head 1.8; access; symbols pkgsrc-2013Q2:1.8.0.22 pkgsrc-2013Q2-base:1.8 pkgsrc-2012Q4:1.8.0.20 pkgsrc-2012Q4-base:1.8 pkgsrc-2011Q4:1.8.0.18 pkgsrc-2011Q4-base:1.8 pkgsrc-2011Q2:1.8.0.16 pkgsrc-2011Q2-base:1.8 pkgsrc-2009Q4:1.8.0.14 pkgsrc-2009Q4-base:1.8 pkgsrc-2008Q4:1.8.0.12 pkgsrc-2008Q4-base:1.8 pkgsrc-2008Q3:1.8.0.10 pkgsrc-2008Q3-base:1.8 cube-native-xorg:1.8.0.8 cube-native-xorg-base:1.8 pkgsrc-2008Q2:1.8.0.6 pkgsrc-2008Q2-base:1.8 pkgsrc-2008Q1:1.8.0.4 pkgsrc-2008Q1-base:1.8 pkgsrc-2007Q4:1.8.0.2 pkgsrc-2007Q4-base:1.8 pkgsrc-2007Q3:1.6.0.6 pkgsrc-2007Q3-base:1.6 pkgsrc-2007Q2:1.6.0.4 pkgsrc-2007Q2-base:1.6 pkgsrc-2007Q1:1.6.0.2 pkgsrc-2007Q1-base:1.6 pkgsrc-2006Q4:1.5.0.2 pkgsrc-2006Q4-base:1.5 pkgsrc-2006Q3:1.3.0.2 pkgsrc-2006Q3-base:1.3 pkgsrc-2006Q2:1.2.0.14 pkgsrc-2006Q2-base:1.2 pkgsrc-2006Q1:1.2.0.12 pkgsrc-2006Q1-base:1.2 pkgsrc-2005Q4:1.2.0.10 pkgsrc-2005Q4-base:1.2 pkgsrc-2005Q3:1.2.0.8 pkgsrc-2005Q3-base:1.2 pkgsrc-2005Q2:1.2.0.6 pkgsrc-2005Q2-base:1.2 pkgsrc-2005Q1:1.2.0.4 pkgsrc-2005Q1-base:1.2 pkgsrc-2004Q4:1.2.0.2 pkgsrc-2004Q4-base:1.2 pkgsrc-2004Q3:1.1.1.1.0.2 pkgsrc-2004Q3-base:1.1.1.1 pkgsrc-base:1.1.1.1 TNF:1.1.1; locks; strict; comment @# @; 1.8 date 2007.12.14.08.58.02; author bjs; state dead; branches; next 1.7; 1.7 date 2007.10.26.14.08.25; author jmmv; state Exp; branches; next 1.6; 1.6 date 2007.01.02.11.09.57; author jmmv; state dead; branches; next 1.5; 1.5 date 2006.12.11.08.13.13; author dan; state Exp; branches; next 1.4; 1.4 date 2006.11.23.17.38.35; author jmmv; state dead; branches; next 1.3; 1.3 date 2006.09.22.11.14.06; author dan; state Exp; branches; next 1.2; 1.2 date 2004.11.28.10.35.45; author jmmv; state dead; branches; next 1.1; 1.1 date 2004.08.15.14.38.38; author drochner; state Exp; branches 1.1.1.1; next ; 1.1.1.1 date 2004.08.15.14.38.38; author drochner; state Exp; branches; next ; desc @@ 1.8 log @ Update to version 0.38. Fix one line of post-install target to allow for DESTDIR support. Also, patch-ab is no longer necessary; patch-aa is merely a one-liner now. Changes since 0.37: Changes - mtn log now prints a single dot for a project's root directory instead of an empty string. - mtn now warns if changes to a file will be ignored because the file has been deleted on one side of a merge. - mtn now errors if your chosen private key doesn't match the public key of the same name in your database. - mtn now checks for your key before a merge action takes place to ensure that any manually merged file isn't lost in an error case Bugs fixed - a bug introduced in 0.37 prevented an external merger from being executed unless the MTN_MERGE environment variable was set - mtn read successfully reads revision data, and cert packets again - mtn consistently supports certs with empty values (fixed 'ls certs' and 'read') Internal - Update Botan to 1.7.2. - Moved the gzip implementation out of the Botan directory. Other - Added the scripts of the following Lua-based contributed Monotone extension commands to contrib/command/: "mtn base", "mtn fuse", "mtn revision", "mtn conflicts". - Added a hooks version of the contributed ciabot script, contrib/ciabot_monotone_hookversion.lua - The monotone manual is now licensed under the GPL rather than the GFDL. @ text @$NetBSD: patch-ab,v 1.7 2007/10/26 14:08:25 jmmv Exp $ --- pcrewrap.cc.orig 2007-10-26 00:40:22.000000000 +0200 +++ pcrewrap.cc @@@@ -15,7 +15,6 @@@@ // This dirty trick is necessary to prevent the 'pcre' typedef defined by // pcre.h from colliding with namespace pcre. #define pcre pcre_t -#include "pcre_config.h" #include "pcre.h" #undef pcre @ 1.7 log @Update to 0.37: Changes - mtn db kill_rev_locally now checks for an existing workspace before the revision is killed and tries to apply the changes of this particular revision back to the workspace to allow easy re-committing afterwards - the "--brief" switch for mtn annotate has been renamed to "--revs-only" for clarity - mtn help now lists the commands (and their aliases) available within a group, so its easier to get an overview which commands are available at all - the "MTN_MERGE=diffutils" merger (provided by std_hooks.lua) was improved. It now accepts a MTN_MERGE_DIFFUTILS environment variable which can be used to control its behaviour through comma-separated "key[=value]" entries. Currently supported entries are "partial" for doing a partial batch/non-modal 3-way merge conflict "resolution" which uses embedded content conflict markers and "diff3opts=[...]" and "sdiffopts=[...]" for passing arbitrary options to the used "diff3" and "sdiff" tools. When used in combination with "mtn merge_into_workspace" this way one especially can achieve a CVS/SVN style non-modal workspace-based merging. - There is a new revision selector: "p:REV" selects the parent(s) of revision REV. For example, if a revision has one parent, mtn diff -r p:REV -r REV will show the changes made in that revision. - Monotone now uses the Perl-Compatible Regular Expression (PCRE) library for all regular expressions, instead of the boost::regex library. This means that external Boost libraries are no longer required to build or use Monotone. If building from source, you will still need the Boost headers available somewhere. See INSTALL for details. PCRE's syntax for regular expressions is a superset of boost::regex's syntax; it is unlikely that any existing .mtn-ignore files or other user uses of regexps will break. The manual now contains detailed documentation of the regexp syntax, borrowed from PCRE itself. - the format of "mtn automate inventory" has changed to basic_io. This fixes a couple of corner cases where the old format returned wrong information and introduces new capabilities like restricted output, recognized attribute changes, and more. For a complete overview on the new format, please take a look in the appropriate manual section. Bugs fixed - mtn automate heads called without a branch argument now properly returns the head revisions of the workspace's branch if called over mtn automate stdio - mtn commit no longer crashes if it creates a revision whose roster already exists, i.e. was left behind by the command `mtn db kill_rev_locally REV` (savannah #18990) Documentation changes - the documentation of the "--revs-only" (formerly "--brief") switch for the annotate command didn't match its actual behavior, this has been fixed - documentation for the "ssh_agent_add" command was missing and has been added Other - contrib/usher.cc has been removed. Please use the net.venge.monotone.contrib.usher branch instead. Internal - Update SQLite to 3.4.1. - Update Lua to 5.1.2 plus latest bug fixes. - Update Botan to 1.5.10. - Internal use of regular expressions has been almost eliminated. (Regular expressions are still used for .mtn-ignore and the --show-encloser feature of mtn diff, and are still available to Lua hooks.) @ text @d1 1 a1 1 $NetBSD$ @ 1.6 log @Update to 0.32. Tested under NetBSD 4.99.7 and Mac OS X Tiger: Changes - "mtn serve" no longer takes patterns on the command line. Use the permissions hooks instead. - the name of the option that denoted the revision from which "mtn log" should start logging was renamed from "--revision" to "--from" - author selectors no longer have implicit wildcarding - if you manually add entries to MTN/log while you are working, in preparation for an eventual commit, you will now be required to remove a "magic" template line from the file before the commit will succeed. This, like the test for an empty log file, helps to prevent accidents. - the "db regenerate_caches" migration command replaces the previous "db regenerate_rosters", generalising the task of rebuilding or generating cached data that may be added across an upgrade. Like "db migrate", which upgrades the database schema, this command fills in the data for new features. In this release, as well as rosters, it also adds "heights" information used to speed up topology operations. Speed improvements - "mtn annotate file" and "mtn log file" are generally much faster now, dependant on the number of revisions that changed the file. Both commands as well as "mtn automate toposort" make use of data called "heights" caching the topological order of all revisions. In order to create and use this data, the following must be run once for each db after upgrading: $ mtn -d mydb.mtn db regenerate_caches New features - "mtn automate content_diff" - "mtn automate get_file_of" (same as get_file, but expects a file path and optionally a revision) - "mtn import" command - "mtn log --to" - netsync_note_* hooks are given much more information, inlcuding a http/smtp/etc style status code - includedirpattern(dir, fileglob) function for hooks Bugs fixed - bug in "automate stdio" that would result in monotone garbling its input in some circumstances fixed - "mtn annotate file" and "mtn log file" are generally much faster now, dependant on the number of revisions that changed the file. Both commands as well as "mtn automate toposort" make use of data called "heights" caching the topological order of all revisions. - spawn_redirected hook function now understands a blank filename to mean not to redirect that stream - "mtn log" is now in proper topological order, also due to the use of cached "heights" data - reset options between "automate stdio" commands - another compile fix for gcc3 - bug in localization code where option strings where not always properly translated Other - botan library upgraded to 1.6.0 - accommodate changes in boost 1.34 - documentation for "mtn automate get_option" - notes/ directory @ text @d1 1 a1 1 $NetBSD: patch-ab,v 1.5 2006/12/11 08:13:13 dan Exp $ d3 9 a11 28 --- cmd_automate.cc eaa1ad0948253c507dfafc1521b8885260ba2db6 +++ cmd_automate.cc 5b90227ffcfe111cead5ab503b2df2ed81899835 @@@@ -28,7 +28,7 @@@@ namespace automation { static map * automations; automate::automate(string const &n, string const &p, options::options_type const & o) - : name(n), params(p), options(o) + : name(n), params(p), opts(o) { static bool first(true); if (first) @@@@ -350,7 +350,7 @@@@ AUTOMATE(stdio, "", options::opts::autom try { options::options_type opts = options::opts::globals(); - opts = opts | find_automation(cmd, help_name).options; + opts = opts | find_automation(cmd, help_name).opts; opts.instantiate(&app.opts).from_key_value_pairs(params); automate_command(cmd, args, help_name, app, os); } @@@@ -402,7 +402,7 @@@@ commands::cmd_automate::get_options(vect { if (args.size() < 2) return options::options_type(); - return find_automation(idx(args,1), idx(args,0)()).options; + return find_automation(idx(args,1), idx(args,0)()).opts; } @ 1.5 log @ fix from monotone mainline fbfd9b42c9e9fdb5d45e806cab07ad0dd57f7bf8: cmd_automate.cc, cmd.hh, ...: renamed member variable "options" to "opts" to make (hopefully) the gcc-3.3 build slaves happy This went in just after the 0.31 release, but never made it to pkgsrc until now. @ text @d1 1 a1 1 $NetBSD$ @ 1.4 log @Update to 0.31: 0.31 release. Code cleanups and bug fixes. New features: - If multiple --message (or -m) arguments are passed to 'commit', then they will be concatenated on separate lines. - The validate_commit_message hook is now told what branch the commit is on. Bugs fixed: - The typo that prevented building with gcc 3.3 has been fixed. - Attempting to commit without a signing key available now fails earlier. - Command-line option parsing has been redone yet again; this should fix a number of bugs caused by the use of boost::program_options. For instance, command line error messages are now l10nized again, "--depth=asdf" now gives a sensible error message instead of crashing, and --key= now works as an alternative to -k "". - A bug in the new roster caching logic that caused assertion failures on very large trees has been fixed. - A rare bug in the "epoch refinement" phase of the netsync protocol has been fixed. - Accidental (and undocumented) change to 'automate inventory' output format reverted; documentation is now correct again. - Some obscure error conditions with 'pivot_root' fixed. Many fixes to 'automate stdio': - IO handling has been rewritten, to remove some obscure bugs and clean up the code. - automate commands can now take options (even when used with 'automate stdio'). - The default block size has been increased to 32k (which should considerably reduce overhead). - Many automate commands were flushing their output far too often, causing major slowdowns when used with 'automate stdio'; this has been fixed. - Syntax errors now cause 'automate stdio' to exit, rather than attempting to provide usage information for the calling program to read. Other: - New large-coverage random testsuite for delta reconstruction path finding algorithm. - Miscellaneous code cleanups and improved error messages. - Enhancements to debian packaging. - New translation to es (Spanish). @ text @d1 1 a1 1 $NetBSD: patch-ab,v 1.3 2006/09/22 11:14:06 dan Exp $ d3 27 a29 21 --- lru_writeback_cache.hh 4ac60f38211fc4ef11ee1f968bfb88fee5446937 +++ lru_writeback_cache.hh c2fa384b9b29d6c5162fad70ebe96e6479b8aa3e @@@@ -212,19 +212,24 @@@@ public: { // Remove the last element. liter = _list.end(); - // Unless it's the only element (i.e., the one we just inserted) -- we - // never empty ourselves out completely + I(liter != _list.begin()); + --liter; + // liter now points to the last element. If the last element is also + // the first element -- i.e., the list has only one element, and we + // know that it's the one we just inserted -- then never mind, we + // never want to empty ourselves out completely. if (liter == _list.begin()) break; - --liter; this->_remove(liter->first); } + I(exists(key)); } a30 6 inline void insert_dirty(Key const & key, const Data & data) { insert_clean(key, data); safe_insert(_dirty, key); + I(is_dirty(key)); } a31 1 private: @ 1.3 log @Fix a regression introduced between 0.29 and 0.30 detected when handling large repositories like NetBSD src. Problem found by riz@@, fix from mainline a:njs/d:2006-09-22T07:01:26, tested by me. bump rev to nb1 ChangeLog: 2006-09-21 Nathaniel Smith * lru_writeback_cache.hh (clean_insert): Correct the cache overflow logic so as to _actually_ never remove the last element. Add an invariant to verify this. (insert_dirty): Add an invariant here too. @ text @d1 1 a1 1 $NetBSD$ @ 1.2 log @Changes in the package, coming from a monotone package I started creating yesterday w/o having checked for this one: - Beautify DESCR. - Take maintainership. - Handle monotone.info properly. - Mark the package as C/C++. - Use boost's toolset.mk to get the compiler name. - Simplify patches by using CPPFLAGS. - Add missing dependencies. While here, update to 0.15: - overhauled the internal representation of changes. see README.changesets for details. - fixed bugs in merkle trie synchronization code. - fixed echoing and progress UI bugs (helps when using in emacs). - upgraded cryptopp to 5.2.1. - fixed bug 8715, diff hunk coordinate reporting. - added figures, new tutorial to manual. - improve accuracy of log command. - several build, configure, and linkage fixes. - some OSX / PPC portability fixes. @ text @d1 1 a1 1 $NetBSD: patch-ab,v 1.1 2004/08/15 14:38:38 drochner Exp $ d3 21 a23 3 --- Makefile.in.orig 2004-08-14 14:45:22.000000000 +0200 +++ Makefile.in @@@@ -843,7 +843,7 @@@@ unit_tests_CPPFLAGS = -DBUILD_UNIT_TESTS d25 8 a32 7 @@STATIC_BOOST_FALSE@@monotone_LDADD = lib3rdparty.a libplatform.a $(am__append_6) @@STATIC_BOOST_TRUE@@monotone_LDADD = lib3rdparty.a $(BOOSTLIBS) libplatform.a $(am__append_6) -@@STATIC_BOOST_FALSE@@unit_tests_LDADD = lib3rdparty.a -lboost_unit_test_framework libplatform.a $(am__append_7) +@@STATIC_BOOST_FALSE@@unit_tests_LDADD = lib3rdparty.a -lboost_unit_test_framework${BLIBSF} libplatform.a $(am__append_7) @@STATIC_BOOST_TRUE@@unit_tests_LDADD = lib3rdparty.a $(BOOSTLIBS) \ @@STATIC_BOOST_TRUE@@ @@BOOST_LIBDIR@@/libboost_unit_test_framework.a\ @@STATIC_BOOST_TRUE@@libplatform.a $(am__append_7) @ 1.1 log @Initial revision @ text @d1 1 a1 1 $NetBSD$ @ 1.1.1.1 log @import monotone-0.14, another distributed revision control system @ text @@