-bit
systems, the CAF header is about 300 bytes, leaving about 200 bytes
for the free bitmap index (the remaining of a 512-byte blocksize). On
64-bit systems, the size of the CAF header could exceed 512 bytes,
thus leaving no room for the free bitmap index. A 1 KB blocksize is
then used, or a larger size if need be.
* A new CNFS version has been introduced by Miquel van Smoorenburg in
the CNFS header. CNFSv4 uses 4 KB blocks instead of 512 bytes, which
more particularly makes writes faster. CNFSv4 supports
files/partitions up to 16 TB with a 4 KB blocksize.
Existing CNFS buffers are kept unchanged; only new CNFS buffers are
initialized with that new version.
* grephistory -l now returns the contents of the expires history field
as well as the hash of the message-ID. Besides, when the storage API
token does not exist, grephistory -v now also returns the hash of the
requested message-ID.
* The check on cancel messages when *verifycancels* is set to true in
inn.conf has been changed to verify that at least one newsgroup in the
cancel message can be found in the article to be cancelled. This new
feature is from Christopher Biedl.
The previous behaviour was to check whether the cancel message is from
the same person as the original post, which is extremely easy to
spoof; besides, RFC 5537 (USEPRO) mentions that "cancel control
messages are not required to contain From: and Sender: header fields
matching the target message. This requirement only encouraged cancel
issuers to conceal their identity and provided no security".
* The way the "/remember/" line in expire.ctl works has changed.
History retention for an article was done according to its original
arrival time; it is now according to its original posting date.
Otherwise, unnecessary data may be kept too long in the history file.
To achieve that, the HISremember() function in history API now expects
a fourth parameter: the article posting time.
Note that article expiration has not changed and is still based on
arrival time, unless the -p flag is passed to expire or expireover, in
which case posting time is used.
* The default value for "/remember/" has changed from 10 to 11 because
it should be one more than the *artcutoff* parameter in inn.conf, so
that articles posted one day into the future are properly retained in
history.
* auth_krb5 has been rewritten by Russ Allbery to use modern Kerberos
APIs. Note that using ckpasswd with PAM support and a Kerberos PAM
module instead of this authenticator is still recommended.
* A new -L flag has been added by Jonathan Kamens to makehistory so as
to specify a load average limit. If the system load average exceeds
the specified limit, makehistory sleeps until it goes below the limit.
* As UTF-8 is the default character set in RFC 3977, "ctlinnd pause",
"ctlinnd readers", "ctlinnd reject", "ctlinnd reserve", "ctlinnd
throttle" and "nnrpd -r" commands now require the given reason to be
encoded in UTF-8, so that it can be properly sent to news readers.
The creator's name given to "ctlinnd newgroup" is also expected to be
encoded in UTF-8.
* The output of consistency checks for article storage and the history
file no longer appears by default when "cnfsstat -a" is used. A new
-v flag has been added to cnfsstat so as to see it.
* The default path for TLS certificates has changed from *pathnews*/lib
to *pathetc*. It only affects new INN installations or generations of
certificates with "make cert". Besides, a default value has been
added to *tlscapath* because it is required by nnrpd when TLS is used.
* gzip(1) is now the default UUCP batcher in send-uucp instead of
compress(1) because gzip is more widely available than compress, due
to old patent issues. Note that there is no impact on decompression
as it is handled by rnews.
* cnfsheadconf now uses the Perl core module "Math::BigInt" rather than
the deprecated bigint.pl library. When used without specifying a CNFS
buffer, it now properly displays the status of all CNFS buffers.
|
|
(reported by Geoff Wing <gcw@pobox.com>)
|
|
|
|
|
|
|
|
This changes the buildlink3.mk files to use an include guard for the
recursive include. The use of BUILDLINK_DEPTH, BUILDLINK_DEPENDS,
BUILDLINK_PACKAGES and BUILDLINK_ORDER is handled by a single new
variable BUILDLINK_TREE. Each buildlink3.mk file adds a pair of
enter/exit marker, which can be used to reconstruct the tree and
to determine first level includes. Avoiding := for large variables
(BUILDLINK_ORDER) speeds up parse time as += has linear complexity.
The include guard reduces system time by avoiding reading files over and
over again. For complex packages this reduces both %user and %sys time to
half of the former time.
|
|
instead of install -l. Plus, fix a nit in the startup script.
|
|
|
|
|
|
to trigger/signal a rebuild for the transition 5.8.8 -> 5.10.0.
The list of packages is computed by finding all packages which end
up having either of PERL5_USE_PACKLIST, BUILDLINK_API_DEPENDS.perl,
or PERL5_PACKLIST defined in their make setup (tested via
"make show-vars VARNAMES=...").
|
|
|
|
|
|
|
|
expect that which we installed (ie, fix PLIST)
also make sure that our installation destination is under PREFIX
fixes PR 39165
|
|
|
|
- major change of directory structure
- two new options (uucp and perl now both optional)
- change of maintainer
|
|
on packages that are affected by the switch from the openssl 0.9.7
branch to the 0.9.8 branch. ok jlam@
|
|
|
|
and to support the "inet6" option instead.
Remaining usage of USE_INET6 was solely for the benefit of the scripts
that generate the README.html files. Replace:
BUILD_DEFS+= USE_INET6
with
BUILD_DEFS+= IPV6_READY
and teach the README-generation tools to look for that instead.
This nukes USE_INET6 from pkgsrc proper. We leave a tiny bit of code
to continue to support USE_INET6 for pkgsrc-wip until it has been nuked
from there as well.
|
|
the owner of all installed files is a non-root user. This change
affects most packages that require special users or groups by making
them use the specified unprivileged user and group instead.
(1) Add two new variables PKG_GROUPS_VARS and PKG_USERS_VARS to
unprivileged.mk. These two variables are lists of other bmake
variables that define package-specific users and groups. Packages
that have user-settable variables for users and groups, e.g. apache
and APACHE_{USER,GROUP}, courier-mta and COURIER_{USER,GROUP},
etc., should list these variables in PKG_USERS_VARS and PKG_GROUPS_VARS
so that unprivileged.mk can know to set them to ${UNPRIVILEGED_USER}
and ${UNPRIVILEGED_GROUP}.
(2) Modify packages to use PKG_GROUPS_VARS and PKG_USERS_VARS.
|
|
Patch provided by Sergey Svishchev in private mail.
|
|
|
|
INN on my home server.
|
|
and add a new helper target and script, "show-buildlink3", that outputs
a listing of the buildlink3.mk files included as well as the depth at
which they are included.
For example, "make show-buildlink3" in fonts/Xft2 displays:
zlib
fontconfig
iconv
zlib
freetype2
expat
freetype2
Xrender
renderproto
|
|
of the order in which buildlink3.mk files are (recursively) included
by a package Makefile.
|
|
syntax as specified in pkgsrc/mk/install/bsd.pkginstall.mk:1.47.
|
|
that they look nicer.
|
|
|
|
RECOMMENDED is removed. It becomes ABI_DEPENDS.
BUILDLINK_RECOMMENDED.foo becomes BUILDLINK_ABI_DEPENDS.foo.
BUILDLINK_DEPENDS.foo becomes BUILDLINK_API_DEPENDS.foo.
BUILDLINK_DEPENDS does not change.
IGNORE_RECOMMENDED (which defaulted to "no") becomes USE_ABI_DEPENDS
which defaults to "yes".
Added to obsolete.mk checking for IGNORE_RECOMMENDED.
I did not manually go through and fix any aesthetic tab/spacing issues.
I have tested the above patch on DragonFly building and packaging
subversion and pkglint and their many dependencies.
I have also tested USE_ABI_DEPENDS=no on my NetBSD workstation (where I
have used IGNORE_RECOMMENDED for a long time). I have been an active user
of IGNORE_RECOMMENDED since it was available.
As suggested, I removed the documentation sentences suggesting bumping for
"security" issues.
As discussed on tech-pkg.
I will commit to revbump, pkglint, pkg_install, createbuildlink separately.
Note that if you use wip, it will fail! I will commit to pkgsrc-wip
later (within day).
|
|
- Previous versions of INN had an optimization for handling XHDR
Newsgroups that used the Xref header from overview. While this does
make the command much faster, it doesn't produce accurate results and
breaks the NNTP protocol, so this optimization has been removed.
- Fixed a bug in innd that allowed it to accept articles with duplicated
headers if the header occurred an odd number of times. Modified the
programs for rebuilding overview to use the last Xref header if there
are multiple ones to avoid problems with spools that contain such
invalid articles.
- Fixed yet another problem with verifying that a user has permissions
to approve posts to a moderated group. Thanks, Jens Schlegel.
- Increase the send and receive buffer on the Unix domain socket used by
ctlinnd. This should allow longer replies (particularly for innstat)
on platforms with very low default Unix domain socket buffer sizes.
- rnews's handling of articles with nul characters, NNTP errors, header
problems, and deferrals has been significantly improved.
- Thomas Parmelan added support to send-uucp for specifying the funnel
or exploder site to flush for feeds managed through one and fixed a
problem with picking up old stranded work files.
- INN is now licensed under a less restrictive license (about as
minimally restrictive as possible shy of public domain), and the
clause similar to the old BSD advertising clause has been dropped.
- make install and make update now always install the newly built
binaries, rather than only installing them if the modification times
are newer. This is the behavior that people expect. make install now
also automatically builds a new (empty) history database if one
doesn't already exist.
- The embedded Tcl filter code has been disabled (and will be removed
entirely in the next major release of INN). It hasn't worked for some
time and causes innd crashes if compiled in (even if not used). If
someone wants to step forward and maintain it, I recommend starting
from scratch and emulating the Perl and Python filters.
- ctlinnd should now successfully handle messages from INN up to the
maximum allowable packet size in the protocol, fixing problems sites
with many active peers were having with innstat output.
- Overview generation has been fixed in both makehistory and innd to
follow the rules in the latest NNTP draft rather than just replacing
special characters with spaces. This means that the unfolding of
folded header lines will not introduce additional, incorrect
whitespace in the overview data.
- nnrpd now uniformly responds with a 480 or 502 status code to attempts
to read a newsgroup to which the user does not have access, depending
on whether the user has authenticated. Previously, it returned a 411
status code, claiming the group didn't exist, which confuses the
reactive authentication capability of news readers.
- If a user is not authorized to approve articles (using the A access
control in readers.conf), articles that include Approved headers will
be rejected even if posted to unmoderated groups. Some other site may
consider that group to be moderated.
- The configuration parser used for readers.conf and others now
correctly handles "#" inside quoted strings and is more robust against
unmatched double quotes.
- Messages mailed to moderators had two spaces after the colons in the
headers, rather than one. This bug has been fixed.
- A bug that could cause heap corruption and random crashes in innd if
INN were compiled with Python support has been fixed.
- Some problems with innd's tracking of article size and enforcement of
the configured maximum article size have been fixed.
- pgpverify will now correctly verify signatures generated by GnuPG and
better supports GnuPG as the PGP implementation.
- INN's code should now be more 64-bit clean in its handling of size_t,
pointer differences, and casting of pointers, correcting problems that
showed up on 64-bit platforms like AMD64.
- Improved the error reporting in the history database code, in inews,
in controlchan, and in expire.
- Many other more minor bug fixes, optimization improvements, and
documentation fixes.
|
|
|
|
in PR pkg/32299.
|
|
automatically detects whether we want the pkginstall machinery to be
used by the package Makefile.
|
|
Okayed by maintainer.
|
|
CONFIGURE_ARGS.
|
|
example MAKE_ENV+=FOO=${BAR} is changed to MAKE_ENV+=FOO=${BAR:Q}. Some
other changes are outlined in
http://mail-index.netbsd.org/tech-pkg/2005/12/02/0034.html
|
|
"radius.conf.5".
|
|
|
|
for GNU_CONFIGURE automatically.
(This has been in my personal pkgsrc for a while.)
|
|
of this.
|
|
backslashes anymore. A single backslash is enough. Changed the
definition in all affected packages. For those that are not caught, an
additional check is placed into bsd.pkginstall.mk.
|
|
of Perl files to deal with the perl-5.8.7 update that moved all
pkgsrc-installed Perl files into the "vendor" directories.
|
|
perl5/buildlink3.mk.
|
|
arguments anymore.
|
|
from including perl5/buildlink3.mk. These packages just need the Perl
interpreter, and can just add "perl" to USE_TOOLS instead.
|
|
make does not understand the "W" option in :C modifiers, and it
handles substitution in strings containing spaces somwheat differently.
Bump PKGREVISION.
|
|
Bump package revision once more.
|
|
variable name that can change between autoconf versions, just set
VARNAME, which should always be correct.
Also, note that inn needs yacc to build, and uses gzip in the installed
scripts. Bump the PKGREVISION to 4.
|
