# ChangeLog for eix - Ebuild IndeX for portage

*eix-0.34.4
	Martin Väth <martin at mvath.de>:
	- Simplify eix.proto:
	  - Use eix_proto namespace, avoid nesting messages unnecessarily
	  - Rename the main proto to not misleadingly indicate the source
	- Minor change of the proto wire format (extension w/ deprecation) to
	  allow distinction between empty and unset/equal effective keywords
	- Manpage typo fixes, and clarify that --proto is setting independent
	- Deal with some new gcc-10 warnings and clang-10
	- Put originally missing second part of eix-0.34.3 patch into repo

*eix-0.34.3
	Martin Väth <martin at mvath.de>:
	- Fix meson build problem with sqlite + protobuf

*eix-0.34.2
	Martin Väth <martin at mvath.de>:
	- Do not require protoc for autotools --without-protobuf

*eix-0.34.1
	Martin Väth <martin at mvath.de>:
	- Rework protobuf output: Expose the actual eix-internal storage for
	  masks and stability instead of the simplification for xml. This is
	  an exceptional change since I suppose that the proto from eix-0.34.0
	  was not used yet. In future, presumably only predicates will be added
	  to the output without changing the tag numbers.

*eix-0.34.0
	Martin Väth <martin at mvath.de>:
	- Support Google's protobuf output with --proto, currently untested.
	  Please report a bug if you have a use case for a modification or
	  (preferrably) extension of the proto, or it does not work.
	- Interpret /0 as no subslot. https://github.com/vaeth/eix/issues/70

*eix-0.33.11
	Martin Väth <martin at mvath.de>:
	- Bugfix: Fix typo in --with-portdir-default
	- Bugfix: Fix version for meson
	- Remove those sanitizers which cause linking problems on my systems

*eix-0.33.10
	Martin Väth <martin at mvath.de>:
	- Add --with-portdir-default with default /var/db/repos/gentoo,
	  replacing the previously hardcoded fallback /usr/portage,
	  see https://github.com/vaeth/eix/issues/69

*eix-0.33.9
	Martin Väth <martin at mvath.de>:
	- Support --longopt=parameter options
	- Introduce --color {never,always,auto} for overriding, see
	  https://github.com/vaeth/eix/issues/67
	- Remove -m*=thunk due to https://github.com/vaeth/eix/issues/64.
	  In gcc-9, this collides with -fcf-protection=full anyway.

*eix-0.33.8
	Martin Väth <martin at mvath.de>:
	- Output errors produced when reading repos.conf, see
	  https://github.com/vaeth/eix/issues/60
	- Add some missing NOEXCEPT
	- meson: add flags immediately to detect collisions
	- Adapt to gcc-9: Remove obsolete -mmpx -fcheck-pointer-bounds -Wchkp
	- Adapt to gcc-9: Prefer -fcf-protection=full over
	  colliding -mindirect-branch=thunk -mfunction-return=thunk

	Jan Ziak <0xe2.0x9a.0x9b at gmail.com>:
	- Improve performance of StringHash by using an unordered map

	Marco Leogrande <dark.knight.ita at gmail.com>:
	- Manpage typo fix

	Dmitriy Demura <dmitriy at demura.info>:
	- Russian translation fix

*eix-0.33.7
	Martin Väth <martin at mvath.de>:
	- Support relative pathnames in make.conf, see
	  https://github.com/vaeth/eix/issues/57
	- Update sqlite links, see https://github.com/vaeth/eix/issues/58

	Wolfgang Müller (https://github.com/vehk):
	- Add support for italics in COLORSTRING

*eix-0.33.6
	Martin Väth <martin at mvath.de>:
	- meson: rename optimization -> normal-optimization to avoid clash
	- meson: fix message for forced optimization
	- meson: internal: avoid clashing variable name continue, cosmetics

*eix-0.33.5
	Martin Väth <martin at mvath.de>:
	- New search options --{available,installed}-{deps,{,r,p,b}depend}
	- Support SRC_URI, https://bugs.gentoo.org/show_bug.cgi?id=666546
	  In detail: Add SRC_URI, <srcuri> {havesrcuri}, --src-uri,
	  extend DEFAULT_MATCH_FIELD by src-uri, output by default with -vl,
	  output in xml, new xml version 15, new eix database version 38
	- handle installed deps even without DEP
	- default to --with-deps-default --with-required-use-default
	- bugfix: empty LINGUAS w/ meson https://github.com/vaeth/eix/issues/54
	- do not let installed deps default to available deps if DEPS_INSTALLED
	- add yet another ATTRIBUTE_PURE
	- warn about not re-adding the ATTRIBUTE_PURE removed in 0.33.1
	- manpage: add a missing occurrence of BDEPEND
	- Makefile, meson.build: avoid non-ascii symbols

*eix-0.33.4
	Martin Väth <martin at mvath.de>:
	- Bugfix: eix-update: set reponame in version to correctly apply
	  package.mask from overlays https://bugs.gentoo.org/658658
	- Bugfix: do not treat leading * in system entries as part of pattern
	- Bugfix: Read vars like EAPI also with ONLY_KEYWORDS_SLOT.
	  This fixes cache method ebuild* for non-first versions of a package,
	  see https://github.com/vaeth/eix/issues/52
	- Use a variant of Levenshtein (Hirschberg) which requires less memory
	- zsh completion: Use standard style; use eix --format; simplify
	- Minor code simplification

*eix-0.33.3
	Martin Väth <martin at mvath.de>:
	- Rename hdepend->bdepend (EAPI=7)
	- Add configure options --with-eix-{user,group,uid,gid}
	- Fix configure test for emplace_back test to include vector<bool>
	  and prepend (not append) dialect flags when testing for consistency
	  with later usage, see https://bugs.gentoo.org/show_bug.cgi?id=649558
	- Add contrib/iwyu.sh experimentally
	- Include README.md, eix.png into distribution
	- Configure diag for -Wsuggest-final-methods -Wgnu-statement-expression
	- Until a good test is known, do not add -fno-plt: it breaks with clang
	- eix-functions.sh: Fix npgettext lookup test.
	- eix-functions.sh: Fix {eval_,}ngettext fallback: 0 has plural form
	- Update compiler-flags (e.g. include gcc-8)
	- meson.buid: Unify flag adding
	- Add dummy initialization to avoid gcc-8 false positive warning
	- Add ATTRIBUTE_NONNULL_ to template specialications

*eix-0.33.2
	Martin Väth <martin at mvath.de>:
	- Bugfix: Fix typo in FORMATLINE_BEST (I18N_BESTVERSIONSSLOT) which
	  prevented printing the text "Best versions/slot:" (or a translation)
	  with eix -v
	- Introduce jumbo build and use it by default: This avoids a splitting
	  of defaults.cc into several compilation units which existed only for
	  compile-time memory reasons.
	  Use --disable-jumbo-build for compilation on low-memory systems.
	  (The produced binary might be longer and slower.)
	- Use all available compiler spectre protection with --enable-security.
	  Some are overkill for some processors, but better avoid leaking data.

*eix-0.33.1
	Martin Väth <martin at mvath.de>:
	- Fix meson tests which require dialect arg
	- Use unordered_{map,set} if available (presumably faster)
	- Use emplace, emplace_back, std::move if available (faster)
	- Prefer count() != 0 over find() != end() (presumably faster)
	- Use final and override where possible/useful; add diagnostics
	- Add memory saving flags
	- Remove an ATTRIBUTE_PURE which is suspicious to be related to
	  https://github.com/vaeth/eix/issues/42
	- Improve ccache usage (only for meson): Remove version from config.h
	  (included everywhere) to increase hit rate for upgrades
	- manpage: more verbose description of eix-postsync

*eix-0.33.0
	Martin Väth <martin at mvath.de>:
	- Provide eix-postsync as a simple eix-sync alternative.
	  To use it with >=portage-2.3.7 create /etc/portage/postsync.d/ and
	  ln -s /usr/bin/eix-postsync /etc/portage/postsync.d/50-eix-postsync
	  After syncing, call eix-diff manually if required
	- Provide a wrapper for eix-function.sh
	- Add support for meson/ninja build system (fast, but less compatible).
	  Probably autotools and meson will both continue to be supported.
	- Make new dialect the default
	- Pass flags to disable dead code. Use --disable-dead-code-elimination
	  (with meson -Ddead-code-elimination=false) to disable this feature.
	- Fix return value in _eix completion. Thanks to
	  Mikael Magnusson <mikachu at gmail.com>
	- Fix option and description typos in configure.ac
	- Replace /etc/make.conf by the new path /etc/portage/make.conf in docs
	- eix-layman help text: Avoid mentioning source commands in make.conf
	  since most users will use nowadays repositories.conf

*eix-0.32.10
	Martin Väth <martin at mvath.de>:
	- Fix subversion guessing without CAREMODE=true
	- Fix wrong prepending of CXXFLAGS and LDFLAGS
	- Fix stress test for C++11 (fixes compilation with clang & gcc++ lib)
	- Fix adding of -fstack-protector-strong and friends
	- Add -fstack-check
	- Fix stress test in ./configure to not spit warning which breaks
	- Fix some sign type conversions
	- Support fallthrough attribute
	- Correctly support mpx with --enable-strong-security if available
	- Fix get_geometry() error check
	- Initialize some variables redundantly to make stupid compiler happy

*eix-0.32.9
	Martin Väth <martin at mvath.de>:
	- Support -* in profile "packages" files, see
	  https://bugs.gentoo.org/show_bug.cgi?id=610670
	- Fix remove_line(): Do not consider removing removed lines a change
	- Only treat 0 and - as "false" if nothing else follows, see
	  https://bugs.gentoo.org/show_bug.cgi?id=616972
	- internal: Add some pure attributes where possible
	- internal: Switch off format-nonliteral warning for strftime

*eix-0.32.8
	Martin Väth <martin at mvath.de>:
	- IMPORTANT: For testing scripts, it is strongly recommended to use
	  DEFAULT_MATCH_FIELD=error
	  DEFAULT_MATCH_ALGORITHM=error
	  (and to always specify match field/algorithm explicitly).
	  This way the scripts are independent of current and/or future changes
	  of the default heuristics for these values.
	- Change MATCH_ALGORITHM_EXACT so that --exact is default if the match
	  field contains _only_ things like USE-flags, sets, EAPI, or SLOT
	  and search string does not look like a regex/pattern.
	  This may cause a surprising change of behaviour and might perhaps be
	  reverted.
	- Change MATCH_ALGORITHM_SUBSTRING to apply if match field contains
	  only category/version and search string does not look like a
	  regex/pattern.
	- Introduce a default match algorithm/field "error" so that the user
	  can be required to explicitly specify algorithm/field, thus avoiding
	  any dangerous heuristics.
	- Extend DEFAULT_MATCH_ALGORITHM to allow checking for match field(s).
	- Fix segfault (regression from eix-0.32.7): Provide correct copy/move
	  constructor/assignment in poor man's (not C++-11) eix::forward_list,
	  https://github.com/vaeth/eix/issues/39

*eix-0.32.7
	Martin Väth <martin at mvath.de>:
	- internal: provide and use eix::array instead of hacks
	- internal: provide rudimentary eix::forward_list if necessary
	- internal: use std::.. for documented functions from #include <c...>
	- internal: remove UNUSED hackery
	- contrib/make.sh: Improve clang filtering

*eix-0.32.6
	Martin Väth <martin at mvath.de>:
	- Fix compilation with C++98 (regression from eix-0.32.5), see
	  https://github.com/vaeth/eix/issues/35
	- Make default values of MATCH_FIELD_LICENSE and MATCH_FIELD_EAPI
	  identical to MATCH_FIELD_CATEGORY_NAME so that they never match and
	  are de facto disabled. This hopefully avoids user confusion without
	  breaking user configs. https://bugs.gentoo.org/show_bug.cgi?id=610424
	- Let default value of MATCH_ALGORITH_EXACT be less aggressive

*eix-0.32.5
	Martin Väth <martin at mvath.de>:
	- Fix completion with >=zsh-5.3
	- Find first valid substring for some algorithms when matching package
	  paths, see https://bugs.gentoo.org/show_bug.cgi?id=605978
	- Modify heuristics for default choice of match algorithm
	- Treat category/package:slot/subslot[::repo] as valid for --pipe
	  Previously, words with more than one / had been tacitly ignored
	- Fix duplicate detection with overlays for parsing cache methods
	- Add configure option for treatment of const virtuals
	- xml output: use shortcuts like <description/>. Omit some quoting
	- eix-header: Update help text for recommended functions.sh usage
	- internal: Use eix::format throughout for output
	- internal: Use C++11 if available: delete copy constructors, override
	- internal: Use constexpr for const arrays; gcc might otherwise produce
	  unexpected copies: https://forums.gentoo.org/viewtopic-t-1057674.html
	- internal: Finally implement ExplodeAtom in a sane manner
	- internal: Use array class if available (might require C++11)
	- internal: forward_list instead of list if available (C++11)
	- internal: Use reserve() (when it presumably really optimizes)
	- internal: Introduce generic ptr_container
	- internal: Use vector instead of list when presumably more efficient
	- internal: Place attributes in front. Use attribute.h. Use C++11 attr
	- internal: Use bugchecking include order

*eix-0.32.4
	Martin Väth <martin at mvath.de>:
	- Fix --disable-swap-remote (which falsely was equivalent to
	  --enable-swap-remote; only omitting the argument could switch it off)
	  Be aware of the following simultaneous change.
	- Interchange the two remote paths (i.e. invert in a sense the meaning
	  --{en,dis}able-swap-remote or of not specifying it, respectively):
	  http://dev.gentooexperimental.org is dead since quite a while;
	  so it makes sense to choose it only as the alternative update2/fetch2
	  by default. To revert to the previous defaults, use
	  --enable-swap-remote (or omit now using --{en,dis}able-swap remote,
	  respectively; note that --disable-swap-remote was faulty, see above).
	- Introduce EIX_TMPDIR to allow overriding of TMPDIR, see
	  https://bugs.gentoo.org/show_bug.cgi?id=600648

	Wim Muskee:
	- Manpage typo fix

*eix-0.32.3
	Martin Väth <martin at mvath.de>:
	- Explain on manpage how to retain functionality of previous USE-flags
	  which were removed from the ebuild for political reasons
	- In particular, explain how to install with CXXFLAGS/LDFLAGS which are
	  recommended by the eix maintainer for optimization and security
	- Modify instructions in segfault message to match USE-flags crippling
	- --enable-new-dialact only in case of optimizations

*eix-0.32.2
	Martin Väth <martin at mvath.de>:
	- Make eix-functions.sh a program to be used with "eval"

*eix-0.32.1
	Martin Väth <martin at mvath.de>:
	- eix-remote: Use https://github.com/vaeth/quoter/ instead of hacks
	- contrib: reintroduce contrib/mkrelease.sh as a wrapper to git-tag

*eix-0.32.0
	Martin Väth <martin at mvath.de>:
	- eix-diff now requires much less memory:
	  https://bugs.gentoo.org/show_bug.cgi?id=597380
	- Remove DIFF_PRINT_HEADER (package count can no longer be predicted)
	- New database format 37 is guaranteed to be alphabetically sorted
	- Add eix-header -C
	- scripts: fix getopts usage
	- scripts: handle export Bourne shell compatible
	- Fix typos in zsh completion. Thanks to longlene
	- Fix various operator overloads
	- Define some move operations not provided by defaults
	- Remove git helpers; rely on https://github.com/vaeth/git-wrappers-mv

*eix-0.31.10
	Martin Väth <martin at mvath.de>:
	- Avoid redundant newlines in eix -t
	- Add --nowarn to eix and masked-packages
	- Print portage warnings only once, in particular in eix-test-obsolete:
	  https://bugs.gentoo.org/show_bug.cgi?id=593774
	- Always count empty lines for line numbers in errors/warnings
	- Print line numbers for warnings with bad category names
	- eix-sync: Prepend /etc/eix-sync.conf only once, see
	  https://github.com/vaeth/eix/issues/30
	- scripts: speed up ReadBoolean

*eix-0.31.9
	Martin Väth <martin at mvath.de>:
	- scripts: set -f in loops over lines:
	  https://bugs.gentoo.org/show_bug.cgi?id=591422
	- Improve yesno detection (0 or - is false)

*eix-0.31.8
	Martin Väth <martin at mvath.de>:
	- Important change: Add -u for all shell scripts.
	  This can trigger currently hidden bugs ("unbound variable" or
	  "parameter not set" indicate such bugs); please report them if you
	  encounter them. The hope is that in the long run, this step will help
	  to eliminate hidden bugs, although in the moment it can cause pain.
	- Thanks already for some such reports to:
	  Michał Bartoszkiewicz <mbartoszkiewicz at gmail.com>
	- Add etcat script to resemble somewhat etcat -v behaviour
	- New FORMAT_ETCAT and FORMAT_VERSION_ETCAT default variables
	  for etcat type output
	- Support colors in unquoted runtime variables
	- New attributes overlayname, overlay{plain,ver}name{,*}
	- Honour ebuild md5 whenever provided (e.g. by >=portage-2.3.1)
	- Make timestamp optional for assign method (e.g. for >=portage-2.3.1)
	- Honour TMPDIR as requested by POSIX
	- Some more Bourne shell compatibility
	- Internal change: Turn readers into a class
	- Add gcc-6 options

*eix-0.31.7
	Martin Väth <martin at mvath.de>:
	- Support EAPI and (optionally) REQUIRED_USE
	- Extend MATCH_ALGORITHM_EXACT default value to let a pure number match
	  exactly by default (useful for EAPI matching)
	- Reintroduce extended version globbing to match portage:
	  https://github.com/vaeth/eix/issues/21
	  https://bugs.gentoo.org/show_bug.cgi?id=572178
	- Fix metadata-{flat,assign} confusion
	- Fix: version parsing must omit [
	  https://github.com/vaeth/eix/issues/20
	- Fix memory leak of eix -t (delete all_packages) and sqlite
	- Fix xml output with illegal use flags

*eix-0.31.6
	Martin Väth <martin at mvath.de>:
	- Fix further memory leak in PrintFormat
	- Bugfix: Manually destroy root_node; do not do this in FormatParser.
	  This hopefully fixes the segfault of bug 571108

*eix-0.31.5
	Martin Väth <martin at mvath.de>:
	- Adapt version wildcard matching to pms and portage implementation
	- Fix memory leaks in PrintFormat. Thanks to
	  Agostino Sarubbo <ago at gentoo.org> for reporting and for
	  pointing out that PAX kernel blocks --fsanitize=address
	- Add --enable-nopie-security to make --fsanitize=address work better
	- Switch back to .tar.xz releases as default (github supports it now)

*eix-0.31.4
	Thomas D. <whissi at whissi.de>:
	- Fix forgotten slash in eix-sync breaking hooks support (regression
	  from eix-0.31.3), see https://bugs.gentoo.org/show_bug.cgi?id=565504

	Martin Väth <martin at mvath.de>:
	- Support repo-name from metadata/layout.conf
	- Fix --without-nls output of eval_pgettext (regression from
	  eix-0.31.3), see https://forums.gentoo.org/viewtopic-t-1034606.html
	- Fix forgotten slash in eix-sync (regression from eix-0.31.3)

*eix-0.31.3
	Martin Väth <martin at mvath.de>:
	- Use context for internationalization of variable defaults,
	  statusline, and others
	- No utf8 in statusline https://bugs.gentoo.org/show_bug.cgi?id=518800
	- eix-functions.sh: StatusInfo(): argument must now be one string;
	  optionally an alternative statusline can be added
	- eix-functions.sh: RunCommand() has Statusline as a new 1st argument;
	  the special value "=" as 1st arg means the 2nd argument is used
	- eix-functions.sh: provide {eval_,}{n,}pgettext
	- internal: remove line numbers from .po

*eix-0.31.2
	Daniel M. Weeks <dan at danweeks.net>:
	- Initial implementation for FEATURES=binpkg-multi-instance

	Martin Väth <martin at mvath.de>:
	- Internationalization and German translation of default FORMATSTRING
	- Force fuzzy Russian translation of default FORMATSTRING and others.
	  Linguistic fixes are probably needed and welcome.
	- Add --multi-binary
	- manpage: document how to translate default FORMATSTRING to
	  other languages
	- Use UTF8 quotes in German and Russian translation
	- Extend support for FEATURES=binpkg-multi-instance: Be stricter in
	  filename syntax, add further attributes, use new colors
	- Update documentation for FEATURES=binpkg-multi-instance
	- internal: Rework eix::format: support positional arguments like %2$s.
	  Also remove %v but add inferior %d (not quoting strings) instead.
	  Both changes are needed for clean internationalization with gettext.
	- Fix --disable-nls cleanly so that inclusion order plays no role and
	  thus N_() can be used also in .h (although it currently isn't)
	- Build system: remove unneeded checks for endianness

*eix-0.31.1
	Martin Väth <martin at mvath.de>:
	- Urgent release with a hackish fix for --disable-nls, see
	  https://bugs.gentoo.org/show_bug.cgi?id=563680

*eix-0.31.0
	Martin Väth <martin at mvath.de>:
	- Restore sh bracing for eval_gettext to fix poor man's implementation,
	  see https://forums.gentoo.org/viewtopic-t-1029872.html
	- Add tmux to TERM_{SOFT,}STATUSLINE
	- Lock EIX_CACHEFILE (currently implemented without any timeout), see
	  https://bugs.gentoo.org/show_bug.cgi?id=553574
	- Fix handling of magic * in EIX_LOCAL_SETS, see
	  https://bugs.gentoo.org/show_bug.cgi?id=553262
	- Update manpage: Suggest usage (and example) of
	  /etc/portage/repo.postsync.d/50-egencache
	- Improve English in some output
	- Use ngettext when possible
	- Use AM_GNU_GETTEXT_REQUIRE_VERSION which is finally available, hence:
	- Require >=gettext-0.19.6 to create the tarball

*eix-0.30.11
	Martin Väth <martin at mvath.de>:
	- Honour profiles/package.unmask, https://github.com/vaeth/eix/issues/9
	- Verify (and in trivial cases: fix) content of profiles/categories
	  https://github.com/vaeth/eix/issues/8
	- Fix ATTRIBUTE_NONNULL in PortageUserConfig constructor
	- Remove redundant sh quoting

*eix-0.30.10
	Martin Väth <martin at mvath.de>:
	- Let only non-main profiles/package.mask mask repo-locally
	- Bump to gettext-0.19.4

*eix-0.30.9
	Martin Väth <martin at mvath.de>:
	- Fix https://bugs.gentoo.org/show_bug.cgi?id=551108
	  This was a regression of eix-0.30.0 by resetting default repo. Fix by
	  using only local logic: Allow default repo only in parseMask()
	- Add gcc-5 options
	- Add other optimization options which should have been added earlier
	- Fix ATTRIBUTE_NONNULL in OverlayIndent constructors and get_destcat
	- Remove some redundant NULL tests
	- Work around a gcc-5 regression when comparing empty vectors
	- eix-remote: Support combination of -x and -X,
	  see https://forums.gentoo.org/viewtopic-t-1016652.html

*eix-0.30.8
	Martin Väth <martin at mvath.de>:
	- Order mask lists (/etc/portage/package.*) more similar to portage:
	  a) Wildcards are treated alphabetically
	  b) More specifications (slot, repo, version, ...) increase priority
	  However, b) is only distinguished for identical wildcards;
	  this slightly deviates from portage.
	- Fix eix-header: Let -f create fresh data, not add to previous -f
	- eix-header: Special treatment of empty OV value
	- eix-remote: Compatibility fix for recent changes in zugaina, see
	  https://github.com/vaeth/eix/issues/6

*eix-0.30.7
	Martin Väth <martin at mvath.de>:
	- Support @profile: Add it to database (new database version: 34)
	- new option --profile{,+,-}
	- new package property "profile"
	- include profile in "allsetnames"
	- encode @profile as colors in default format strings
	- separate PORTDIR_OVERLAY by newlines (for output in other programs)
	- internal tools: avoid GREP_OPTIONS for compatibility with grep-2.21

*eix-0.30.6
	Martin Väth <martin at mvath.de>:
	- Fix: pass EAPI in environment for cache method ebuild*.
	  Thanks to Michał Górny for the analysis, see
	  https://bugs.gentoo.org/show_bug.cgi?id=532952
	- internal: silence some stupid cpplint warnings

*eix-0.30.5
	Martin Väth <martin at mvath.de>:
	- Let PORTAGE_REPOSITORIES _override_ (not only add to) repos.conf
	  to match with portage's behaviour
	- zsh-completion: fix checks for _tags
	- internal: Replace type(...) casting by static_cast<type>(...)
	- German translation: Many fixes and extensions by
	  André Jaenisch <andre.jaenisch at openmailbox.org>
	- Bump gettext

*eix-0.30.4
	Martin Väth <martin at mvath.de>:
	- Fix for lines ending with EOF (regression introduced in eix-0.30.3),
	  see https://github.com/vaeth/eix/issues/3
	- Support PORTAGE_REPOSITORIES
	- internal: Support parsing from mem for varsreader.cc

*eix-0.30.3
	Martin Väth <martin at mvath.de>:
	- Replace default [D] by [?] to avoid confusion with portage output
	- Remove EXEC_EBUILD, EXEC_EBUILD_SH, PORTAGE_BIN_PATH
	- Remove --with-ebuild-sh-default, --with-portage-bin-path
	- Use instead portageq, PATH, and/or EPREFIX_PORTAGE_EXEC to obtain
	  paths, see https://bugs.gentoo.org/show_bug.cgi?id=520436
	- eix-sync: Add -vn (verbose dry-run) options, see
	  https://bugs.gentoo.org/show_bug.cgi?id=509750
	- Fix more wrong usages of getline(), see
	  https://bugs.gentoo.org/show_bug.cgi?id=506106
	- Fix include files for ioctl() on solaris, see
	  https://bugs.gentoo.org/show_bug.cgi?id=510120
	- Trigger clang/gcc/glibc incompatibility when testing for c++1y, see
	  https://bugs.gentoo.org/show_bug.cgi?id=510102
	- Fix gdb instructions
	- manpage: add remarks about stability of names in future eix versions
	- manpage: fix typo in SPEEDUP
	- Russian translation: Improve spacing for eix-test-obsolete, and
	  https://developer.berlios.de/bugs/?func=detailbug&bug_id=19366&
	  group_id=11597
	- cosmetical changes to be cpplint compliant
	- Add CPPLINT.cfg

*eix-0.30.2
	Martin Väth <martin at mvath.de>:
	- Move repository to GitHub since BerliOS closes
	- Fix wrong usage of getline(), see
	  https://bugs.gentoo.org/show_bug.cgi?id=506106
	- Prefer eix::format over printf
	- Introduce REQUIRE_DROP, NODROP_FATAL, see
	  https://bugs.gentoo.org/show_bug.cgi?id=499944
	- Use gcc-4.9 features if available
	- Initialize bools to make gcc-4.9 diagnostics happy
	- Add tmpfiles.d/eix.conf default
	- scripts: act as with USE_NLS=no if gettext.sh cannot be found, see
	  https://bugs.gentoo.org/show_bug.cgi?id=500130
	- Several translation fixes for German. Thanks to
	  André Jaenisch <andre.jaenisch at openmailbox.org>,
	  Dominik Geyer <dominik.geyer at gmail.com>, and
	  Justin Lechner <jlec at gentoo.org>
	- manpage: Add reference to color variables in FAQ and fix typos from
	  https://bugs.gentoo.org/show_bug.cgi?id=499932

*eix-0.30.1
	Martin Väth <martin at mvath.de>:
	- Add maskreason to --xml output
	- Update email
	- internal: remove redundant ATTRIBUTE_NONNULL
	- contrib/make.sh: Fix ccache handling
	- Check whether dropping of permissions worked
	- Fix message if EIX_LIMIT{,_COMPACT} equals number of all packages
	  output, see https://bugs.gentoo.org/show_bug.cgi?id=497232
	- eix-header: Add comment that -f is non-optional for portable scripts,
	  see https://bugs.gentoo.org/show_bug.cgi?id=497572
	- Use -fPIE -pie with --enable-security, see
	  https://bugs.gentoo.org/show_bug.cgi?id=497546

*eix-0.30.0
	Martin Väth <martin at mvath.de>:
	- Introduce {mainrepo} attribute; "Find open bugs" only for those, see
	  https://developer.berlios.de/bugs/?func=detailbug&bug_id=19160&
	  group_id=11597
	- Introduce OMIT_EXPAND and <{colli,}use0> attributes to omit output of
	  USE_EXPAND variables: https://bugs.gentoo.org/show_bug.cgi?id=489306
	- Let --version output only version number
	- scripts: Let -i also reset -H properly
	- Avoid #include of algorithm.h in packagetest and of packagest.h in
	  mask.h. This avoids "phantom" code with some gcc versions and flto:
	  https://developer.berlios.de/bugs/?func=detailbug&
	  bug_id=19182&group_id=11597
	- Use constexpr for some constants (if supported by dialect, of course)
	- Use gettext-0.18.3: allows to specify GNU sed
	- Remove --with-nongnu-cxx (no kludges needed anymore for current clang
	  and/or ccache; it is now the task of the user to export appropriate
	  CCACHE_CPP2 or CXXFLAGS if he needs them for older versions.)
	  This fixes https://bugs.gentoo.org/show_bug.cgi?id=455078
	- Remove --enable-quickcheck (makes no sense without manual distinction
	  of compilers which was just removed)
	- Dump /usr/bin/env shebangs unless --with-eprefix-default is set.
	  Thanks to SteveL for convincing me.
	  Introduce --with{,out}-sh-shebang for overriding if necessary
	- Code cleanup: Pass Google's current cpplint #296/r116 with braces,
	  avoid default arguments, add more typedefs, put some member
	  functions into separate file, standard reference syntax Type&

*eix-0.29.6
	Martin Väth <martin at mvath.de>:
	- --format now overrides all format related options/variables like
	  e.g. FORMAT or DEFAULT_FORMAT. In contrast, without --format, which
	  of the FORMAT* variable takes effect still depends on DEFAULT_FORMAT.
	- --format-verbose and --format-compact have been removed:
	  Use --format or set FORMAT_{VERBOSE,COMPACT} instead.
	- New option -N (--normal) instead of letting -c or -v be toggling
	- Fix: Do not cut rest of dependency strings after ${DEPEND}/${RDEPEND}
	  (Regression of eix-0.29.5)
	- Avoid strcasecmp() which is not in C++ standard and unspecified
	  outside POSIX locale. Also be more tolerant (match substrings only).
	- manpages: Prefer --format in the examples now since it is simpler
	- manpages: Avoid some warnings with groff-1.22.2

*eix-0.29.5
	Martin Väth <martin at mvath.de>:
	- Shorten database size dramatically (sort StringHashs by frequency)
	- versionsort: Fix versionsort -[pn], introduce -[fV], let -? support
	  more than one argument, ignore some non-alphanumeric rubbish
	- fix calculation of output of revision text
	- eix -R/-Z: Print clearer error message,
	  see https://bugs.gentoo.org/show_bug.cgi?id=485488
	- eix-sync: Fix checking EIX_CACHEFILE when path changed,
	  see https://bugs.gentoo.org/show_bug.cgi?id=485152
	- Improve storage of depends (simpler and better compression)
	- --print-all-depends works now as documented (no doublequote symbols)

*eix-0.29.4
	Martin Väth <martin at mvath.de>:
	- Fix wrong usage of ATTRIBUTE_NONNULL for Node *root (caused even
	  segfaults), see https://bugs.gentoo.org/show_bug.cgi?id=483392
	- Fix cache-method sqlite: Set overlay_key in read versions, see
	  https://bugs.gentoo.org/show_bug.cgi?id=480604
	- Fix: deep StringList comparison; fixes duplicate occurence of
	  mask reasons in certain cases

*eix-0.29.3
	Martin Väth <martin at mvath.de>:
	- Add EIX_LIMIT{,_COMPACT} as a safety measurement for typos.
	  This solves somewhat https://bugs.gentoo.org/show_bug.cgi?id=346665

*eix-0.29.2
	Martin Väth <martin at mvath.de>:
	- Fix random bugs for masks with slot and no subslot
	- Fix eix-update: Make long option --override-method work
	- Fix eix-update: ebuild*: export PORTAGE_REPO_NAME and a rudimentary
	  PORTAGE_ECLASS_LOCATIONS (>=portage-2.2.0_alpha89 needs it),
	  see https://bugs.gentoo.org/show_bug.cgi?id=478320
	- Fix colorless output: https://bugs.gentoo.org/show_bug.cgi?id=478818
	- Read also PORTAGE_REPOS_CONF /usr/share/portage/config/repos.conf
	- Fix zsh-completion: use _call_program, force DEFAULT_FORMAT=normal
	- manpage: cosmetical fixes

*eix-0.29.1
	Martin Väth <martin at mvath.de>:
	- Fallback to PORTDIR=/usr/portage, see
	  https://forums.gentoo.org/viewtopic-p-7349870.html
	- Make --installed-slot work again (regression from 0.26.0).
	  Thanks to Vikraman <vikraman at gentoo.org> for finding and fixing,
	  see https://bugs.gentoo.org/show_bug.cgi?id=475564
	- Fix md5 algorithm for files of 8192 bytes or longer (serious bug
	  from the beginning which made md5-cache for longer ebuilds invalid).
	- Fix possible memory leaks.

*eix-0.29.0
	Артём Воротников <skybon at gmail.com>:
	- Russian translation updates

	Martin Väth <martin at mvath.de>:
	- eix-sync now calls eix-remote add1/add2 by default if necessary
	- Support /etc/portage/repos.conf, honouring priority values
	- Currently, --print PORTDIR and --print PORTDIR_OVERLAY are modified
	  by repos.conf: useful for scripts, and faster than portageq.
	  If possible, this feature will be kept in future versons of eix
	- PORTDIR and PORTDIR_OVERLAY are probably going to be supported
	  as input variables for eix, even if perhaps deprecated by portage,
	  see https://bugs.gentoo.org/show_bug.cgi?id=474590
	- Let dependency attribute work for installed versions
	- Introduce --deps-installed, DEPS_INSTALLED
	- Introduce new tool eix-header
	- Remove now redundant --print-overlay-{path,label,data}, --is-current
	- eix-remote: speedup of add action due to single call of eix-header
	- Fix --*depend in zsh-completion
	- Allow /etc/eix-sync.conf not ending with newline. Use read -r
	- Fix some paths in manpage (regression from eix-0.28.3)
	- cleanup: Transform io namespace to class: automatic close on cleanup
	- German translation updates, improvements, and fixes, see e.g.
	  https://bugs.gentoo.org/show_bug.cgi?id=475076

*eix-0.28.5
	Martin Väth <martin at mvath.de>:
	- Allow that eix/portage config files are (recursively read) dirs,
	  see e.g. https://bugs.gentoo.org/show_bug.cgi?id=467276
	  Make /etc/eixrc a directory by default
	- Support for /etc/portage/package.accept_restrict, see
	  https://bugs.gentoo.org/show_bug.cgi?id=467622
	- eix-remote: add option -x/-X to exclude redundant local/remote
	  layman overlays. -x used to be the default in earlier eix versions,
	  but this was reverted since users do not expect it,
	  see https://bugs.gentoo.org/show_bug.cgi?id=467958
	  Put EIX_REMOTE_OPTS=-x into /etc/eixrc to keep the old behaviour.
	- eix-remote: make LOCAL_LAYMAN heuristic work with layman-2.0.0
	- Fixes in zsh completion for --print, --*overlay*, eix-installed-after
	- -fstrict-enums -fno-rtti -fno-threadsafe-statics -fvect-cost-model
	- eix-sync: Exit with status 3 if neither main tree changed nor
	  option -a was used nor have_changed=: is set, see
	  https://developer.berlios.de/bugs/?func=detailbug&
	  bug_id=18973&group_id=11597

*eix-0.28.4
	Артём Воротников <skybon at gmail.com>:
	- Russian translation fixes and updates

	Martin Väth <martin at mvath.de>:
	- Ethan Schoonover's Solarized has now its own color schemes in eix:
	  Set SOLARIZED if you use solarized (and omit setting TERM_ALT3 or
	  COLORSCHEME? as was recommended previously: this would choose a
	  standard scheme instead of the special adaption to solarized
	  system colors)
	- Use -fno-fat-lto-objects

*eix-0.28.3
	Артём Воротников <skybon at gmail.com>:
	- Russian translation fixes and updates

	Martin Väth <martin at mvath.de>:
	- Support secondary remote database:
	  Default: gentoo and zugaina (http://gpo.zugaina.org)
	  New option -Z, --remote2; modify REMOTE_DEFAULT correspondingly
	- Fix eix-update with REMOTE_DEFAULT, see
	  https://bugs.gentoo.org/show_bug.cgi?id=462040
	- Fixes in zsh-completion: content, tags, use always _description,
	  add some forgotten options like -R (and new -Z)
	- Make copy-constructors non-explicit as required in new C++ standard,
	  see https://bugs.gentoo.org/show_bug.cgi?id=456064
	- Pass Google's current cpplint-3.231
	- Be nice to gcc-4.8 (e.g. -Wno-long-long for broken pre-includes)
	- Use gettext-0.18.2
	- Compile also with <automake-1.13

*eix-0.28.2
	Артём Воротников <skybon at gmail.com>:
	- Russian translation fixes and updates

	Martin Väth <martin at mvath.de>:
	- Introduce -# as short form of --only-names, see
	  https://bugs.gentoo.org/show_bug.cgi?id=453180
	- Fix regression of re-enumeration of overlays from eix-0.28.0, see
	  https://bugs.gentoo.org/show_bug.cgi?id=452548
	- Allow full mask syntax in --pipe.
	- Introduce --pipe-mask.
	- Remove --pipe-name --pipe-version hacks (--pipe-mask is better).
	- Removed <installedmarkedversions:*> since new --pipe matches only
	  available versions. As a substitute use e.g. <markedversion:VAR>
	  with VAR={installedversion}...{}
	- Introduce --enable-strong-security
	- Make \C<number> slower but reliable (OutputString::..._smart())

*eix-0.28.1
	Martin Väth <martin at mvath.de>:
	- Fix regression {*var} and {!*var} from eix-0.28.0
	- Simplify eix-diff default format (use new \C<number> feature)
	- Improve parsing in --pipe: cut garbage after versions, see
	  https://forums.gentoo.org/viewtopic-p-7222402.html

*eix-0.28.0
	Martin Väth <martin at mvath.de>:
	- Add --print-profile-paths
	- Add --print USE.profile for profile's USE; USE is only make.conf's
	- manpage: add section how to speed up eix-update / handle metadata
	- new powerful tabulator escape sequence: \C<number>
	- Honour RESTRICT_INSTALLED as documented
	- Extend COLORFGBG_DARK default for rxvt-unicode with XPM support.
	- Treat (none;black) correctly
	- eix-remote -q: Pass --no-verbose to wget and redirect its output,
	  see https://forums.gentoo.org/viewtopic-p-7208430.html
	- eix-functions.sh: Introduce Echo
	- Support AC_CONFIG_MACRO_DIR[S] of automake-1.13, keep ACLOCAL_AMFLAGS
	- Remove support for <automake-1.11.2
	- Move tools to contrib/

*eix-0.27.6
	Martin Väth <martin at mvath.de>:
	- PLEASE READ THIS IF YOU HAVE TROUBLES WITH COLORS/BACKGROUND:
	  Try the hints in the beginning of the BUGS section of the eix
	  manpage.
	- The background guess heuristics defaults now to black background
	  unless there are sure signs that the background is white.
	  This makes the default choice more consistent with <=eix-0.27.4.
	- Revert BG1=black, BG3=white since it is the only way to guarantee
	  readability if the heuristic fails.
	- Remove rxvt from 256 color terminals: If your rxvt supports
	  256 colors but sets only TERM=rxvt, you have to change this to
	  TERM=rxvt-256color or put TERM_ALT1_ADD=rxvt to /etc/eixrc
	- Users of transparent terms should therefore set BG1=none BG3=none
	  in /etc/eixrc, see the BUGS section of the eix manpage.
	- Make RESET_ALL_LINES=true the default to make forced background color
	  (reversion of BG?-defaults) as non-intrusive as possible.
	- Change TERM_DARK meaning to allow a finer heuristics
	  (e.g. tune whether COLORFGBG is respected).
	- Clear comment lines correctly in non-mask files, see
	  https://bugs.gentoo.org/show_bug.cgi?id=446078 and
	  https://developer.berlios.de/bugs/?func=detailbug&
	  bug_id=18819&group_id=11597
	- Recognize file-ends better for comment separation

*eix-0.27.5
	Martin Väth <martin at mvath.de>:
	- Support mask reason (by default output with --versionlines), see
	  https://developer.berlios.de/feature/?func=detailfeature&
	  feature_id=5605&group_id=11597
	- If you have inconvenient colors (e.g. on dark or transparent terms),
	  this is probably due to the new DARK heuristics (see below).
	  Put DARK=true (or false) into /etc/eixrc to avoid it.
	  See the BUGS section on the eix manpage.
	- To improve readability by forcing the background colors, set
	  BG0=black, BG1=black, BG2=white, BG3=white in /etc/eixrc:
	  The defaults are all reverted to "none" due to DARK heuristics
	  and to eliminate problems with broken terminals (see below).
	  See the BUGS section on the eix manpage.
	- Renamed variables to allow more cases of terminals:
	  TERM_ALT{1..3} corresponding to COLORSCHEME{0..3}
	  Previous TERM_ALT and COLORSCHEME{,_ALT} variables are obsolete.
	- Example: To select 8/16 colors on all terminals and dark background
	  (e.g. for solarized) set TERM_ALT3=. and e.g. DARK=true or
	  COLORSCHEME3=0.
	  See the BUGS section on the eix manpage.
	- Let DARK=auto use a heuristic based on TERM_DARK, COLORFGBG_DARK,
	  see https://bugs.gentoo.org/show_bug.cgi?id=438076
	- COLORSCHEME{0..3} can now contain two numbers (dark and light).
	- Move rxvt-unicode without -256color to ALT_TERM2, since it supports
	  only 88 colors which could deserve its own color scheme.
	  For the moment, COLORSCHEME0 is used as a poor man's 88 color scheme.
	  See https://bugs.gentoo.org/show_bug.cgi?id=438076
	- Add reset-sequences before some/all (depending on RESET_ALL_LINES)
	  newlines to the defaults to fix the worst issues on terminals
	  breaking with changed background colors, see
	  https://bugs.gentoo.org/show_bug.cgi?id=445392
	- Fix color output of overlay list if no colors are used.
	- Add WIDETERM, COLUMNS, and a heuristic for the defaults.
	- Support =foo/bar-* version atoms as reported by bugmenot2:
	  https://developer.berlios.de/bugs/?func=detailbug&bug_id=18788&
	  group_id=11597

*eix-0.27.4
	Martin Väth <martin at mvath.de>:
	- Change COLOR_INST_VERSION to work with solarized. Thanks to khayyam:
	  https://forums.gentoo.org/viewtopic-p-7167646.html
	- Cosmetical fixes in the manpages
	- Add unused default variables {UPDATE_,DROP_}PRINT_IUSE
	  to formally complete *PRINT_IUSE references.
	- Use again readdir() instead of readdir_r() because of Solaris, see
	  https://developer.berlios.de/bugs/?func=detailbug&bug_id=18768&
	  group_id=11597 Thanks for alx_s for reporting.
	- Improve zsh completions, e.g. add descriptions

*eix-0.27.3
	Martin Väth <martin at mvath.de>:
	- It is now possible to use eix colors with white background terminals:
	  To use the light color schemes, set DARK=false (e.g. in /etc/eixrc);
	  or set COLORSCHEME=2 or COLORSCHEME_ALT=3 for usage with ordinary
	  or 256 colors, respectively; see the description of these variables
	  in the eix manpage. The light schemes are not carefully designed:
	  improvements or complete rework are welcome!
	- Alternatively, you can force black background colors for the default
	  scheme with BG0=black. (For the 256 color dark scheme, this is even
	  the default since there were many complaints about its unreadability;
	  use BG1=none for the previous behavior).
	  See the description of the COLORSCHEME variables in the eix manpage
	- At least the color scheme 0 should now work also with solarized.
	- Minor fixes/improvements to dark colorscheme
	- Introduce --256l{,0,1} and --256d{,0,1} (obsoletes --256f{,0,1}).
	- Introduce COLOR_{KEYEND,OVERLAYNAME,OVERLAYNAMEEND,NUMBERTEXT}
	- Extend TERM_ALT and TERM_STATUSLINE default somewhat
	- eix-installed-after: New option -F File to use mtime for reference
	- Introduce EIX_INSTALLED_AFTER to set default args; corresponding new
	  eix-installed-after option -i to ignore. Previous -i is renamed to -I
	- Avoid strcasestr(), see
	  https://bugs.gentoo.org/show_bug.cgi?id=437706

*eix-0.27.2
	Martin Väth <martin at mvath.de>:
	- Introduce %{var[,;]...} for delayed substitution to simplify colors.
	- Fix colorscheme choice for eix-diff
	- Introduce FORMAT_{BEFORE,AFTER}_COLL_EXPAND*
	- Choose scheme 0 if COLORSCHEME{,_ALT} is too high.
	- Fix/improve colorschemes
	- Internal cleanup: remove code duplication in delayed substitution.
	- Support HDEPEND (of >=portage-2.1.11.22 or >=portage-2.2.0_alpha133)
	- New database and xml formats due to HDEPEND, but keep support
	  for reading previous database format 31 of eix-0.27.1.

*eix-0.27.1
	Martin Väth <martin at mvath.de>:
	- Support 256 color terminals: COLORSCHEME, COLORSCHEME_ALT, TERM_ALT.
	  If you want to use it with screen or tmux use TERM=screen-256colors
	  Currently, only a dark scheme is available on 256 color terminals.
	  Volunteers for a light or solarized color scheme are welcome!
	  If you do not get the dark scheme, although your terminal should
	  support 256 colors or you do not get the standard scheme on a
	  non-256 color terminal, set TERM_ALT in /etc/eixrc correspondingly.
	  If you always want the standard scheme, put COLORSCHEME_ALT=0 there.
	- Introduce --ansi: reset a rather standard ansi 256 color palette.
	- Introduce --256{,f{,0,1},b}: print a 256 color ansi palette.
	- Fix --pipe with versions (regression in 0.26.3)

*eix-0.27.0
	Martin Väth <martin at mvath.de>:
	- Support USE_EXPAND. Set EIX_USE_EXPAND=false if you do not like it.
	  <{coll,}use*> and FORMAT_{BEFORE,AFTER}_IUSE_EXPAND{,_START,_END}
	- Improve handling of local cascading profiles; add local make.defaults
	- Make USE_EXPAND an accumulating variable
	- Speed up eix-remote add
	- New eix --print-overlay-data
	- Let eix --print-overlay-* honor PRINT_APPEND
	- Default to eix-sync -M if PORTDIR_CACHE_METHOD is sqlite{,*} or flat,
	  see https://bugs.gentoo.org/show_bug.cgi?id=434616
	- Fix --with-nongnu-cxx=clang if ccache is used: sets CCACHE_CPP2=true
	- Add -Werror also for LDFLAGS checks. Fixes indirectly CXXFLAGS checks
	- Introduce paranoic/static asserts and use them only for debugging
	- Use __attribute__ ((nonnull)) excessively. Dangerous optimization:
	  a bug here will often lead to segfaults; therefore a major release.
	- Fix Makefile.am for top_builddir != top_srcdir

*eix-0.26.4
	Martin Väth <martin at mvath.de>:
	- eix-remote now generates an alternative database by default.
	  To access this database, the new eix option -R has been included.
	  You can make -R the default by setting REMOTE_DEFAULT=true.
	  With the new default, it is a good idea to put the line
	  "@@eix-sync add" to your /etc/eix-sync.conf
	  Set EIX_REMOTE="" to keep the old behavior with only one database.
	- Introduce search options --nonvirtual, --virtual and attributes
	  {virtual} {havevirtual} {havenonvirtual} and in xml: virtual=1
	- Make configure cross-compile friendly, see
	  https://bugs.gentoo.org/show_bug.cgi?id=432122
	- Fix --without-sqlite compile regression, see
	  https://bugs.gentoo.org/show_bug.cgi?id=432238
	- include config.h in all non-headers, and only there.
	  This should fix https://bugs.gentoo.org/show_bug.cgi?id=432478

*eix-0.26.3
	Martin Väth <martin at mvath.de>:
	- Follow most of Google's C++ style guidelines:
	  eix now avoids using static or global classes. It should at least
	  pass src/cpplint.sh (i.e. cpplint.py up to minor whitespace
	  conventions). Exceptions (false positives) are marked with NOLINT.
	  This is actually a major rewrite which makes many things cleaner.
	  Thanks a lot to Michał Bartoszkiewicz <mbartoszkiewicz at gmail.com>
	  for analyzing a hard to find bug introduced by changing types.
	- Fix regression from eix-0.26.1 of wrong <availableversions:VAR>
	  (used e.g. with --versionlines in default format).
	- Use cache of true filenames instead of heuristic recursion bounds.
	- Improve varsreader: Treat ;|&) as additional end symbols (like shell)
	- Fix frequent warnings (mainly with pragmas).
	- Fix mkstemp name and a possible memory leak.
	- Allow --disable-silent-rules to follow Gentoo's policy, although
	  personally, I dislike it, and it is not necessary for eix.

*eix-0.26.1
	Martin Väth <martin at mvath.de>:
	- eix-test-obsolete: Add -q, "quick", use FORMAT_TEST_OBSOLETE,
	  support TEST_FOR_REDUNDANCY better
	- Fix masked-packages: Treat repository name in arguments correctly.
	- masked-packages: Treat - as standard input.
	- Fix USE=tools linking error (regression of eix-0.26.0), see
	  https://bugs.gentoo.org/show_bug.cgi?id=425512
	- Allow KEYWORDS='*' and KEYWORDS='~*' in ebuilds
	- Support profile-format=portage-2, see
	  https://bugs.gentoo.org/show_bug.cgi?id=426256
	- Introduce DEFAULT_ARCH and ./configure --with-default-arch
	- Use function pointers instead of case cascades
	  for FORMAT execution and for file parsing.
	  This is cleaner and might speed up things.
	- manpage: improve description of masked-packages
	- varsreader: Fix hang for DOS line ends, see
	  https://forums.gentoo.org/viewtopic-t-929536.html
	- Do not let (...) automagically choose the dep search, see
	  https://bugs.gentoo.org/show_bug.cgi?id=427644

*eix-0.26.0
	Артём Воротников <skybon at gmail.com>:
	- Continued Russian translation and add fixes suggested by
	  Stanislav Cymbalov <dummyunit at gmail.com>.

	Martin Väth <martin at mvath.de>:
	- Add masked-packages
	- Introduce versionsort -[npvr]
	- Introduce plainversion and revision attributes
	- Introduce --pipe-name and --pipe-version
	- Support for subslots
	- Do not use C++ exceptions. This is a major internal change.
	- Improve error handling of version parsing
	- Remove NEWLINE
	- Extend varsreader.cc to allow VAR=\n and (for portage escapes)
	  VAR[\t ]*=[\t ]*value https://bugs.gentoo.org/show_bug.cgi?id=417387
	- support --with-nongnu-cxx=clang++
	- separate eix-drop-permissions binary with --enable-separate-tools
	- Improve slot finding
	- Do not cache version strings to save memory

*eix-0.25.5
	Артём Воротников <skybon at gmail.com>:
	- Russian translation. (Missing yet: eix --dump and most of manpage)

	Martin Väth <martin at mvath.de>:
	- Depend on https://github.com/vaeth/push/ instead of implementing
	  a poor substitute.
	- Add German manpage
	- Improve file error handling. Make some errors fatal.
	- scripts: More compatible usage of cd/mkdir, speed up tests for -z/-n
	- eix-functions.sh: Make eval_gettext safe
	- Use waitpid() in a more compatible manner
	- No longer support gnu++0x or c++0x which breaks with gcc-4.4, see
	  https://bugs.gentoo.org/show_bug.cgi?id=414979

*eix-0.25.4
	Martin Väth <martin at mvath.de>:
	- Introduce -y (--any) to select any match field: don't miss anything.
	- Default format: new properties/restrictions separator: */^
	  Also add $sep (a tab) after version with --versionlines
	  Remove duplicate FORMAT_AFTER_IUSE (a long standing regression)
	- Fix eix-installed-after (regression introduced in eix-0.24.0)
	- Set umask 002 in eix-update unless option -o is used, see
	  https://bugs.gentoo.org/show_bug.cgi?id=408509
	- Try to fix permissions on interix, see
	  https://bugs.gentoo.org/show_bug.cgi?id=410205
	  Thanks to Gordon Cowie <gcowie at live.com> and
	  Fabian Groffen <grobian at gentoo.org>
	- Rely on working delete 0; should speed up things slightly
	- Avoid std::auto_ptr
	- Default to most current C++ dialect and use nullptr if possible
	- Add some more ATTRIBUTE_PURE
	- configure.ac: Fix usage of MV_APPEND
	- scripts: Simplify quoting slightly

*eix-0.25.3
	Martin Väth <martin at mvath.de>:
	- Introduce eix-drop-permissions and use it for eix-remote
	  instead of su/sudo hacks.
	- Fix regressions: eix-remote add|fetch without temporary file,
	  and eix-remote update with temporary file.

*eix-0.25.2
	Martin Väth <martin at mvath.de>:
	- Introduce EIX_PREVIOUS as default for eix-diff; remove eix-sync -d.
	- Introduce EIX_REMOTEARCHIVE; remove filename handling of eix-remote.
	- eix-remote: Pass more variables to new user.
	- Update docu to new paths and magic string (forgotten in eix-0.25.1)

*eix-0.25.1
	Martin Väth <martin at mvath.de>:
	- Drop all permission stuff: Only become EIX_{USER,GROUP,UID,GID} ASAP.
	  This fixes several security issues and simplifies code enormously.
	- Let DEP work with sqlite cache method (bug in eix-0.25.0).
	- Update zsh completion (including missing changes from eix-0.25.0).
	- Move all cache files into /var/cache/eix/
	- Start cachefile with magic string "eix\n"
	- Introduce EXIT_STATUSLINE, see the discussion in
	  https://bugs.gentoo.org/show_bug.cgi?id=407473

*eix-0.25.0
	Martin Väth <martin at mvath.de>:
	- Add DEP (and ./configure option --with-dep-default for the default)
	  to support DEPEND, RDEPEND, PDEPEND ({r,p,}depend{,*} properties).
	  By new default, the output happens with "eix -lv" (or "eix -lvx")
	  if DEP=true and the cachefile was generated with DEP=true.
	  Set VERSION_DEPS_NORMAL=true if you want output also with "eix -l".
	  Usage of DEP=true roughly doubles disk resp. memory requirements.
	- New match-fields --depend --rdepend --pdepend (and --deps for
	  all three). This requires DEP=true (also for cachefile generation).
	- Output dependencies also to xml (if DEP=true), bump xml version.
	- Fix xml-schema (e.g. mask can be repeated).
	- Change defaults (VERSION_KEYWORDS_NORMAL=false, PRINT_BUGS=true).
	- Put keywords/effective keywords handling to formatstring level:
	  New properties versionkeywords*, versionekeywords;
	  remove internal PRINT_{BEFORE,AFTER}{,_EFFECTIVE}_KEYWORDS.
	- Rework default FORMAT: Split into more variables and handle case
	  distinction compact/verbose/normal by runtime variable "$modus".
	  Give runtime variables saner names. Improve naming of some
	  configuration variables and descriptions.
	- Deal with version output differently in verbose mode.
	- Several bugfixes/improvements in German translation.
	- Remove --with-bzip2 (enabling /var/db/pkg/CAT/NAME/environment.bz2 if
	  /var/db/pkg/CAT/NAME/repository is unavailable). This was not used in
	  the ebuild from the portage tree anyway and simplifies internal logic
	- Remove haveversionuse and NOT_FULL_USE to shorten database format.

*eix-0.24.0
	Martin Väth <martin at mvath.de>:
	- Remove support for obsolete old-style virtuals.
	  This is a major change: Remove many search options, print-commands,
	  many default variables, and changes the database format (even xml).
	- scripts: Make PushEscaped work with more shells (fixes eix-layman).
	  Note that this causes a change in the optional last argument.
	- Remove OBSOLETE_MINUSASTERISK and RED_MINUSASTERISK
	- Remove cache method cdb
	- Remove --enable-obsolete-symlinks --enable-obsolete-reminder
	- Remove --enable-extra-cache
	- Support TERM_SOFTSTATUSLINE for screen.
	- Do not add -D* flags already contained in CPPFLAGS.
	- scripts: Simplify some quoting in assignment and case

*eix-0.23.10
	Martin Väth <martin at mvath.de>:
	- eix-sync: Make layman and hook errors non-fatal, by default.
	  Use -F, FATAL_HOOKS, or the die function to revert old behavior.
	- Suppress status line if TERM does not begin with TERM_STATUSLINE,
	  see the discussion in https://bugs.gentoo.org/show_bug.cgi?id=402993

*eix-0.23.9
*eix-0.23.8
	Martin Väth <martin at mvath.de>:
	- Provide status line also for eix-update
	- Do not update status line if output is not a terminal
	- Introduce {NO,FORCE_}{COLORS,STATUSLINE,PERCENTAGE}.
	  The previous FORCE_USECOLORS and new NOCOLOR default them all.
	- eix-remote: Change ownership before calling wget, see
	  https://developer.berlios.de/bugs/?func=detailbug&bug_id=18493&
	  group_id=11597
	- Make correct fix for https://bugs.gentoo.org/show_bug.cgi?id=399451.
	  Thank Arfrever Frehtes Taifersar Arahesis <arfrever.fta at gmail.com>

*eix-0.23.7
	Martin Väth <martin at mvath.de>:
	- Fix MetadataCache::use_prefixport(), see:
	  https://bugs.gentoo.org/show_bug.cgi?id=399233
	- Add {tbz2} also for default compact output
	- Fix typo in eix-sync: https://bugs.gentoo.org/show_bug.cgi?id=399117
	- Fix and document behavior if path is specified for cache md5-or-*
	- Remove UTF-8 symbol from my name in some docs to allow building with
	  docutils + python-3.2: https://bugs.gentoo.org/show_bug.cgi?id=399451

*eix-0.23.6
	Martin Väth <martin at mvath.de>:
	- Scripts: Support for status line update; minor output improvements
	- Introduce new cache method metadata-md5-or-flat and make it default
	  for PORTDIR_CACHE_METHOD:
	  https://developer.berlios.de/feature/?func=detailfeature&
	  feature_id=5483&group_id=11597
	- Include unistd.h for isatty(). Thanks to <gokturk at binghamton.edu>,
	  see https://bugs.gentoo.org/show_bug.cgi?id=397589

*eix-0.23.5
	Martin Väth <martin at mvath.de>:
	- scripts: Be less rigorous with set -f
	  This fixes some regressions of eix-0.23.4 (e.g. eix-remote).

*eix-0.23.4
	Martin Väth <martin at mvath.de>:
	- scripts: Fix possible globbing problems
	- varsreader.cc: Improve error output for failed "source".
	  Thanks to Sergei Trofimovich <slyfox at gentoo.org> for the patch
	- autotools: Require automake-1.11.2 and disable dependency tracking
	  by default.
	- No unmap for empty md5 files
	- Fix minor include issues

*eix-0.23.3
	Martin Väth <martin at mvath.de>:
	- Now --print exits with 1 if variable is not known
	- Introduce --known-vars and use it with zsh-completion
	- Major rewrite of varsreader.cc: Allow concatenation of strings
	- scripts: Remove portageq usage completely (for runtime)

*eix-0.23.2
	Martin Väth <martin at mvath.de>:
	- New cache method metadata-md5 to support metadata/md5-cache
	  in overlays
	- eix-update: Add option -v (and UPDATE_VERBOSE) for printing
	  effective cache method for every ebuild
	- scripts: Handle quoting of spaces in PORTDIR_OVERLAY properly
	- scripts: Do not rely on portageq being available
	- eix-functions.sh: Add PushEscaped, InitPortdirOverlay

*eix-0.23.1
	Martin Väth <martin at mvath.de>:
	- Support basic /etc/portage/package.license checking
	- Add rudimentary bash-completion of
	  Marcel Philipp Partap <marcel.partap at rwth-aachen.de>, see
	  https://bugs.gentoo.org/show_bug.cgi?id=304819
	- Avoid install -D (does not work on BSD), see
	  https://bugs.gentoo.org/show_bug.cgi?id=382539

*eix-0.23.0
	Martin Väth <martin at mvath.de>:
	- Make profiles/* files default only on their repositories, see
	  https://bugs.gentoo.org/show_bug.cgi?id=380979
	  This required a major code reorganization
	- Bugfix: Let -... in /etc/portage/package.accept_keywords remove
	  previous keywords of /etc/portage/package.accept_keywords
	- Use a function-call for testing *FLAGS (e.g. -fstack-protector), see
	  https://bugs.gentoo.org/show_bug.cgi?id=379717
	  Thanks to Dustin Polke for helping with debugging this issue
	- Cosmetic code changes

*eix-0.22.11
	Martin Väth <martin at mvath.de>:
	- Bugfix: Let -... in /etc/portage/package.accept_keywords remove
	  ACCEPT_KEYWORDS, see https://bugs.gentoo.org/show_bug.cgi?id=378767
	- eix-test-obsolete: Set DEFAULT_FORMAT=compact instead using -c
	  (chooses compact format absolutely, not toggling...), see https://
	  developer.berlios.de/bugs/?func=detailbug&bug_id=18284&group_id=11597

*eix-0.22.10
	Martin Väth <martin at mvath.de>:
	- Add -Wl,-z,noexecstack with strong-optimization and security:
	  gcc-4.6 -flto turns on execstack which is probably a bug, see
	  https://bugs.gentoo.org/show_bug.cgi?id=369795
	- In case of -flto add CXXFLAGS to LDFLAGS unconditionally, see
	  https://bugs.gentoo.org/show_bug.cgi?id=374217
	- scripts: Use readlink only if it supports -f

*eix-0.22.9
	Martin Väth <martin at mvath.de>:
	- Support -~* keywords: Introduce {isminusunstable}, {wasminusunstable}
	- Support alternative /etc/portage/make.profile location, see
	  https://bugs.gentoo.org/show_bug.cgi?id=358647
	- Support layman-2
	- Fix handling of non-matching /etc/portage/package.accept_keywords:
	  https://bugs.gentoo.org/show_bug.cgi?id=370799
	- Make eix-test-obsolete more intelligent concerning skipping of
	  checks, see https://bugs.gentoo.org/show_bug.cgi?id=358861
	- Fix eix-test-obsolete's support for REDUNDANT_IF_IN_CFLAGS.
	  Thanks to Michał Bartoszkiewicz <mbartoszkiewicz at gmail.com>,
	  see https://bugs.gentoo.org/show_bug.cgi?id=372921
	- Provide poor man's einfo, ewarn, eerror, ebegin, eend to avoid
	  dependency on openrc: https://bugs.gentoo.org/show_bug.cgi?id=372887
	- Introduce ROOT_UID for usage with prefix, see
	  https://bugs.gentoo.org/show_bug.cgi?id=364351
	  Thanks to Jacob Godserv at <jacobgodserv at gmail.com> for the patch.
	- Add several ATTRIBUTE_PURE
	- Treat provide as a list of strings, see
	  https://bugs.gentoo.org/show_bug.cgi?id=358729
	- Scripts: Avoid test -e for compatibility reasons
	- Do not add -flto on gcc-4.6.0 as it seems broken there, see
	  https://bugs.gentoo.org/show_bug.cgi?id=364317
	- Make -flto tests more reliable
	- Add CXXFLAGS to LDFLAGS in case of -flto
	- Use AS_VAR_COPY instead of AS_VAR_SET if possible
	- Use -fno-common -fnothrow-opt -flto-partition=none and new -W...

*eix-0.22.8
	Martin Väth <martin at mvath.de>:
	- Fix non-casting of rend() -> reverse_const_iterator on some systems,
	  see https://bugs.gentoo.org/show_bug.cgi?id=354071
	- Do not rely on argreader.end() not to change, even if argreader is
	  not accessed in a writable way. Thanks to
	  Tom Lloyd <napalmllama at gmail.com> for letting me debug on his
	  machine, see https://bugs.gentoo.org/show_bug.cgi?id=355623
	- Set always default search field, even if string is unspecified
	- Introduce ACCEPT_KEYWORDS_AS_ARCH=full and make it default, see
	  https://bugs.gentoo.org/show_bug.cgi?id=353562
	- Also add ./configure --with-always-accept-keywords
	- Remove --with-rst and ship *.html in tarball, see
	  https://bugs.gentoo.org/show_bug.cgi?id=353705
	- Add --with-extra-doc and install by default nothing from doc

*eix-0.22.7
	Martin Väth <martin at mvath.de>:
	- Add option -f to eix-installed-after
	- Fix segfault for tests beginning with --end, see
	  https://bugs.gentoo.org/show_bug.cgi?id=350321
	- Minor fixes in zsh-completion
	- Avoid casting away const to be compatible with future C++ standard;
	  this goes at the cost of making many arguments non-const Package *

*eix-0.22.6
	Martin Väth <martin at mvath.de>:
	- Add some experimental zsh-completion
	- eix-sync: New option -V, cancel -m, fix description of -u
	- eix-test-obsolete: Output {,accept_}keywords instead of keywords:
	  https://bugs.gentoo.org/show_bug.cgi?id=349352
	- Honour order of {,accept_}keywords
	- Use largefile support if available
	- Fix minor variable and manpage description issues
	- No longer support translations for true/yes/on in variable names
	- Make code c++0x clean

*eix-0.22.5
	Martin Väth <martin at mvath.de>:
	- Fix mask syntax if slots and repositories are used in the same mask,
	  see https://forums.gentoo.org/viewtopic-p-6479886.html
	- Fix NONEXISTENT_IF_MASKED (check not only keywords but also mask)
	- manpage: Explain logic of EXPRESSION more verbosely.
	- Divide memory usage of global*.cc during compiling more even.

*eix-0.22.4
	Martin Väth <martin at mvath.de>:
	- Introduce eix-installed.
	- Rewrite eix-test-obsolete almost completely.
	- Do not require SA_SIGINFO (thanks Markus Duft <mduft at gentoo.org>)
	  https://bugs.gentoo.org/show_bug.cgi?id=343071
	- Add a forgotten translation to eix-installed-after

*eix-0.22.3
	Martin Väth <martin at mvath.de>:
	- Support repository names in config files:
	  https://bugs.gentoo.org/show_bug.cgi?id=340098
	  The current implementation stores reponames in every version and
	  thus increases memory usage, but this is perhaps cleaner than to tie
	  the names to the database headers.
	- Convert manpage a-diaresis back to groff format (no utf8)

*eix-0.22.2
	Martin Väth <martin at mvath.de>:
	- Improve error handling when reading profile masks
	- Fix sqlite.cc: https://bugs.gentoo.org/show_bug.cgi?id=336896
	- Fix sqlite automagic

*eix-0.22.1
	Martin Väth <martin at mvath.de>:
	- use /etc/portage/package.nowarn instead of
	  /etc/portage/package.*.nowarn; the latter is now obsolete.
	  If you want continue to use it, set OBSOLETE_NOWARN=true
	- Support wildcards in /etc/portage/package.nowarn
	- Ignore CVS/RCS/SCCS dirs in recursive dir reading
	- Support /etc/portage/make.conf

*eix-0.22.0
	Martin Väth <martin at mvath.de>:
	- Support wildcards in config-files.
	  This required a rewrite of the whole mask handling: Even if no
	  wildcards are used, a crucial change in the data structure is needed
	  (one can no longer use only a sorted list of names for lookup)
	  which has some overhead in memory and time requirements.
	- Switch off REDUNDANT_IF_{IN,DOUBLE}_{USE,ENV,CFLAGS} by default:
	  Usually, double entries are intentional here, and most people will
	  want to leave some settings after deinstallation for later usage.
	- Put EIX_TEST_OBSOLETE_OPTS=-d since most people will want it.
	- Security-fix: Some auto-destruction in std::vector-elements was
	  removed (this would also have led to segfaults in bad cases).
	- Fix: Handle sets in local {un}masks and for -t

*eix-0.21.0
	Martin Väth <martin at mvath.de>:
	- Support for virtuals in @system and worldfile:
	  --world* --selected* --system* include now by default all virtuals.
	  Use --world-pure etc. to avoid inclusion of virtuals.
	  Similarly for {world} {world_set} {system} which have now
	  {worldpure} {world_setpure} {systempure} counterparts.
	- Support for sets in profile masks/keywords
	- Support profile/package.accept_keywords
	- Support /etc/portage/package.accept_keywords
	- Support escaping of "-" on the commandline by using "---" instead.

*eix-0.20.7
	Martin Väth <martin at mvath.de>:
	- Add new demo script (eix-installed-after) which outputs packages
	  installed after/before a given package.
	- Choose color defaults more similar to portage
	- Redundancy checks for /etc/portage/package.env
	- scripts: Provide cd -P fallback if there is no readlink command
	- scripts: Improve emulated mktemp
	- scripts: Make functions definitions more standard (no space)
	- Use -fgraphite-identity of gcc-4.5
	- Improve treatment of __attribute__
	- Carefully begin to use __const__, __pure__, and __signal__
	- Use sigaction() where available

*eix-0.20.6
	Martin Väth <martin at mvath.de>:
	- Fix wrong test for masks ...*:SLOT (regression from eix-0.19.0), see
	  https://bugs.gentoo.org/show_bug.cgi?id=326143

*eix-0.20.5
	Martin Väth <martin at mvath.de>:
	- Interpret escapes for portage config files in the way portage does,
	  see https://bugs.gentoo.org/show_bug.cgi?id=316907
	  Thanks to Michał Górny <gentoo at mgorny.alt.pl> for the main part
	  of the patch.
	- Fix lots of typos in German translation. Thanks to
	  Boris Petersen <transacid at gmail.com>
	- scripts: Avoid race conditions in traps.
	- scripts: Avoid "local" keyword as it is not POSIX.
	- configure: Treat sqlite with PKG_CHECK_MODULES if possible.

*eix-0.20.4
	Martin Väth <martin at mvath.de>:
	- Support RESTRICT=parallel.
	- Support RESTRICT=primaryuri in xml output.
	- Ignore hidden files/dirs in local set dirs like /etc/portage/sets
	  See https://bugs.gentoo.org/show_bug.cgi?id=315611
	- Fix linking of standalone versionsort with gettext, see
	  https://bugs.gentoo.org/show_bug.cgi?id=313267
	  Thanks a lot to Javier Villavicencio <the_paya at gentoo.org>
	  for examining the cause. This should also fix
	  https://bugs.gentoo.org/show_bug.cgi?id=305737
	- Transfer manpage to utf8 to make it work with heirloom-doctools.
	  This makes the manpage work only with .utf8 locales. If somebody
	  knows a better fix for https://bugs.gentoo.org/show_bug.cgi?id=294402
	  please let me know.
	- Use -flto and -fwhole-program of gcc-4.5
	- Improve automagic in warning flags (test without -Werror);
	  this could prevent previously e.g. -Wall from being auto-added.

*eix-0.20.3
	Martin Väth <martin at mvath.de>:
	- Output in --xml up to two mask_type/unmask_type:
	  One for the keyword type and one for the hard/profile mask:
	  https://sourceforge.net/tracker/?
	  func=detail&atid=710608&aid=2968195&group_id=128101
	- eix-remote/eix-layman: work around bug of newer layman or python
	  which misinterprets python -c ... as option -c for layman, see
	  https://bugs.gentoo.org/show_bug.cgi?id=309685
	- eix-remote/eix-layman: support >=python-3
	- manpage: Fix indentation

*eix-0.20.2
	Martin Väth <martin at mvath.de>:
	- Move hosting to BerliOS as a protest on the influence of the
	  American Government against the freedom of projects on Sourceforge.
	- Support for BUILD_TIME (>=portage-2.2_rc63):
	  Introduce USE_BUILD_TIME and let it default to "true".
	- Quote in --dump and --dump-defaults: It is now guaranteed that the
	  output can be reread.
	- Fix segfault of <markedversions:*> without pipe argument, see
	  https://bugs.gentoo.org/show_bug.cgi?id=307479
	- Extend xml output by <licenses> and <provide>:
	  https://sourceforge.net/tracker/?
	  func=detail&aid=2965496&group_id=128101&atid=710611
	- Introduce eix-test-obsolete [no-]buildtime
	- Being unable to fix USE=nls on OS X,
	  https://bugs.gentoo.org/show_bug.cgi?id=305737
	  Thanks to Stuart Shelton <srcshelton at gmail.com> for testing.
	  If somebody has an idea, please write to that bug.

*eix-0.20.1
	Martin Väth <martin at mvath.de>:
	- Fix magic adding of -a / -o between successive tests
	  (regression introduced in 0.20.0).
	- Introduce NOFOUND_STATUS, MOREFOUND_STATUS to make eix -q useful:
	  https://bugs.gentoo.org/show_bug.cgi?id=304705
	  Best combined with:
	- Introduce --brief, --brief2 (mainly to make eix -q faster);
	  best combined with COUNT_ONLY_PRINTED=false
	- varsreader does no longer introduce accessed undefined variables:
	  There might be an advantage in speed and memory, but mainly it is
	  cleaner, avoiding e.g. https://bugs.gentoo.org/show_bug.cgi?id=299904
	- Minor compatibility fixes.

*eix-0.20.0
	Martin Väth <martin at mvath.de>:
	- Introduce -( ... -) command line parsing.
	- Be aware if you rely in scripts on logical connectives -a -o -!
	  or the previous strange side effects of --pipe
	  This has dramatically changed now (therefore the version jump):
	- Treat -a and -o left-associative (of equal precedence as previously).
	- Treat -! as unary operator (not as a TEST_OPTION).
	- Treat --pipe as a TEST_OPTION.
	- Unify shell-script style.
	- Fix missing includes and type ambiguity on IRIX
	  https://bugs.gentoo.org/show_bug.cgi?id=302626
	  Thanks to Stuart Shelton <srcshelton at gmail.com>

*eix-0.19.2
	Martin Väth <martin at mvath.de>:
	- eix-functions.sh: Replace: returns true/false, uses $replace_count
	- eix-functions.sh: New function ReadVar as --print wrapper
	- Introduce PRINT_APPEND and let it default to \n. Hence:
	- --print now outputs a trailing newline by default!
	  Be aware that this might break some scripts which trickily
	  attempted to preserve trailing spaces in variables.
	  Set PRINT_APPEND='' in /etc/eixrc to revert to the old behaviour
	  or better patch the scripts to use ReadVar.
	- Fix description of @/@@-hook order in manpage and help page.
	  Fix corresponding eix-remote example in manpage.
	  Thanks to navid.zamani at googlemail.com
	  https://bugs.gentoo.org/show_bug.cgi?id=296722
	- eix-sync: Add option -L for passing options to layman,
	  see https://sourceforge.net/tracker/index.php?
	  func=detail&aid=2921161&group_id=128101&atid=710611
	- "Export" EPREFIX when sourcing make.{globals,conf} so that e.g.
	  $EPREFIX can be used in etc/make.conf
	- Fix incremental keys hack in varsreader.cc at the cost of speed:
	  Variables now can be exported for sourcing with APPEND_VALUES.
	  In particular, they are not clean(ed) when accessed which caused
	  https://bugs.gentoo.org/show_bug.cgi?id=299904

*eix-0.19.1
	Martin Väth <martin at mvath.de>:
	- Let eix-remote and KEEP_VIRTUAL=true use the original
	  overlay label's (repo_name)
	- Introduce --repo-name and REPO_NAMES
	- eix-update: Treat eix-caches for the same file simultaneously.
	  This dramatically speeds up KEEP_VIRTUALS=true
	  When the first overlay accessing the file is displayed, actually
	  all overlays are handled - the later ones only print a message.
	- eix --pipe: Also accept leading "=", e.g. =sys-devel/gcc-4.4.2.
	  Also accept several packages in one line (separated by spaces).
	- Provide xml-schema for the --xml output: eix-xml.xsd
	- Rename xml-format.* and format.* into eix-xml.* and eix-db.*
	- manpage: Give some useful examples at the very beginning.
	- Fix xml-description (add description, homepage, missing_keyword)
	- Fix ./configure --disable-security (regression of 0.19.0,
	  was fixed by patch in eix-0.19.0-r1).
	- Improve tests in case of --enable-security.

*eix-0.19.0
	Martin Väth <martin at mvath.de>:
	- This is a major rewrite (although hopefully not visible to the user;
	  perhaps it is slightly faster which however is hardly visible, since
	  the main bulk is the I/O governed by the kernel).
	     Visible changes:
	- Rename options:
	    --world     -> --selected-file
	    --world-set -> --selected-set
	    --world-all -> --selected
	  Introduce new options --world, --world-file, --world-set (yes, the
	  former two have now a different meaning than in <eix-0.19.0) which
	  behave like the corresponding --selected* but include @system.
	  The intention is to make --world/--selected behave consistently with
	  @world/@selected of portage-2.2_rc47, i.e. world_sets are included
	  for both, and @system is included in @world.
	  The new systematic is: Unless you specify *-file/*-set, both are
	  admissible, and --world* includes @system, but --selected* does not.
	- New cache method sqlite* (analogous to eix*).
	- eix-update: improve percentage output.
	- Allow --nocolor/--force-colors also for eix-update (although this is
	  a misnomer, since it is about the percentage status; the variables
	  stay separated with appropriate names: FORCE_PERCENTAGE/FORCE_COLORS)
	- Fix name output on --help.
	- Fix: KEEP_VIRTUALS=true should now quote ":" correctly, see
	  https://bugs.gentoo.org/show_bug.cgi?id=294773
	- Fix: Now --world-file --world really behaves like --world-file as it
	  should be according to the logic.
	- Fix: Check effective uid instead of uid.
	- configure: Be stricter with testing flags and enable less defaults:
	  https://bugs.gentoo.org/show_bug.cgi?id=293263
	- configure: More verbose output concerning options for flag mangling
	- configure: Remove libtool again which was only added for libintl
	  support, since I guess this is not needed when gettext is installed
	  on non-GNU systems. Please, file a bug if my guess is wrong and you
	  are suddenly without the German translations on some system.
	     The rest are internal changes of the code:
	- Change of #include style: Do not rely on implicit includes;
	  include in the file exactly those things needed, ignoring that
	  include files might include other files.
	  This slows down compilation but is cleaner w.r.t. dependencies.
	- Use likely()/unlikely() inflationary to hopefully improve
	  branch prediction. If you provide code make sure that
	  likely()/unlikely() is only used with Boolean arguments
	  (although gcc supports also int arguments, but do not rely on this).
	- Make use of attribute unused/noreturn if supported.
	- Avoid usage of std::unique() since usually it is better to use a
	  set anyway. Moreover, since it is broken on some systems, the
	  configure test can now be avoided.
	- Other style changes: Prefer constructor argument over assignment,
	  because it *may* be faster on some systems.
	- Use clean comparison with NULL instead of Boolean test.
	  This might be slightly slower on badly optimizing compilers but is
	  cleaner and can be used with likely()/unlikely().
	- Fix memory leak with percentage output
	- Use autoconf-2.64; set PACKAGE_URL. Remove some obsolete macros.
	- autoconf: Use *_ONCE() where possible. Use AC_PROG_SED().
	- In Scripts: Avoid incompatible usage of !
	     "Hacks" for speed increase:
	- Use string::string/assign/append with args instead of substr()
	  where possible.
	- Make BasicPart a class of its own with sane constructors, allowing
	  avoidance of substr().
	- Pass return by reference in split_string(). Actually, split_string()
	  (and similarly resolve_plus_minus()) do not clear() the results now
	  which is not only faster by itself but also sometimes avoids extra
	  calls of push_backs() for the result.
	- Similarly, let also join_to_string act on references and not clear()
	- Make packagetree a std::map of categories; this also avoids the ugly
	  fast_access()-hacks which Emil hates so much. Another advantage is
	  that now only this tree is needed for BasicCache::readCategories().
	  Since the category names are not stored redundantly, they must be
	  passed separated to BasicCache::readCategoryPrepare() which is now
	  used anyway to avoid unnecessary percentage updates.
	- Use switch()es instead of subsequent tests in eixrc.cc
	- Add only finished packages to a Category instead of hackishly
	  erasing afterwards if unfinished.
	- In Scripts: Use ":" instead of "true"

*eix-0.18.3
	Martin Väth <martin at mvath.de>:
	- Add ./configure --with-remote-file=STR
	- Modify CXXFLAGS/LDFLAGS. The default can be modified by:
	  --enable-debugging --enable-strong-optimization --enable-warnings
	  --disable-optimization --disable-security --enable-strong-warnings
	- Make ./configure slightly be more verbose
	- Code cleanup (nonconst->const/mutable, prefer prefix ++/--, headers,
	  use localeC cctype, configure.in->configure.ac, mvmacros)
	- Fix WEAKER detection in -T/eix-test-obsolete which was apparently
	  broken since many versions...
	- Fix: Do not use obsoleted {markedversions} in the defaults.
	- Introduce {havemarkedversion}.
	- Fix some segfaults when unknown attributes are used by mistake.
	- Fix linking versionsort
	  https://bugs.gentoo.org/show_bug.cgi?id=288514
	- Let *.tbz2 recognition also work with symlinks (needed for older
	  portage versions): https://bugs.gentoo.org/show_bug.cgi?id=288031

*eix-0.18.2
	Martin Väth <martin at mvath.de>:
	- PLEASE NOTICE: The adding of +- data to USE-Flags to the cachefile
	  in eix-0.18.1 came without a version format jump to not force you to
	  regenerate it. Of course, you will only get the new information if
	  you regenerate the cachefile with eix-update/eix-sync. If you use
	  KEEP_VIRTUALS=yes (or configured eix to use cache method eix::...)
	  this information is of course not magically upgraded either.
	  For KEEP_VIRTUALS this means typically that the information is only
	  upgraded after the next (successfull) run of eix-remote.
	- Introduce {slots}, {slotted}.
	- Introduce NEWLINE to avoid printing the magic newline after a
	  package, and even let it default to false (the default formats are
	  now adapted to work with both setting of NEWLINE; but if you use
	  your own formats and do not want to modify these, you will want to
	  set NEWLINE=true in /etc/eixrc to get the previous behavior).
	  The new default NEWLINE=false should make the life simpler for
	  authors of complex format strings who now do not have to care
	  about avoiding to print the final newline: I recommend to use
	  NEWLINE=true only as a temporary workaround until you upgrade your
	  formatstrings. Maybe NEWLINE will be passed out in some future
	  version of eix (when this happens, the behaviour is certainly that
	  of NEWLINE=false).
	- Introduce DATESORT example.
	- NAMESLOT now prints the slot only if there are at least two slots.
	  The old behaviour of printing the slot always is now available
	  as NAMEASLOT.
	- Feature Request https://bugs.gentoo.org/show_bug.cgi?id=288031
	  Introduce --binary, {binary}, {isbinary} and use it in defaults.
	- Introduce XML_DATE.
	- Fix missing initialization leading to "unknown property ''" errors.
	  A big thanks to Michał Bartoszkiewicz <mbartoszkiewicz at gmail.com>
	  https://bugs.gentoo.org/show_bug.cgi?id=287186
	- Fix some false error messages about variables calling themselves.
	  Thanks again to Michał Bartoszkiewicz <mbartoszkiewicz at gmail.com>
	  for analyzing this terrible spaghetti-recursion code.
	  https://bugs.gentoo.org/show_bug.cgi?id=287288
	- Fix bestslotupgradeversions{,*} (in 0.18.1 it worked like
	  bestslotversions{,*}).
	- Fix IUSE storing in eix-update (0.18.1 stored false empty entries);
	  this causes --xml to falsely output an additional leading space in
	  the output of the first useflag tag of each version.

*eix-0.18.1
	Martin Väth <martin at mvath.de>:
	- Allow also properties and variable names on the right-hand side
	  of comparisons or runtime variable assignments.
	- Store/show +- of IUSE.
	- Change iuse/use output for --xml to reflect the new available data.
	  Thanks to Bob Shaffer II <bob.shaffer.2 at gmail.com>
	- Document the XML output format.
	- Introduce {havecolliuse}
	- Fix defaults in formatstrings (particularly for eix-diff).
	- Fix manpage: Add some forgotten properties.
	- Fix possible overflow in string assignment/comparison.
	- Improve parsing of <*:*>.
	- Fix implementation of -q which was broken in a subtle way.
	- Use fseek/ftell on systems without fseeko/ftello, and
	- improve error messages for failed permission tests,
	  for both see https://bugs.gentoo.org/show_bug.cgi?id=286229
	- ./configure: --with-null-device, get default EPREFIX by portageq.
	- Fix bug in variable parsing for --xml. Thanks to
	  Bob Shaffer II <bob.shaffer.2 at gmail.com>

*eix-0.18.0
	Martin Väth <martin at mvath.de>:
	- XML output was implemented by
	  Bob Shaffer II <bob.shaffer.2 at gmail.com>
	  Thanks for the contribution.
	- Introduce runtime variables in format-string:
	  Now the format-string is almost a programming language;
	  without loops or recursion, but one can loop over versions:
	- Finally, dump the mess with available*/best*/installed*/empty*/full*
	  and the distinction of printable and "test-only" properties.
	  Instead of the flood of variations of properties for version output
	  there are now essentially only two properties with slight variations:
	  <availableversions:VAR> <installedversions:VAR>
	  They interpret the variable VAR for each version as a format.
	  Special properties like {last} are available in VAR to ease writing
	  separators between versions or in special cases.
	  Special care has been taken to parse VAR only once so that
	  eix should not be dramatically slower after this change.
	  Of course, a slight loss in the output speed was unavoidable, but
	  this should be worth the gained flexibility and especially the
	  simplification of writing formats.
	- For the previous empty*/full* use cases there are now example
	  variables NAMEVERSIONS, EQNAMEVERSION, NAMESLOT, ANAMEASLOT,
	  and ANAMESLOT; see eix --dump for further details on these.
	  To see the effect of such an example try e.g.
	  eix --format '<availableversions:ANAMESLOT:ANAMESLOT>' '-*' gcc
	- Let = in conditionals work as documented (a long outstanding bug).
	- The --versionlines cases are now completely shifted to the formats.
	- Many variables like PRINT_IUSE, TAG_FOR_* etc. would now be
	  superfluous, since the format can now directly be modified.
	  Some of these variables thus have been either removed, but for
	  backward compatibility and "simple" modifications, most of them
	  are still honored by means of delayed substitutions in the formats.
	- Fix segfaults for certain malformed delayed references.

*eix-0.17.1
	Martin Väth <martin at mvath.de>:
	- Fix missing include: https://bugs.gentoo.org/show_bug.cgi?id=283549
	  Thanks to Michael Yang <yangofzeal at gmail.com>.
	- autotools: Tweaks, updates, and minor fixes.
	- manpage: Describe eix-diff output defaults.
	- Support for package{s,.mask,.unmask,.keywords}.d in profiles.
	  Use with care: Even portage does not support it yet.

*eix-0.17.0
	Martin Väth <martin at mvath.de>:
	- Renames: update-eix -> eix-update, diff-eix -> eix-diff,
	  update-eix-remote -> eix-remote, update-eix-layman -> eix-layman,
	  functions-eix.sh -> eix-functions.sh
	  The purpose is to get a consistent naming scheme: All programs of
	  eix now start with eix-* (except versionsort).
	  If you want to keep the old names, create corresponding symlinks:
	  This is explicitly supported and is not intended to be deprecated.
	  This is done with ./configure --enable-obsolete-symlinks or with
	  ./configure --enable-obsolete-reminder.
	  However, these symlinks are not installed by default, because
	  new users will not need them. So it is advisable to update your
	  scripts to the new names if they should ever run on newer systems.
	  Also, in eix-functions.sh two functions and a corresponding internal
	  variable have been renamed to match the new naming scheme (for the
	  functions, a wrapper is provided if you use the above configure
	  options). See the HISTORY section in "man eix" for further details.
	- Introduce EIX_SYNC_CONF (and slightly modified EIX_SYNC_OPTS)
	  EIX_REMOTE_OPTS, EIX_LAYMAN_OPTS, EIX_TEST_OBSOLETE_OPTS
	  These defaults are now evaluated in the corresponding scripts,
	  so be aware about quoting and security!
	- eix-sync: Introduce options and corresponding calling conventions.
	- eix-remote: Introduce -U -f -a (useful as defaults)
	- eix-remote: Include label names and correct paths in virtual names.
	- eix-remote: Optionally use sudo to change permissions
	- eix-remote: Remove -O since remote script now produces proper files.
	- eix-remote: Rename OVERLAYPARENT into LOCAL_LAYMAN and -o into -l
	  since now not the parent but the full path are required.
	  Improve documentation of that feature.
	- eix-test-obsolete: Require argument and support options.
	- Introduce --print-overlay-{path,label}
	- OVERLAYS_LIST=all now also prints the main tree as 0th entry.
	- Fix multiple declaration of local variable, see
	  https://bugs.gentoo.org/show_bug.cgi?id=275300
	  Thanks to Stuart Shelton <srcshelton at gmail.com> for the patch.
	- Better document eix-layman.
	- Ignore --disable-dependency-tracking without --disable-silent-rules.

*eix-0.16.2
	Martin Väth <martin at mvath.de>:
	- Default to SAVE_WORLD=true: I doubt that the world file is that
	  security relevant, and I guess many users don't know the feature.
	- Build system defaults to silent rules in case of >=automake-1.11.

*eix-0.16.1
	Martin Väth <martin at mvath.de>:
	- Allow optional #metadata-method(s) for cache method parse/ebuild,
	  if up-to-date metadata is found for the ebuild.
	  This speeds up things enormously if metadata is provided by an
	  overlay-maintainer (using: egencache --repo=foo --update).
	  Of course, it is still faster to use the metadata method directly.
	- Introduce CACHE_METHOD_PARSE to add the above automagically.
	- Internal change: merge cache methods parse and ebuild.
	- Fix memory leak in metadata.cc (aux was not freed in some cases).
	- Minor compatibility tweaks for gcc-4.4 and glibc-2.10

*eix-0.16.0
	Martin Väth <martin at mvath.de>:
	- Add NLS support and experimental translation into German.
	- Add versionsort helper tool (and ./configure --enable-separate-tool)
	- Support $PROFILE/package.keywords and $PROFILE/package.unmask
	- Workaround setenv for https://bugs.gentoo.org/show_bug.cgi?id=267807
	- Use non-capital variable names in scripts.
	- Minor bugfixes in scripts.
	- Some autotools/preprocessor cleanup.
	- manpage: Make clear that overlays must be paths, not repo-names.
	- manpage: Give some examples for egencache hooks with eix-sync.

*eix-0.15.7
	Martin Väth <martin at mvath.de>:
	- Allow USE dependencies https://bugs.gentoo.org/show_bug.cgi?id=264798
	- Update citation reference for "portage with sqlite"
	- Introduce EXPORT_PORTDIR_OVERLAY to export a new PORTDIR_OVERLAY with
	  added/excluded overlays to deal with eclasses. This somewhat fixes
	  https://bugs.gentoo.org/show_bug.cgi?id=249843
	  and https://bugs.gentoo.org/show_bug.cgi?id=265854

*eix-0.15.6
	Martin Väth <martin at mvath.de>:
	- Introduce EAPI_REGEX for EAPI-suffixed ebuilds (GLEP 55).
	- Introduce ...installedmarked... and ...availablemarked... to print
	  only those installed/available versions marked by --pipe.
	- Fix reference to obsolete cache type in eix-sync help page.
	- Make scripts more POSIX (e.g. "local" only at a function start).

*eix-0.15.5
	Martin Väth <martin at mvath.de>:
	- Introduce special overlay name ~ for cache methods eix/eix* to allow
	  easily reading of previous overlay data.
	- Fix regression which did not work for overlay labels/numbers for
	  cache method eix/eix*.
	- Be a bit more restrictive for the -H default since there are
	  packages containing "http" in their name.
	- Fix segfault/endless loop when an optional last argument is omitted.

*eix-0.15.4
	Martin Väth <martin at mvath.de>:
	- Hack around https://bugs.gentoo.org/show_bug.cgi?id=255711
	  (cannot compare const_backward_iterator with backward_iterator).
	- Use signed char instead of char for FullFlag, see
	  https://bugs.gentoo.org/show_bug.cgi?id=255994
	- Allow "virtual" for "provide || category-name" in DEFAULT_MATCH_FIELD
	  and prefer this in the defaults to avoid confusion for virtual/, see
	  https://sourceforge.net/tracker/index.php?func=detail&aid=2542891&
	  group_id=128101&atid=710608

*eix-0.15.3
	Martin Väth <martin at mvath.de>:
	- Allow relative (to PORTDIR) paths and magic "*" (for PORTDIR_OVERLAY)
	  in EIX_LOCAL_SETS and let the default contain "*/sets" and "sets".
	  This is to deal with the de facto standard of some overlays to put
	  new sets into $OVERLAY/sets.
	- Introduce FORCE_PERCENTAGE for better cron-job handling
	  https://sourceforge.net/tracker/index.php?func=detail&aid=2391952&
	  group_id=128101&atid=710611
	- Allow passing of options from eix-sync to update-eix.
	- More regexp-defaults (e.g. when @.* is searched).

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fix format string in ansicolor.cc for bug #254412

*eix-0.15.2
	Martin Väth <martin at mvath.de>:
	- Make regex-matches in DEFAULT_MATCH_{FIELD,ALGORITHM}
	  case-sensitive to make LICENSE defaults reasonable, see
	  https://forums.gentoo.org/viewtopic-p-5285010.html
	  or https://bugs.gentoo.org/show_bug.cgi?id=248677
	- Change EPREFIX_SOURCE default to %{EIX_PREFIX}, see
	  https://bugs.gentoo.org/show_bug.cgi?id=247227

*eix-0.15.1
	Martin Väth <martin at mvath.de>:
	- New match algorithms: --substring (-z), --begin (-b), --end
	- Introduce DEFAULT_MATCH_ALGORITHM. This should avoid some surprises
	  like "eix libstdc++-v3" looking for regular expressions as in
	  https://bugs.gentoo.org/show_bug.cgi?id=245487
	  or "eix @kde" also matching e.g. @kde3 as in
	  https://bugs.gentoo.org/show_bug.cgi?id=245773
	- Introduce DEFAULT_MATCH_FIELD; remove MATCH_.*_IF and MATCH_ORDER.
	- Fix order for <full...> and <=full...>. Slot comes after version.
	- Treat ADD_CACHE_METHOD and ADD_OVERRIDE_CACHE_METHOD only by
	  delayed substitution in CACHE_METHOD or OVERRIDE_CACHE_METHOD.
	  Note that this means that you should now add this delayed
	  substitution to redefinitions of the latter in e.g. /etc/eixrc.
	- Let Replace() in functions-eix.sh act on unexpanded arguments.
	- Only read the first set definition file if the same name occurs in
	  several directories.
	- Fix order in EIX_LOCAL_SETS: list local before global set files.
	- Fix missing includes, https://bugs.gentoo.org/show_bug.cgi?id=246270
	  https://bugs.gentoo.org/show_bug.cgi?id=246222
	- Fix LEVENSHTEIN_DISTANCE (actually honour the variable).
	- Name guards in a more standard manner (no dangerous _* names).
	- Internal change: More convenient argument order in split_string()

*eix-0.15.0
	Martin Väth <martin at mvath.de>:
	- Allow @set (from EIX_LOCAL_SETS) in /etc/portage/package.keywords.
	  This is more than portage currently does, although it is planned:
	  https://bugs.gentoo.org/show_bug.cgi?id=244366
	- Introduce /etc/portage/sets.eix in the EIX_LOCAL_SETS defaults,
	  and document in the description of that file and in the bug section
	  of the eix man page that eix' shortcomings for sets can be partially
	  (manually) circumvented with this directory.
	- For --set check also "@set" (i.e. in addition to "set").
	- Let @ default to set check (so that e.g. eix -e @kde is possible),
	  see https://bugs.gentoo.org/show_bug.cgi?id=245773
	- Introduce <empty...> to produce only category/name:slot output.
	- Change <full...> to produce less output (simplifies postprocessing).
	- Introduce <=full...>/<=empty...> to produce output portage likes.
	- Introduce %{\VAR} delayed references to avoid surprises in the
	  expansion of list variables like EIX_LOCAL_SETS or CACHE_METHOD.
	- Document escape mechanism of string lists in the manpage.
	- Turn *_NOWARN *_NONEXISTENT SLOT_UPGRADE_* into string lists.
	- Correct ChangeLog of 0.14.1/2 (segfault for cache method ebuild).
	- Pass PORTAGE_ROOTPATH and PORTAGE_BIN_PATH to ebuild.sh.
	  Note that ebuild.sh uses it to calculate PATH, so be careful!
	- Code cleanup (changes in inline/static/global).
	- Cleanup of header files. Some header file fixes (occur with gcc-4.4).
	- Fix various compilation problems, see
	  https://bugs.gentoo.org/show_bug.cgi?id=243396
	- Fix ERPEFIX_SOURCE (previously it had no effect...)
	- Fix varsreader.cc to deal also with files following DOS conventions;
	  improve handling of unclosed "${".
	- Fix segfault and hash problems with USE=-- (which might occur
	  if e.g. cache method parse fails), see
	  https://bugs.gentoo.org/show_bug.cgi?id=238216
	- Fix minor bugs in escape handling in lists.

*eix-0.14.2
	Martin Väth <martin at mvath.de>:
	- Introduce /etc/portage/package.slot_upgrade_{allow,forbid}
	  to allow exceptions from the general UPGRADE_TO_HIGHEST_SLOT policy.
	  This can be used as manual workaround for the missing dependency
	  resolving which will never be available in eix. See
	  https://bugs.gentoo.org/show_bug.cgi?id=236036
	- Fix scripts to use proper printf instead of echo.
	- Introduce MARK_UPGRADE.
	- Allow combining marker attributes, e.g. MARK_UPGRADE=bold,underline
	- Allow escaping for all variables containing space-separated paths.
	- Introduce <fullbest...> format.
	- Introduce <bestslotsupgrade...> (and <fullbestslotsupgrade...>)
	  format as a reaction to several requests, the latest being
	  https://forums.gentoo.org/viewtopic-t-278819-start-393.html
	  But please please note that if you use this in a script to call
	  emerge you are almost surely doing something wrong.
	  See the manpage for more details.
	- Fix typo in name --with-ebuild-sh-default
	- Fix segfault for cache methods ebuild and parse[*]|ebuild.

*eix-0.14.1
	Martin Väth <martin at mvath.de>:
	- Support sets in masks or sets; provide RECURSIVE_SETS.
	- Provide --installed-unstable, --installed-testing, --installed-masked
	  see https://bugs.gentoo.org/show_bug.cgi?id=240010
	- Make EIX_LOCAL_SETS a list of directories.
	- Fix visible output of cache methods metadata/flat, see
	  https://bugs.gentoo.org/show_bug.cgi?id=240038
	- Introduce ./configure --with-ebuild-sh-default and
	  EXEC_EBUILD, EXEC_EBUILD_SH, EBUILD_DEPEND_TEMP to allow
	  customization to nonstandard paths, see
	  https://bugs.gentoo.org/show_bug.cgi?id=240460
	  Default is autodetection which works in the above case.
	- Remove EPREFIX_PORTAGE_EXEC (except for delayed substitution
	  for the above new variables).
	- Be extremely careful about vfork() usage to fix segfault from
	  https://bugs.gentoo.org/show_bug.cgi?id=240460
	- New cache methods repo-* support paludis' path scheme.
	- Be more verbose on execution failures for update-eix.
	- Cache method ebuild (without "*") and parse[*]|ebuild are broken in
	  this release! (I observed only after the release that they segfault).
	  Please upgrade to >=eix-0.14.2 if you need one of these methods.

*eix-0.14.0
	Martin Väth <martin at mvath.de>:
	- Support for PROPERTIES and RESTRICT.
	- Inquire sqlite database more flexible. This should fix
	  https://bugs.gentoo.org/show_bug.cgi?id=238942
	- Fix ./configure --with-eprefix-default/--with-eix-prefix-default.
	  Thanks to Michał Bartoszkiewicz <mbartoszkiewicz@gmail.com> for
	  investigating, see https://bugs.gentoo.org/show_bug.cgi?id=235011
	- Assume SLOT="" is not ok for parse[*]|ebuild[*], see
	  https://bugs.gentoo.org/show_bug.cgi?id=238882

*eix-0.13.5
	Martin Väth <martin at mvath.de>:
	- Fix segfault in numeric comparison of 0 and number with leading 0,
	  see https://forums.gentoo.org/viewtopic-t-278819-start-375.html
	- Disable autotools maintainer mode which seems to be broken anyway.
	- Fix -u in manpage, see https://bugs.gentoo.org/show_bug.cgi?id=236036

*eix-0.13.4
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Correctly interpret package.keywords: multiple lines with same mask
	  are combined; only use default keywords if combination is empty.
	- Print warning if "source" failed.

	Martin Väth <martin at mvath.de>:
	- Let eix-sync be tacit on hooks and evaluate them in current shell.
	  This makes hooks more powerful and gives the user the ability to
	  treat redirection manually to some extent, see
	  https://sourceforge.net/tracker/index.php?func=detail&aid=2042001&
	  group_id=128101&atid=710611
	- Improve calculation of times in eix-sync; avoid time command, see
	  https://bugs.gentoo.org/show_bug.cgi?id=235269
	- Introduce EIX_PREFIX to split the previous double meaning of EPREFIX,
	  see https://bugs.gentoo.org/show_bug.cgi?id=235011
	  Now EPREFIX has a meaning similar to that of prefix-portage.
	  For chroot environment you should set EIX_PREFIX instead;
	  The EIX_PREFIX environment variable now also wins over
	  PORTAGE_CONFIGROOT for the /etc/eixrc path.
	- Fix cache method eix: Previously, iuse (version-wise) was forgotten.
	- Change default definitions of FORMAT_* and DIFF_FORMAT_* so that it
	  is easier to make "wrappers" around them (e.g. on the commandline).
	- Remove mentioning of /var/log/eix-sync.log from the manpage;
	  this logfile was obsoleted already in eix-0.13.3.

*eix-0.13.3
	Martin Väth <martin at mvath.de>:
	- New cache method "metadata-assign" needed for alt-gentoo, see
	  https://bugs.gentoo.org/show_bug.cgi?id=231808
	- Rename some of the other cache methods to have a consistent scheme
	  (metadata->metadata-flat, portage-2.0->flat, portage-2.1->assign).
	  Also treat internally flat/assign and metadata-flat/assign by the
	  same class, i.e. remove the port2_1_0 directory completely.
	  The obsolete names will keep getting supported for quite a while,
	  but the update-eix output only uses the new names.
	- Rename update-eix-functions.sh into functions-eix.sh
	  I am aware that this will break several user scripts, but these
	  scripts are not so hard to fix either (or you can make a symlink)
	  and getting rid of the obsolete misleading name was long overdue.
	- You now should call "read_functions [args]" after sourcing
	  functions-eix.sh; this will read baselayout's functions.sh with
	  the corresponding [args]. This fixes the error that the "-C"
	  argument was possibly passed to the sourced script in some shells,
	  mistakenly producing non-colored output when the user passed "-C"
	  to some script. In the new manner you can now determine explicitly
	  what should be passed and how.
	- Use python in update-eix-layman to parse layman.cfg for its
	  make.conf, see https://bugs.gentoo.org/show_bug.cgi?id=230788
	- Remove logging support of eix-sync (and drop options -v and -V).
	  This should fix all problems with the emerge --ask option, see
	  https://bugs.gentoo.org/show_bug.cgi?id=233132
	  If you use eix-sync in a cronjob you now have to redirect manually!
	- Be more verbose in ./configure output
	- Improve pkg-config error treatment
	- Improve format description of world/world_sets
	- Use inference rules in Makefile where $< is used (needed by POSIX).
	  This also avoids separate rules for separate *.in files.

*eix-0.13.2
	Martin Väth <martin at mvath.de>:
	- Extend the manpage concerning the description of expressions, see
	  https://bugs.gentoo.org/show_bug.cgi?id=229625
	- Fix buggy want_slash=false in normalize_path(), see
	  https://bugs.gentoo.org/show_bug.cgi?id=230341
	- Do not demand canonicalize_file_name() or realpath():
	  The profile reading should even work without symlink resolving.

*eix-0.13.1
	Martin Väth <martin at mvath.de>:
	- Fix defaults for FORMATLINE_PACKAGESETS, see
	  https://bugs.gentoo.org/show_bug.cgi?id=229355
	- Fix broken replacement update-eix-functions.sh path and a bug in
	  normalize_path(), see https://bugs.gentoo.org/show_bug.cgi?id=229419

*eix-0.13.0
	Martin Väth <martin at mvath.de>:
	- Support for package sets.
	- Support for world file. You need execute permissions for
	  /var/lib/portage or set SAVE_WORLD=true and run update-eix with
	  such permissions to make use of this feature.
	- Introduce --installed-slot, MATCH_IF_*SLOT
	- The new defaults for eix-sync now neither call emerge --metadata
	  nor clean /var/cache/edb. You might want to set up an appropriate
	  /etc/eix-sync.conf with previous defaults if you use a cache method
	  depending on this like "portage-2.1" (most users are not affected).
	  See the manpage and eix-sync -h for details. See the discussion in
	  https://bugs.gentoo.org/show_bug.cgi?id=202535
	- Allow "!!" lines in eix-sync to override environment variables like
	  FORCE_USECOLORS, see https://bugs.gentoo.org/show_bug.cgi?id=225963
	- Treat /etc/make.profile with normalize_path(). This should fix
	  https://bugs.gentoo.org/show_bug.cgi?id=229091
	- Fix treatment of "-" in slotnames.
	  Thanks to Edward "Hades" <edward.hades at gmail.com> for the patch,
	  see https://bugs.gentoo.org/show_bug.cgi?id=225773
	- Fix PATH_MAX headers on FreeBSD-7. Thanks to
	  Javier Villavicencio <the_paya at gentoo.org> for the patch, see
	  https://bugs.gentoo.org/show_bug.cgi?id=226499
	- Fix recognition of system packages according to local settings.
	- Provide is now HashedWords. Thanks to
	  Charlie Shepherd <masterdriverz at gmail.com> for reporting.
	- Minor simplifications and compatibility fixes.
	- Finally updated AUTHORS and TODO (Sorry, I had forgotten about
	  these files for a long time) and reformatted ChangeLog.

*eix-0.12.5
	Martin Väth <martin at mvath.de>:
	- Introduce DIFF_SEPARATE_DELETED and let it default to true.
	- Introduce DIFF_PRINT_HEADER.
	- Ignore leading + or - for /var/db/pkg/*/*/{,I}USE for paludis.
	  Do the same with IUSE data in packages for EAPI=1. Thanks for the
	  hint to Leonardo Valeri Manera <l.valerimanera at gmail.com>
	- Fix error handling in pushback_lines_file(), see
	  https://bugs.gentoo.org/show_bug.cgi?id=220119
	- Fix some bsd shell incompatibilities in scripts, see
	  https://bugs.gentoo.org/show_bug.cgi?id=220397
	- Introduce MAKE_GLOBALS default /usr/share/portage/config/make.globals
	  See https://bugs.gentoo.org/show_bug.cgi?id=221271
	- Reorganize eixrc/defaults.cc
	- Export PATH and other portagesettings variables to ebuild.sh, see
	  https://bugs.gentoo.org/show_bug.cgi?id=219066
	- Minor documentation fixes.

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Simplify pushback_lines_file()

*eix-0.12.4
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Drop EIX_WIKI and use PACKAGE_BUGREPORT; Label Wolfgang and Roland
	  as inactive developers, so they are less likely to be bothered with
	  emails.
	- Fix https://bugs.gentoo.org/show_bug.cgi?id=216483, a mask like
	  =app-portage/eix-0* triggered an exception.
	  Affected versions are >=eix-0.12.0.

	Martin Väth <martin at mvath.de>:
	- Treat parent profiles before children, see
	  https://bugs.gentoo.org/show_bug.cgi?id=217120
	- Let update-eix first process options like --print.
	  This also fixes script problems with KEEP_VIRTUAL=true.
	  Thanks to tzopik <tzopik at gmail.com> for reporting and
	  investigating. See https://bugs.gentoo.org/show_bug.cgi?id=216972
	- Conversion of all latin1 characters to unicode.
	- Fix segfault/endless loop when loops in profile/**/parent data occur.
	- Decrease required number of simultaneously open files.

*eix-0.12.3
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fix sqlite error messages
	- Add SysError exception and use it in database code, so we throw
	  errors and don't hang if there is something wrong with the database.
	- Make io::read_packagetree a little faster by reading the database as
	  categories with packages, not just as a stream of packages.
	  Only do seek in PackageReader::skip() if we didn't read all
	  information.
	- Add support for multiple inheritance in "parent" files of profiles.

	Martin Väth <martin at mvath.de>:
	- Avoid scandir(), to allow compilation under Solaris 9. Thanks to
	  Jeremy Olexa <darkside at gentoo.org> for the hints, see
	  https://bugs.gentoo.org/show_bug.cgi?id=195411
	- Fix NormalizeNames() of update-eix-function.sh, see
	  https://bugs.gentoo.org/show_bug.cgi?id=215417
	- Fix possible endless loop for cache method metadata and portage-2.1

*eix-0.12.2
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Replace "$@.in" with "$<" in Makefile.am (which gives us the real
	  path, not just the name of the file); Remove configmake.h and add
	  SYSCONFDIR to AM_CXXFLAGS (as mentioned in the autoconf manual);
	  "make distcheck" works again.
	- BasicVersion: Store m_parts as list (avoids reallocation overhead).
	- Version: hold iuse as vector, not as string so we don't need to
	  split/join so much. Join vector only on request and cache result.
	- Package: Do not split/join collected iuse everytime a version is
	  added, instead store a set of useflags in the package, and add the
	  new iuse from the version. Only join the flags into a string if
	  iuse() is called (and cache the result for subsequent calls).
	  Also, do not rebuild the slotlist everytime a version was added, but
	  rather only build a slotlist if one is actually requested by
	  slotlist() (and cache the resulting slotlist).
	- database: Remove splitting/joining of strings/vectors, since we now
	  have vectors/sets in Package, Version, etc. This brings a huge speed
	  improvement for version reading.
	- Move stdint.h stuff into eixTk/inttypes.h and include. Add static
	  assertion macro EIX_STATIC_ASSERT. Add database/types.h and include
	  if needed.
	- Remove uint, see https://bugs.gentoo.org/show_bug.cgi?id=213769
	- Kill --as-needed, see https://bugs.gentoo.org/show_bug.cgi?id=213199

	Martin Väth <martin at mvath.de>:
	- Fix REDUNDANT_IF_DOUBLE_{UN,}MASKED in eix-test-obsolete.
	- Introduce metadata:PATH and flat:PATH to allow usage of metadata
	  calculated by paludis. Thanks for the idea to
	  Leonardo Valeri Manera <l.valerimanera at gmail.com>
	- Fix signal handling for ebuild* (and parse{,*}|ebuild*) cache.
	- Honour EPREFIX_PORTAGE_EXEC in parse{,*}|ebuild{,*}. This should fix
	  https://bugs.gentoo.org/show_bug.cgi?id=213248
	  Thanks to Fabian Groffen <grobian at gentoo.org> for finding out.
	- Improve treatment of argv[0] (for --help).
	- Merge implementation of metadata and flat/portage-2.0.
	- Improve interplay of error messages and percent status.
	- Avoid some hangs on damaged eix files.

*eix-0.12.1
	Martin Väth <martin at mvath.de>:
	- Do not die on obsolete eix cache files, see
	  https://bugs.gentoo.org/show_bug.cgi?id=212637
	- Default to create only one binary file. This needs less disk space
	  but more memory. Moreover, it means that the eix binary needs shared
	  libraries which are otherwise only needed for update-eix
	  (e.g. sqlite). The ./configure options --enable-separate-binaries or
	  --enable-separate-update can be used to get the previous behaviour
	  or to split only update-eix into a separate binary, respectively.
	- Avoid libtool usage, since no libraries have to be built anymore.

*eix-0.12.0
	Martin Väth <martin at mvath.de>:
	- BE AWARE!!
	  The new default OVERLAY_CACHE_METHOD=parse|ebuild*
	  is a security risk if you have untrusted ebuilds in your overlay!
	  Use OVERLAY_CACHE_METHOD=parse if you do not trust your overlays!
	- Introduce new cache methods parse{,*}|ebuild{,*}
	- Introduce EBUILD_{USER,GROUP,UID,GID} to make cache method ebuild{,*}
	  reasonably secure.
	- Rename cache methods none{,*} to parse{,*} to force users of "none"
	  to observe that the new cache method parse|ebuild* exists (and
	  probably switch to it).  The name "none" was misleading anyway.
	- Introduce *-versions of the best/bestslot format strings to avoid
	  printing of all available versions for diff-eix or eix -c.
	- Introduce *NOBEST_* to make diff-eix or eix -c more compact.
	- Introduce --with-rst configure option for database docu html.
	- Fix spaces in results for cache method none{,*}/parse{,*}
	- Remove broken cache method portage-2.1* unless a corresponding
	  ./configure option is used (which is mainly for testing purposes).
	- Introduce --enable-as-needed to override the -Wl,--as-needed test,
	  see the comments in https://bugs.gentoo.org/show_bug.cgi?id=209239
	- Fix some pedantic compiler warnings; make basicversion.cc using std.
	- Make update-eix slightly more verbose.

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Add documentation for database format.

*eix-0.11.0
	Emil Beinroth <emilbeinroth at gmx.net>:
	- kill SAVE_VERSIONTEXT
	- rewrite parts of BasicVersion and store versions as a vector of Parts

	Martin Väth <martin at mvath.de>:
	- Introduce /etc/portage/package.*.nonexistent
	- Introduce /etc/portage/package.*.nowarn
	- Introduce --print-all-* options
	- Introduce "detail" option for eix-test-obsolete, see
	  https://forums.gentoo.org/viewtopic-t-666071.html
	- Shrink database file by hashing common words.
	- Fix REDUNDANT_IF_DOUBLE_LINE, REDUNDANT_IF_MINUS_ASTERISK and also
	  fix corresponding parts in eix-test-obsolete.
	- Fix update-eix-layman (broken since eix-0.10.4), see
	  https://bugs.gentoo.org/show_bug.cgi?id=211093
	  Thanks to Roberto Castagnola <roberto.castagnola at gmail.com>
	  for the fix.
	- Fix update-eix-remote for users without shell (like "nobody"), see
	  https://bugs.gentoo.org/show_bug.cgi?id=211348
	- Install "empty" /etc/eixrc to avoid that users overlook the new
	  recommendations and thus get problems like e.g. the one leading to
	  https://bugs.gentoo.org/show_bug.cgi?id=211192
	- Support prefix-portage subrevisions
	  https://www.gentoo.org/proj/en/
	  gentoo-alt/prefix/techdocs.xml#doc_chap2
	  See also https://bugs.gentoo.org/show_bug.cgi?id=209239
	- Better honour SYSCONFDIR. Use configmake.h to pass to C++ code.
	- Let autoconf try to define uint*_t

*eix-0.10.5
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Use STL for revision parsing.

	Martin Väth <martin at mvath.de>:
	- Fix some problems with EOF after variable names or assignments, see
	  https://bugs.gentoo.org/show_bug.cgi?id=210106
	- Work around a dash bug ("VAR=e func_call" does not restore VAR).

*eix-0.10.4
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Add --slot for matching the slot of versions. eix --slot kde-4 will
	  search packages that have at least one version with a slot that
	  matches kde-4.
	- Remove all uses of __VA_ARGS__, see
	  https://bugs.gentoo.org/show_bug.cgi?id=209239
	- Fix https://bugs.gentoo.org/show_bug.cgi?id=209934

	Martin Väth <martin at mvath.de>:
	- Make REDUNDANT_IF_MIXED=false the default since it confused people:
	  https://bugs.gentoo.org/show_bug.cgi?id=203067
	- Make defaults independent of bzlib usage (considering overlays of
	  installed packages makes always sense in current portage versions).
	- Finally use POSIX shell throughout, making bash obsolete.
	- Clean up manpage formatting.
	- Some compatibility fixes, see
	  https://bugs.gentoo.org/show_bug.cgi?id=209239

*eix-0.10.3
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fix compilation for ia64-hpux, see
	  https://bugs.gentoo.org/show_bug.cgi?id=201117
	- Minor changes (size() == 0 -> empty(), ...)

	Martin Väth <martin at mvath.de>:
	- Introduce PRINT_KEYWORDS, FORMAT_{BEFORE,AFTER}_KEYWORDS
	- Let -T check overlay for all-installed and some-installed
	- Fix segfault for unknown repository and unreadable environment.bz2
	- Use POSIX' command -v instead of type/type -P in scripts
	- Clearer error messages for newer cachefile format, see
	  https://forums.gentoo.org/viewtopic-t-614015.html

*eix-0.10.2
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fix gcc-4.3 warnings
	- Include ICC and `make distcheck` in contrib/compile-test.sh
	- Remove useless OOM_ASSERT, WARNING and THROW macros.
	- Replace PrintVar macro with function, simplify get_eixrc()

	Martin Väth <martin at mvath.de>:
	- Fix REDUNDANT_IF_WEAKER and REDUNDANT_IF_MIXED which were
	  completely broken in eix-0.10.1 due to a stupid typo.
	- New treatment of scandir, see
	  https://bugs.gentoo.org/show_bug.cgi?id=195411
	- Introduce PRINT_BUGS, see
	  https://bugs.gentoo.org/show_bug.cgi?id=195204
	- New number format in cache file

*eix-0.10.1
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Replace full license headers with the following:
	   This file is part of the eix project and distributed under the terms
	   of the GNU General Public License v2. + Copyright notice
	  This change removes about 2000 useless lines from the project.

	Martin Väth <martin at mvath.de>:
	- Support for RESTRICT="fetch mirror", see
	  https://forums.gentoo.org/viewtopic-t-278819-start-355.html
	- Introduce ALWAYS_ACCEPT_KEYWORDS and let it default to "false" to
	  imitate the behavior of <eix-0.10.0 for ~ARCH in ACCEPT_KEYWORDS,
	  see https://bugs.gentoo.org/show_bug.cgi?id=193719
	- Improve internal treatment of SetStability.
	- Many fixes due to Emil Beinroth <emilbeinroth at gmx.net>.
	- Decrease memory usage during compilation.
	- More standard-conformal treatment of header files:
	  eix should now compile with gcc-4.3

*eix-0.10.0
	Martin Väth <martin at mvath.de>:
	- Rewrite stability setting, fixing thereby a lot of bugs and
	  inconsistencies. In particular, this fixes ACCEPT_KEYWORDS=~ARCH
	  and -~ARCH in /etc/portage/package.keywords (thanks to
	  Emil Beinroth <emilbeinroth at gmx.net> for reporting).
	  This also fixes problems with several ARCHes in ACCEPT_KEYWORDS.
	- Treat leading zeroes of version-related numbers in the portage way,
	  see https://bugs.gentoo.org/show_bug.cgi?id=192807
	- Store all version-related numbers as strings. This avoids overflows
	  and actually saves space and (for the new comparison) even time.
	- Store string-length more efficiently.
	- Remove redundant data from the database:
	  (a) The full version string is now calculated at run-time.
	  (b) Only the original keyword string is stored, no stability flags.
	      They have to be recalculated at run-time anyway.
	  However, the profile-mask/system-packages flags are still saved for
	  treatment with diff-eix.
	- Due to (b) above, the SMALL_EIX_DATABASE variable was canceled.
	- Fix progress output. Thanks to Mark J. Olah <mjo at cs.unm.edu> for
	  reporting. See https://bugs.gentoo.org/show_bug.cgi?id=192458
	- Fix https://bugs.gentoo.org/show_bug.cgi?id=192707

*eix-0.9.12
	Martin Väth <martin at mvath.de>:
	- Introduce eix-test-obsolete [no-]repository
	- Re-introduce optional parameter for eix-test-obsolete to set
	  (but not toggle!) CHECK_INSTALLED_OVERLAYS: Users of older portage
	  versions might find it convenient.
	- Better document repository behaviour.
	- Introduce SORT_INST_USE_ALPHA
	- Introduce TEST_{KEYWORDS,MASK,UNMASK,USE,CFLAGS,REMOVED}.
	  In particular, users might want to switch off the package.cflags
	  tests if they use a more sophisticated format for it in
	  /etc/portage/bashrc
	- Only scan /etc/portage/package.{cflags,use} if -T really has to:
	  This is faster and avoids possible undesired error messages.
	- Consider profile/package.mask and overlay/profile/package.mask as
	  (first or last) part of profile w.r.t. '-' entries. This fixes
	  https://bugs.gentoo.org/show_bug.cgi?id=191716

*eix-0.9.11
	Martin Väth <martin at mvath.de>:
	- Support for repository names.
	- Introduce <fullinstalled>.
	- Introduce <fullavailable>, <fullavailableshort>, <fullavailablelong>.
	- Bugfix: Do not shorten /var/log/eix-sync.log near end of eix-sync.
	- Bugfix: eix no longer considers output from <...> as "no output"
	  (for COUNT_ONLY_PRINTED or trailing newline). This fixes
	  https://sourceforge.net/tracker/index.php?
	  func=detail&aid=1780283&group_id=128101&atid=710608
	- eix-sync: introduce -V to make -v in /etc/eix-sync.conf convenient.
	- Avoid "bashisms" in scripts except for arrays (which are necessary).

*eix-0.9.10
	Martin Väth <martin at mvath.de>:
	- Fix compilation issues with gcc-3.3.6
	- Do not treat # as comment symbol in /var/db/pkg/*/*/{USE,IUSE,SLOT}.
	  This fixes https://sourceforge.net/tracker/index.php?
	  func=detail&aid=1756510&group_id=128101&atid=710608
	- Fix some crashes if there is bad data in /var/db/pkg
	- Fix some memory allocations

*eix-0.9.9
	Martin Väth <martin at mvath.de>:
	- Allow slot names for masks, see
	  https://bugs.gentoo.org/show_bug.cgi?id=181244
	- Fix compilation issues with some C compilers, see
	  https://sourceforge.net/tracker/index.php?
	  func=detail&aid=1731374&group_id=128101&atid=710608
	- Fix DIFF_FORMAT_HEADER_NEW default (was broken in eix-0.9.8;
	  thanks to Seemant Kulleen <seemant at gentoo.org> for reporting.)
	  See also https://bugs.gentoo.org/show_bug.cgi?id=181629
	- Introduce SKIP_PERMISSION_TESTS to deal with NSS/LDAP or other
	  setups, see https://sourceforge.net/tracker/index.php?
	  func=detail&aid=1735722&group_id=128101&atid=710611

*eix-0.9.8
	Martin Väth <martin at mvath.de>:
	- Fix calculation of needed data which was broken since eix-0.9.5.
	  This brings "cheap" searches (for names etc) back to the speed of
	  eix-0.9.4. See https://sourceforge.net/tracker/index.php?
	  func=detail&aid=1728310&group_id=128101&atid=710608
	- New default format for diff-eix: Print * tag for stabilized packages,
	  see https://forums.gentoo.org/viewtopic-p-4078742.html
	- Introduced delayed TAG_* variables in default formats.
	- Suppress newline when the FORMATSTRING outputs nothing. This allows
	  to use conditionals in the formatstring to select packages.
	  Examples are given in the FORMATSTRING section of the manpage.
	- Introduce corresponding COUNT_ONLY_PRINTED
	- Fixed update-eix-remote remove
	- manpage recommends to set KEEP_VIRTUALS=true for update-eix-remote
	- Fix (once more) color output
	- Major internal cleanup of masking functions, in particular:
	- Avoid local profile masking if /etc/portage/profile does not
	  contain any relevant data.

*eix-0.9.7
	Martin Väth <martin at mvath.de>:
	- eix-sync now defaults to layman instead of deprecated gensync. Thanks
	  for the patch to Roberto Castagnola <roberto.castagnola at gmail.com>
	  see https://bugs.gentoo.org/show_bug.cgi?id=179482
	- Introduce PRINT_COUNT_ALWAYS=never
	- Introduce --pure-packages and --only-names
	- Introduce +- variants of --stable, --testing, --non-masked, --system.
	- Rename --update into --upgrade
	- Introduce --upgrade+ --upgrade-
	- Replace UPGRADE_ALWAYS_LOCAL and RECOMMEND_ALWAYS_LOCAL by the
	  more powerful UPGRADE_LOCAL_MODE and RECOMMEND_LOCAL_MODE
	- UPGRADE_ALWAYS_LOCAL=true was not working correctly in the previous
	  versions. The new UPGRADE_LOCAL_MODE=+ should fix this
	- Fix a serious memory leak in PackageTest::match()

*eix-0.9.6
	Martin Väth <martin at mvath.de>:
	- Fix diff-eix: Flag whether local settings should be
	  obeyed was inverted by mistake in 0.9.5.
	- Fix logic error in {recommend} when new slot vanishes.
	- Minor fixes and simplifications in default formats.

*eix-0.9.5
	Martin Väth <martin at mvath.de>:
	- Add Bugs/FAQ section in manpage.
	- Introduce OBSOLETE_MINUSASTERISK and set it to false,
	  to deal with the soon coming portage-2.1.3.
	- Introduce corresponding REDUNDANT_IF_MINUSASTERISK.
	- Introduce --stable, --testing, --non-masked, --system, see
	  https://forums.gentoo.org/viewtopic-t-556942.html
	- Honour /etc/portage/profile. This fixes
	  https://bugs.gentoo.org/show_bug.cgi?id=176840
	- Remove FORMAT_INSTALLED_USE for security reasons.
	- Some further security fixes.
	- Remove asprintf usage.
	- Fix possible endless loop in color parsing.

*eix-0.9.4
	Martin Väth <martin at mvath.de>:
	- Introduce PRINT_IUSE, FORMAT_{BEFORE,AFTER}_{,COLL_,SLOT_}IUSE
	- Introduce --installed-with-use, --installed-without-use
	- Introduce ./configure option --use-bzip2 to enable the following:
	- Print overlay from which a version was installed.
	- eix-test-obsolete now tests also for versions installed from
	  obsolete overlays by default. To make this test reliable you have
	  to pass an argument to eix-test-obsolete (or you have to set
	  CHECK_INSTALLED_OVERLAYS). However, then the test takes much longer.
	  See the manpage for details.
	- Try to mark versions as "installed" only for appropriate overlay.
	  (If no matching overlay exists, fall back to old marking.)
	- Change <installed:...> syntax by adding support for overlay output.
	- Change -J to mean really installed *from* some overlay.
	  The old meaning is available as --installed-in-some-overlay.
	- Introduce --installed-from-overlay (corresponding to new -J option).
	- Introduce NONEXISTENT_IF_OTHER_OVERLAY.
	- Introduce CHECK_INSTALLED_OVERLAYS.

*eix-0.9.3
	Martin Väth <martin at mvath.de>:
	- Fix nonexistent asprintf and broken FNM_CASEFOLD handling
	  for solaris, see https://bugs.gentoo.org/show_bug.cgi?id=173763
	- Fix manpage typo --update:
	  https://bugs.gentoo.org/show_bug.cgi?id=174489
	- Fix some missing include unistd.h (required for linux 2.4), see
	  https://bugs.gentoo.org/show_bug.cgi?id=174697
	- Use extended path for functions.sh to work with baselayout-2.
	  Thanks to Matthias Schwarzott for pointing out.

*eix-0.9.2
	Martin Väth <martin at mvath.de>:
	- Fix partly https://bugs.gentoo.org/show_bug.cgi?id=169127
	  (compilation error on big-endian architectures).
	- Fix std::unique which is presumably broken on some systems.
	  Perhaps this fixes the above issues completely.
	- Reformulate security recommendations in update-eix-remote. Thanks to
	  Marijn Schouten, see https://bugs.gentoo.org/show_bug.cgi?id=169806
	- Fix nocolor output. Thanks to Rick Morra for reporting
	  https://bugs.gentoo.org/show_bug.cgi?id=170868
	- Fix compilation issues with gcc-3.3.6.
	  Thanks to Helmut Auer <helmut at helmutauer.de>
	- Minor compatibility fixes.

*eix-0.9.1
	Martin Väth <martin at mvath.de>:
	- Use environment's EPREFIX to find /etc/eixrc if
	  PORTAGE_CONFIGROOT is unset.
	- New ./configure options allow customization of default values for
	  EPREFIX and ROOT.
	- Introduce prefix-string-type for variables to make e.g. ROOT=/
	  equivalent to ROOT=''.

*eix-0.9.0
	Martin Väth <martin at mvath.de>:
	- Let * and ~* in /etc/portage/package.keywords accept stable
	  and testing keyword, respectively, of any architecture.
	  Moreover, consider this as weaker (preferrable) than ** and
	  (the soon obsolete) -*.
	- Introduce corresponding TAG_FOR_{EX_,}ALIEN_{UN,}STABLE.
	- Introduce --print option and use it in the scripts to have
	  consistent variable treatment throughout.
	- Introduce ROOT and EPREFIX_* fine granulation, see
	  https://bugs.gentoo.org/show_bug.cgi?id=167961
	- Introduce EIXRC and EIXRC_SOURCE
	- Introduce EIX_CACHEFILE
	- Let -ARCH in /etc/portage/package.keywords stabilize -ARCH keyword.
	- Change default for TAG_FOR_{EX_,}{MINUS_ASTERISK,MISSING_KEYWORD}.
	- Drop {EX_,}TAG_FOR_STRANGE.
	- Fix memory leak in PortageUserConfig::setStability().
	- Fix some rare exceptional cases in redundancy testing of
	  /etc/portage/package.keywords and in marking of keywords.
	- Improve documentation for -I.
	- Let KEEP_VIRTUALS default to false since it confused too many people.
	  Moreover, it is somewhat dangerous if (falsely) used with
	  EPREFIX and friends.
	- Code cleanup (new-style casts and other pedantic compiler warnings).

*eix-0.8.8
	Martin Väth <martin at mvath.de>:
	- Change to new version number scheme:
	  https://bugs.gentoo.org/show_bug.cgi?id=165414
	- Fix sorting of things like _p20060716 (longer numbers)
	- Increase some quantities to 64bit (e.g. 20071042308
	  should still be a valid version number), or to  16bit (e.g. -r9999).
	- Internal cleanup: Proper variable typing.

*eix-0.8.7
	Martin Väth <martin at mvath.de>:
	- Let ** in /etc/portage/package.keywords accept "missing keyword".
	  This corresponds to the change in portage-2.1.2-r4, see
	  comments #20 #24 in https://bugs.gentoo.org/show_bug.cgi?id=160519
	- Let PORTAGE_CONFIGROOT modify the path of /etc/eixrc.
	- Allow PORTAGE_CONFIGROOT also in /etc/eixrc and ~/.eixrc
	  if not set in the environment.
	- Introduce EPREFIX which is similar to PORTAGE_CONFIGROOT but
	  applies for all paths except /usr/bin/update-eix-function.sh,
	  $HOME/eixrc, PORTAGE_PROFILE, PORTDIR, overlay paths,
	  and the cachefile arguments on the command line.
	  This allows support for prefix-portage.
	- Introduce EPREFIX_PORTAGE which is used as prefix for
	  PORTAGE_PROFILE, PORTDIR, and overlay paths.
	- Use true (soft-link resolved) filenames for PORTDIR and overlays.
	- Bugfix: Apply package.mask of profiles. This fixes
	  https://bugs.gentoo.org/show_bug.cgi?id=164695

*eix-0.8.6
	Martin Väth <martin at mvath.de>:
	- Ignore *~ files in /etc/portage/package.*
	  This reflects the change of portage-2.1.2_rc4-r9 of
	  https://bugs.gentoo.org/show_bug.cgi?id=144445
	- Fix <availableversionsshort>.
	- Security improvement: Use random temporary names also on systems
	  without mktemp (in eix-sync and update-eix-remote).
	- Add Push and NormalizeNames to update-eix-functions.sh
	- Code cleanup of scripts, in particular of eix-sync
	- Correct EIX_WIKI to https://eix.sourceforge.net

*eix-0.8.5
	Martin Väth <martin at mvath.de>:
	- Honour PORTAGE_CONFIGROOT
	- Honour PORTAGE_PROFILE
	- Allow environment to override/add to portage config variables
	  like PORTDIR, PORTDIR_OVERLAY, ...
	- Introduce --cache-file
	- Rename cache methods flat/backport to portage-2.0/portage-2.1
	- Experimental portage-2.1* cache method (don't use it!).
	- Prepare database format for perhaps future per-version IUSE data.

*eix-0.8.4
	Martin Väth <martin at mvath.de>:
	- Treat empty regular expression manually. This fixes
	  https://bugs.gentoo.org/show_bug.cgi?id=158222
	- Add option -O to update-eix-remote, see the discussion starting at
	  https://forums.gentoo.org/viewtopic-t-278819-start-343.html
	- Add option -W and more general options -c, -C, -M to eix-sync, see
	  https://bugs.gentoo.org/show_bug.cgi?id=159005

*eix-0.8.3
	Martin Väth <martin at mvath.de>:
	- Fix default FORMAT_HEADER_DELETE (reset color in [D]).
	- Add update-eix-layman
	- Add update-eix-functions.sh
	- Store IUSE in database (currently only for packages).
	- Add option -U (suggested by stephan.ritscher).

*eix-0.8.2
	Martin Väth <martin at mvath.de>:
	- Add options --in-overlay, --only-in-overlay
	- Add options -J, --installed-in-overlay
	- Allow wildcard patterns for overlays in --exclude-overlay and
	  --override-method (and for the corresponding variables) and
	  for the overlay references in cache method eix/eix*
	- Bugfix: Treat lines with identical target in
	  /etc/portage/package.keywords in the way portage does it
	  (i.e. ignore earlier lines if target is completely identical).
	- Introduce REDUNDANT_IF_DOUBLE_LINES (related to previous bugfix).
	- Some code cleanup and fixing of some compiler warnings.
	  Thanks to KIMURA Masaru / hiyuh <hiyuh.root at gmail.com>
	  https://bugs.gentoo.org/show_bug.cgi?id=155160

*eix-0.8.1
	Martin Väth <martin at mvath.de>:
	- Conditional blocks in delayed references.
	- Make use of conditionals in defaults:
	  Introduce delayed PRINT_ALWAYS, DIFF_PRINT_INSTALLED.
	  In particular, PRINT_ALWAYS finally is a way to have a "short"
	  configuration for cyberpatrols wish to print lines even if unused:
	  https://forums.gentoo.org/viewtopic-t-278819-start-292.html
	- Let eix-sync print the time emerge --sync needed.
	- Bugfix: Let "%%{" work as documented in the treatment of
	  delayed variables.
	- Bugfix: Reset colors after each printed version.
	- Bugfix in defaults: color of marked versions/packages.
	- Tiny bugfix in defaults: color of ":" after installed version
	- Fix some comments of variables.

*eix-0.8.0
	Martin Väth <martin at mvath.de>:
	- Major change: delayed reference %{VAR} of config vars possible.
	  This way, *FORMAT_* strings can be input/modified more conveniently.
	  Moreover, most DIFF_* variables were dropped, since the distinction
	  whether eix or diff-eix is run can be made with %{*VAR}.
	  Default *FORMAT* strings make now heavy use of these delayed
	  references. In particular, all colors/strings/characters/lines are
	  stored in separate variables and can such be easily customized.
	- Introduce TAG_FOR_* variables (in particular, it is now possible
	  to completely revert to the old behaviour by setting the
	  TAG_FOR_EX_* variables empty and TAG_FOR_LOCALLY_MASKED=[M]).
	- Enable escaping of ':' in the installedversions* string arguments.
	- Bugfix: Put DIFF_NEW_SLOTS into defaults.cc
	- Fix installation date in default FORMAT_COMPACT (the separating "@"
	  was mistakenly removed in the previous release).
	- Do not include braces in default FORMAT_*INSTALL*:
	  This is better put into the installedversions* string arguments.
	- Change FORMAT_VERBOSE default to let installedversions* display
	  everything in a separate line.
	- Tiny bugfix: "and" for "Upgrade and Downgrade" in default color.

*eix-0.7.9
	Martin Väth <martin at mvath.de>:
	- Since this version, it is now strongly recommended to leave
	  *LOCAL_PORTAGE_CONFIG=true, because the "default" stability
	  can now be seen anyway: The output and default coloring was changed
	  to reflect the non-local settings (concerning the coloring, this
	  new feature is optional and can be switched off, see below).
	- Print output such that also the default (non-local) stability
	  can be seen. This is a suggestion of STEDevil, see
	  https://forums.gentoo.org/viewtopic-t-278819-start-325.html
	- Show the keyword flag also for masked versions.
	- Color according to original settings not to local settings;
	  make a possible exception for local masking (*COLOR_LOCAL_MASK).
	- Introduce *COLOR_ORIGINAL to allow the user to retain the
	  old coloring behaviour.
	- Allow 4 further strings for installedversions* to make it possible
	  to have different coloring for set/unset useflags.
	- Corresponding changes in default FORMAT strings.
	- Sort useflags alphabetically.
	- Fix a misleading text in eix-test-obsolete.
	- Fix manpage typos.

*eix-0.7.8
	Martin Väth <martin at mvath.de>:
	- Main changes are improved possibilities of (colorized!) output of
	  installed versions and corresponding changes in FORMAT_* defaults.
	- Show installed versions in FORMAT_COMPACT default.
	- Change output of marked packages/versions in FORMAT_* defaults
	  consistently to (red,1;underlined).
	- Introduce installedversionsshort, installedversionsdate,
	  installedversionsshortdate, and FORMAT_SHORT_INSTALLATION_DATE.
	- Remove DIFF_FORMAT_INSTALL* as this should better be handled in
	  the FORMAT_* strings (which is now possible also for the date).
	- Allow appending stringlist to installedversion* to influence the
	  color output.
	- Make default colors of [I],* reverse, since the only other
	  consistent choice would be to use [I] color for installed versions;
	  this looked bad since bright default was reserved for package names
	  and all other non-reserved choices remind of the "downgrade" color.
	  Change [N] color with -c correspondingly to allow better optical
	  distinction (not in diff-eix: [N] has a different meaning there).
	- Bugfix: Initialize empty search string in constructor of regex.
	  This fixes https://bugs.gentoo.org/show_bug.cgi?id=151327
	- Bugfix: No splitting of short options when they are option
	  parameters, e.g. in "update-eix -o -oo" the second argument
	  is not interpreted as two options of type -o.
	- Option parameters can follow without space. For instance,
	  instead of "update-eix -o test" you can use "update-eix -otest".

*eix-0.7.7
	Martin Väth <martin at mvath.de>:
	- Fix default DIFF_FORMAT_DELETE (fix "unknown property: old*" errors).
	- Introduce FORMAT_INSTALLATION_DATE and DIFF_FORMAT_INSTALLATION_DATE.
	- Introduce FORMAT_INSTALLED_USE and DIFF_FORMAT_INSTALLED_USE.
	- Reformulate description of cache methods on manpage.
	- Internal Change: Use InstVersion (inherited from BasicVersion) to
	  manage installed versions (to store e.g. installation dates).

*eix-0.7.6
	Martin Väth <martin at mvath.de>:
	- Early release to fix two problems:
	- In the default setting -T had false non-matches (the +/- prefix
	  in REDUNDANT_IF_* had falsely an effect also for some other tests).
	- The structure in the remote overlays database was changed.
	  To be more independent of such changes, update-eix-remote now
	  takes more care about filenames and adds all overlays of the
	  fetched data (although currently this adds a package too much).
	- Add eix-test-obsolete script to give a structured eix -tTc output.
	- Introduce REDUNDANT_IF_DOUBLE_{USE,CFLAGS}.
	- Fix manpage typo (thanks to Mathieu Bonnet).

*eix-0.7.5
	Martin Väth <martin at mvath.de>:
	- By the new default, now all matches of /etc/portage/package.*
	  to non-installed packages are considered as redundant.
	  Also /etc/portage/package.cflags is now checked for redundancy
	  to the same extent as /etc/portage/package.use.
	  This was possible with the following three additions:
	- Introduce REDUNDANT_IF_IN_* to match any entry in
	  /etc/portage/package.*.
	- Introduce +/- prefix for string in REDUNDANT* to test the redundancy
	  only for installed/uninstalled packages.
	- Extend -t to contain also /etc/portage/package.cflags.
	- Let diff-eix by default list all versions if no stable exists.
	- Enable default options for eix-sync (by config-file and environment).
	- Fix eix-sync -s.
	- Let eix-sync -u imply -s "".
	- Let eix-sync not call update-eix if this was done in an emerge hook.
	- Fix tempfile handling of eix-sync if mktemp does not exist.

*eix-0.7.4
	Martin Väth <martin at mvath.de>:
	- Add support for sqlite if portage uses the sqlite backend
	    https://gentoo-wiki.com/TIP_speed_up_portage_with_sqlite
	  Introduce very fast cache method "sqlite" which is available if
	  eix is compiled with ./configure --with-sqlite
	- Fix endless loop and wrong sanitize in update-eix-remote.
	  Thanks to Stefan Schweizer <genstef at gentoo.org> for the report.
	- Allow different server/client PORTDIR directory for eix-sync.
	- Introduce --dump-defaults and fix output of --dump so that
	  commented variables containing newlines are commented in each line.
	- Improve error-handling of update-eix (continue as far as possible).

*eix-0.7.3
	Martin Väth <martin at mvath.de>:
	- Improved determination of "virtual" overlays.
	- Introduce KEEP_VIRTUAL and let it default to true.
	- Introduce CACHE_METHOD and ADD_OVERRIDE_CACHE_METHOD
	  and change override-order:
	  CACHE_METHOD < ADD_CACHE_METHOD < KEEP_VIRTUAL <
	  OVERRIDE_CACHE_METHOD < ADD_OVERRIDE_CACHE_METHOD
	- Let update-eix-remote default to temporary files:
	  With KEEP_VIRTUAL, a "permanent" remote cache is not really needed.
	- Drop permissions for update-eix-remote when fetching.
	- Enable much more customization for eix-sync:
	  rsync from/to other hosts, various hooks, keep old database.
	  Allow almost each step (gensync/emerge --sync/update-eix/...)
	  switched on/off by options.
	- Rename /etc/eix-sync-overlays to /etc/eix-sync.conf,
	  since now hooks etc. are possible in this file.
	- Document update-eix-remote and new details of eix-sync in manpage.
	- It makes now also sense to start eix-sync as a normal user (e.g.
	  for -d or certain other tasks) -> change to normal bin-directory.
	- Better doxygen sourcecode documentation.

*eix-0.7.2
	Martin Väth <martin at mvath.de>:
	- urgent version bump to fix a segfault of update-eix -m
	  when ADD_CACHE_METHOD and OVERRIDE_CACHE_METHOD are both empty
	  (this happened for update-eix-remote if no configuration was used).

*eix-0.7.1
	Martin Väth <martin at mvath.de>:
	- Let cache method "none" substitute variables;
	  fix segfault in substitutions for cases like A=${X-$ARCH}
	- Previous cache method "none" is now called "none*".
	- Define standard ebuild standard variables like $P or $PN;
	- Add cache methods "ebuild" and "ebuild*" to scan properly e.g.
	  SLOTS of typical kde-base packages.
	- Better treatment of overlay eclass (in particular for ebuild*).
	- Let none, none*, ebuild, ebuild* continue after errors.
	- Add option --output to updatedb - can now run with minor permissions.
	- Improved handling of database file permissions.
	- Improved option parsing.
	- Added --override_method option; eliminated ADD_LOCAL_CACHE_METHOD.
	- Avoid memory leak in pushback_files().

*eix-0.7.0_beta
	Martin Väth <martin at mvath.de>:
	- Major changes: remote support and full support for slots.
	  For the former, see
	  https://bugs.gentoo.org/show_bug.cgi?id=144200
	  the idea for the latter is due to a discussion with
	  caster at matfyz.cz
	  All changes in detail:
	- Add OVERRIDE_CACHE_METHOD, ADD_LOCAL_CACHE_METHOD, ADD_CACHE_METHOD
	  to allow e.g. cache method metadata for a specific overlay.
	- Add cache method eix to use "pre-indexed" data for a
	  specific overlay.
	- Add cache method eix* to use "pre-indexed" data even with possibly
	  new categories.
	- Make eix database format architecture-independent.
	- Detect overlays with equivalent paths as identical.
	- Introduce COLOR_VIRTUALKEY
	- Add update-eix-remote.
	- Distinguish old/new overlays in diff-eix.
	- Fix manpage bug: best=bestlong.
	- Introduce COLOR_SLOTS, *COLORED_SLOTS, *COLON_SLOTS
	- Introduce --versionsorted, STYLE_VERSION_SORTED.
	  Let the defaults be false -> different default output.
	- Introduce bestversions* format.
	- diff-eix outputs package also if best versions change in some slot.
	- Implement -u option.
	- Introduce UPGRADE_ALWAYS_LOCAL.
	- Introduce DIFF_ONLY_INSTALLED, DIFF_NO_SLOTS.
	- Introduce formats old* and new* for DIFF_FORMAT_CHANGED.
	- Introduce formats *better, *worse, *differ for DIFF_FORMAT_CHANGED.
	- Introduce formats *upgrade, *upgradeorinstall, *downgrade,
	  *recommend, *recommendorinstall to visualize upgrade recommendations.
	- Change default formatstrings to use new features.
	- Introduce UPGRADE_ALWAYS_LOCAL, RECOMMEND_ALWAYS_LOCAL.
	- Introduce UPGRADE_TO_HIGHEST_SLOT.
	- Introduce DEFAULT_FORMAT.
	- Implement just-in-time reading of slots of installed versions.
	- Implement --quick, --care. Introduce *QUICKMODE, *CAREMODE
	- Implement --quiet for diff-eix. Introduce *QUIETMODE.

*eix-0.6.4
	Martin Väth <martin at mvath.de>:
	- Introduce slot support.
	- Fix memory leak for source command in varsreader.
	- Fix setting of package flags in all cache methods
	  (required splitting of addVersion into two functions).
	- Introduce PRINT_SLOTS and DIFF_PRINT_SLOTS.
	- Introduce options -i, -1, -2, -O.
	- Output nonexistent installed package names with -t.
	- DEFAULT_IS_OR defaults to "false" again.

*eix-0.6.3
	Martin Väth <martin at mvath.de>:
	- Rename EMPTY_IS_BAD -> TEST_FOR_EMPTY.
	  Sorry for that: eix-0.6.2 was not yet intended to be released.
	  (The release happened only by a misunderstanding).
	- Do no install test-programs.
	- Introduce PRINT_COUNT_ALWAYS.
	- Introduce TEST_FOR_NONEXISTENT, NONEXISTENT_IF_MASKED.
	- Introduce TEST_FOR_REDUNDANCY.

*eix-0.6.2
	Martin Väth <martin at mvath.de>:
	- Correct scandir handling according to ./configure in pushback_files.
	- Introduce EMPTY_IS_BAD and corresponding test in -t.

*eix-0.6.1
	Martin Väth <martin at mvath.de>:
	- Reflect portage's change of ignoring hidden files/dirs
	  for recursive reading, see
	  https://bugs.gentoo.org/show_bug.cgi?id=137607
	- Reflect portage's behaviour of reading files/dirs in sorted order
	  for recursive reading.
	- Introduce OVERLAYS_LIST to display only overlays that are
	  really needed.
	- Option --pipe to parse output from emerge -pv (similar to genlop -p)
	- Introduce DEFAULT_IS_OR and let it default to "true".
	  This changes the default behaviour when several arguments are given.
	- Document that variables may be overridden by environment
	- Introduce -d and DUP_*_ONLY_OVERLAYS, see
	  https://bugs.gentoo.org/show_bug.cgi?id=142469
	- Read profile/package.mask of overlays, see
	  https://bugs.gentoo.org/show_bug.cgi?id=142503
	- Mark packages and versions passed with --pipe, see the discussion in
	  https://bugs.gentoo.org/show_bug.cgi?id=142469
	- Introduce "marked" format and "MARK_VERSIONS" to customize the
	  marking of packages and versions passed with --pipe.
	- Change color format to allow also e.g. "underline" in the formats.
	- Change --dump to display current variables (but still defaults).
	- Introduce "markedversions" format.

*eix-0.6.0_beta
	Martin Väth <martin at mvath.de>:
	- Implemented -T to help cleaning up /etc/portage/package.*.
	- Read package information for the newest available versions:
	  https://bugs.gentoo.org/show_bug.cgi?id=138629
	- Let diff-eix respect local setting (DIFF_LOCAL_PORTAGE_CONFIG):
	  https://bugs.gentoo.org/show_bug.cgi?id=140555
	- Document that it is possible to exclude the PORTDIR "overlay".
	- Add --add-overlay (ADD_OVERLAY variable) to update-eix.
	- Fixed passing of recursive flag for /etc/portage/package.*
	  This should fix https://bugs.gentoo.org/show_bug.cgi?id=141818
	- Treat also non-arch Keywords, e.g. x86 on an amd64 architecture.
	  To avoid increasing database size by factor ~1.5 for this feature,
	  you can use the new SMALL_EIX_DATABASE variable.
	- Introduced REDUNDANT_* to allow high customization for -T
	- Implemented -t to help further in cleaning up /etc/portage/package.*

*eix-0.5.6
	Martin Väth <martin at mvath.de>:
	- If /etc/portage/package.* is a directory, read it recursively:
	  https://bugs.gentoo.org/show_bug.cgi?id=137784
	- patch from caster at matfyz.cz to treat version numbers as long ints:
	  https://bugs.gentoo.org/show_bug.cgi?id=133732
	- Support a "source " (and ". ") command in /etc/make.*:
	  https://bugs.gentoo.org/show_bug.cgi?id=132874
	- Mark installed versions in output list. This experimental feature
	  can be disabled with MARK_INSTALLED=none in /etc/eixrc
	- Fixing -D again.

*eix-0.5.5
	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fix fstream.ignore-stuff in metadata and flat.
	- Redone a part of argsreader.h Option .. looks *much* cleaner now.
	- Fixed https://gentooexperimental.org/eix/ticket/3
	  (was previously https://dev.croup.de/proj/eix/ticket/3)
	  profiles can contain profiles/categories
	- fuzzy-search is now ignore-case.

*eix-0.5.4

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Introducing eix::ptr_list and eix::ptr_iterator.
	- Cleaning headers: no using namespace.

*eix-0.5.3

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Added metadata as cache-method.
	- Added a workaround for https://bugs.gentoo.org/show_bug.cgi?id=114628
	  Looks like the problems isn't eix, but some part of the toolchain.

*eix-0.5.1

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fixed trim/rtrim/ltrim: strings that only contain trim-characters are
	  cleared now. This fixes all the 'Unable to split mask "" ..' bugs.
	- portage-2.1 introduced big changes to the depcache-handling, the old
	  cache is no longer removed. Old entries are not removed.
	  But update-eix needs the cache to be ok, thus eix-sync removes the
	  old depcache in /var/cache/edb/dep.
	- More verbose output if problems with files in /etc/portage/ or the
	  profile pop up. This should stop reports like
	  https://bugs.gentoo.org/show_bug.cgi?id=116592
	- Fixed https://bugs.gentoo.org/show_bug.cgi?id=117407 and removed a
	  compilation-error for -pedantic -ansi -W -Wall -Werror
	- Moved the code in ExplodeAtom from regex to short, _fast_ and
	  hand-written parser-code.
	- Added a compare(..) method to basicversion. This compare-method is
	  used in all other comparators, thus decreasing LOC and a bit more
	  efficient.
	- Rewrote big parts of Mask; Be more verbose about parse-errors;
	  Correctly parse wildcard-stuff; Remove regular-expression stuff;
	- Default cache for PORTDIR_CACHE_METHOD can be chosen by
	  configure-parameter.

*eix-0.5

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fixed a small bug regarding -o/-a and atoms with a pattern ..
	  thanks to leonie for bringing this to my attention.
	- Added PROVIDE; bumped db-version
	- Fixing https://bugs.gentoo.org/show_bug.cgi?id=111927
	- Added our own strndup-implementation that will be used for ppc-macos
	  See https://bugs.gentoo.org/show_bug.cgi?id=111912
	- Finally, environment variables can be used to override eixrc-stuff.
	- Applied patch from https://bugs.gentoo.org/show_bug.cgi?id=111865,
	  thanks JeR
	- Applied patch from https://bugs.gentoo.org/show_bug.cgi?id=112776
	- Applied patch from https://bugs.gentoo.org/show_bug.cgi?id=112806
	- Applied patches for scandir, sed, byteorder and regex from
	  https://bugs.gentoo.org/show_bug.cgi?id=111912
	- Fixed some small memory-leaks - valgrind reports one more that leads
	  into the glibc (getpwnam) .. but this is really a leak. See
	  --run-glibc-freeres in valgrind(1).
	- Rewrite BasicVersion (much more efficient) and fixes
	  https://gentooexperimental.org/eix/ticket/1
	  (was previously https://dev.croup.de/proj/eix/ticket/1)
	- Fix https://bugs.gentoo.org/show_bug.cgi?id=113735

*eix-0.5_beta

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Formatstring(Conditions): If a "else" was missing, the if-true part
	  was linked to if-false.
	- Changed the arch-part of the dbheader from char[16] to string.
	- Fixed https://bugs.gentoo.org/show_bug.cgi?id=104083 - once again:
	  unsigned int vs. vector::size_type
	- Moved from #define in basicio.h to a namespace io with generic
	  read/write-functions for POD and {read,write}_string for strings.
	  Slowly moving other read/write-code into that namespace.
	- Bumped to 0.3.1
	- eix-sync hides output of emerge --sync and update-eix per default.
	  Added cli-options -v, -h, -q
	- Fixed a non-existing bug :)
	  https://bugs.gentoo.org/show_bug.cgi?id=97462
	- Fixed a rather *extrem* bug:
	  The read-only-what-we-need-to-check-if-the-package-matches
	  wasn't used at all! This brings an extrem speedup.
	- Fixed https://bugs.gentoo.org/show_bug.cgi?id=105935
	- I just love auto_ptr!

*eix-0.3.0

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Moved explodeShit(..) into a class with static members. Changed
	  RegExClass to Regex and changed some internals. Category was merged
	  into PackageDatabase. Removed 'this->' and ; after methods.
	- After having a nice talk with Kyuu, I added
	  LOCAL_PORTAGE_CONFIG=true/false to eixrc. It's possible to tell eix
	  to ignore all local configuration like /etc/portage and
	  ACCEPT_KEYWORDS.
	  Yes, I know that this could be done by a alias in $SHELL.
	- Fixed (think) https://bugs.gentoo.org/show_bug.cgi?id=103836.
	  offset_type wasn't off_t everywhere.

*eix-0.3.0_beta
*eix-0.2.7

	Emil Beinroth <emilbeinroth at gmx.net>:
	- PortageInstalledPackages now catches a exception thrown by
	  BasicVersion and prints the message rather then dying.
	  For example, lockfiles in /var/db/pkg are no longer lethal for eix.
	- Did the same to Mask parsing (don't die, but print an error)
	- Added target and gcc-version to the output of both, eix -V and
	  update-eix -V.
	- Fixed a bug in the cdb-code: munmap was called for bad pointers.
	- Fixed https://bugs.gentoo.org/show_bug.cgi?id=97914 (Thanks to
	  David Held <andukar at andukar.org> for the patch.
	- Fixed compilation-problems
	  https://bugs.gentoo.org/show_bug.cgi?id=98218.
	  Patch by Cyrille Mescam <titi at oboulo.com>.
	- Masks like '>app-portage/eix-0.2*' are finally accepted by
	  update-eix.
	- Finally changed our formatstring-code! We are no longer parsing the
	  string every time we need to print something. It's now parsed into a
	  list of structs which is much faster processed.
	- Fixed compile-errors about undefined flags from varsreader once and
	  for all!
	- Started working on diff-eix (part of our 'esync' functionality)
	- More progress with diff-eix - currently WIP (work in progress)
	- Added IF-ELSE-FI constructs to the formatstrings (this could be the
	  final state of my featuritis ..)
	- Rewrote our stability-calculation-thingy, KEYWORDS="-* ~ARCH" is now
	  reported as TESTING (yellow color)
	- Read /etc/portage/package.{unmask,mask} on-the-fly
	- Honour ACCEPT_KEYWORDS correctly and started working on code for
	  package.keywords
	- Finished package.keywords code and basic diff-eix stuff.
	- Added sbin-script eix-sync. It's our esync!

*eix-0.2.6

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fixed an old bug in our parameter-parsing. The first flag after an
	  implicit OR was ignored.

*eix-0.2.5

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fixed https://bugs.gentoo.org/show_bug.cgi?id=88148
	- Committed reviewed manpage (Thanks Legoguy)
	- Overlay reading: Check result of explodePkgFilename before using it
	- Added --version/-V options
	- Added patch from https://bugs.gentoo.org/show_bug.cgi?id=89376
	  by Mike Gardiner <obz at gentoo.org>
	- Rewrote big parts of update-eix. We now have different classes for
	  different portage cache-backends.
	- Added a more descriptive section about ~/.eirx to the manpage.

*eix-0.2.2 (2005/03/26)

	Wolfgang Frisch <xororand at users.sourceforge.net>:
	- Applied patch: gentoo releases == 0 now are accepted & changed
	  metadata cache directory to /var/cache/edb/dep/$PORTDIR

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Made the fsm recognize references to variables in declarations ..
	  i.e. USE="$USE -kde".
	  This is important for proper make.conf/make.defaults parsing.
	- Created CascadingProfile as a subclass of PortageSettings
	- Added maskings from profile
	- Added system-profile-stuff (somewhat depends on virtual/ packages ..)
	- Removed split of category into rootcat-subcat
	- And some other pretty radical changes
	- Rewrote args.cc (OOP'ish :) and added bash-completion
	- User root no longer needed to update database
	- Moved some functionality into separate toolkits.
	- Added ~-operator to class Mask

*eix-0.2.1 (2005/02/24)

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Removed deprecated Constructor from ExBasic and changed all calls
	- Removed big ugly patternmachting from portagedb.cc and moved to
	  criteria.cc (now even more ugly :) -- added wildcards.cc for
	  wildcards-matching.
	- Moved DBQuery from portagedb.cc to filedb.cc
	- Moved PortageInstalledPackages to portageconf.cc and made it
	  pseudo-caching
	- Removed getopt-stuff from main.cc -- this is now recursively done in
	  criteria.cc (with help from args.cc )
	- Various other small cleanups/fixes ..
	- Added --dump, --quiet and --dup-versions;
	  Use xororand formatstring.cc for output of results

	Wolfgang Frisch <xororand at users.sourceforge.net>:
	- Added a wiki: https://eixwiki.unfoog.de
	- Added parameter -l,--lineversions for printing versions line-by-line
	  like "etcat -v"
	- Added format strings to allow customized output. See manpage and
	  https://eixwiki.unfoog.de/index.php/Usage#Format_strings
	- Yet another 'version number sorting'-fix. Now: 1.2a < 1.2b
	- Moved Portage settings parts from PortageDB to new class
	  PortageSettings

*eix-0.2.0

	Beholders_Eye (Gentoo Forums)
	- Fixed two warnings

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Fixed various bugs

	Wolfgang Frisch <xororand at users.sourceforge.net>:
	- 'eix -u' crashed when it met files like abc.ebuild.xyz in overlay
	  package directories. This has been fixed.

*eix-0.2.0_beta (2005/02/12)

	Wolfgang Frisch <xororand at users.sourceforge.net>:
	- Fixed a reported bug in version.cc: Some package versions were
	  omitted in eix-0.2.0_alpha and probably in earlier versions too.
	  Example: A package with 2.1 and 2.1.1 only showed 2.1.
	- Fixed a small issue with handling files like "abc.ebuild.xyz" in
	  portage overlay directories

	Kay Ramon "YakS" Sauter <yaks-gentoo at users.sourceforge.net>:
	- Added man page

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Included minor fixes from
	  Christian Heim <christian.th.heim at gmx.de>
	- Big cleanups in readoverlay to fix 2 major bugs
	  https://bugs.gentoo.org/show_bug.cgi?id=80241 and
	  https://forums.gentoo.org/viewtopic-p-2005152.html
	- Added -I to search only installed packages. If no search pattern is
	  given it prints all installed packages.
	- Changed version.cc, added a test for version.cc, added 'keys' for
	  packages/versions from overlays and moved cleaned some classes(moved
	  attributes to private/protected).

*eix-0.2.0_alpha (2005/02/01)

	xororand <xororand at users.sourceforge.net>:
	- Fixed a "versions reported twice"-bug
	- Added --fuzzy search option to allow unsharp search queries
	  (no regular expressions).
	- Closed memory leaks

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Constructed a FSM (finite state machine) to replace the
	  regex-massacre in varsreader.cc.
	- Fixed the "-S does not search the name, just the description"-bug and
	  added warnings if keyUndefined occurs while updating the db
	- Rewrote package::getStability(..) so it can distinguish between ppc
	  and ppc64, x86 and x86-obsd .. and so on
	- Converted the db-read/write code to use c-style FILE*, fread, fwrite.
	  This fixes the ppc-problem described in
	  https://forums.gentoo.org/viewtopic-p-2036103.html
	  Thanks meyerm for testing :)

	xororand <xororand at users.sourceforge.net>:
	- Fixed: packages with ebuilds in multiple overlays are no longer
	  reported twice
	- Increased loading speed of the database

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Various updates to autoconf-stuff including --with-cachefile=FILE for
	  configure and update of VERSION in configure.in
	- Show "file:line" when throwing a exception
	- Rewrote get_arch()

*eix-0.1.4 (2005/01/22)

	xororand <xororand at users.sourceforge.net>:
	- Added: multi-line entries in ebuild and/or /etc/make.conf are parsed
	  variables declared without double quotation marks are recognized
	  correctly as well as tab characters in PORTDIR_OVERLAY
	- Versions which are not tested on your architecture (arch not in
	  KEYWORDS) are now shown by default (previously: only with -v)
	  I changed the prefix for such versions from [M~] to [-]

	Emil Beinroth <emilbeinroth at gmx.net>:
	- Cleaned the command line parser to use the quasi-standard getopts()
	  which is used by almost any Unix program.
	  eix -s is now default, so you can simply type "eix REGEXP" to search
	  for packages.

	xororand <xororand at users.sourceforge.net>:
	- added support for /etc/portage/categories

*eix-0.1.3 (2005/01/17)

	xororand <xororand at users.sourceforge.net>:
	- Fixed bug: multiple spaces in PORTDIR_OVERLAY (/etc/make.conf) were
	  interpreted as / overlays
	- Fixed: Version numbers should finally be sorted properly.
	  Explanation: In 0.1.2 versions were only splitted in [primary, e.g. |
	  1.2.3][suffix | _alpha5][gentoo-release | -r2]
	  Now, the primary string is also splitted into integers =>
	  2.6.2 < 2.6.10
	- Added more detail to the error messages in version.cc (in hopes to
	  resolve kay's problem)

*eix-0.1.2 (2005/01/13)

	xororand <xororand at users.sourceforge.net>:
	- Bekks removed the unnecessary libstdc++v3 dependency from the ebuild
	- The currently installed version of each package is now displayed
	  again
	  Note: Slotted versions are still not supported
	- Corrected a bug of 0.1.0: All versions are displayed again
	- The -v parameter now also shows "arch masked" package versions, e.g.
	  [M~]1.2.3-r4
	- Reverted to the old get_arch code because of an error on an amd64
	  system
	- Fixed a small bug in command-line parsing

*eix-0.1.0 (2005/01/09)

	xororand <xororand at users.sourceforge.net>:
	- Renamed portagedb to eix

# vim:set ts=8 tw=78 noet:
