- Initial release (to help doing apt-BuildPreReq-under-pkdirect-checkinstall
  in repositories where the default build env contains external APT sources).

- Clear APT's sources before the test for repositories where
  there are uncommented external sources in the default build env (like e2k).

- tests:
  + Enhanced to be able to work with test packages containing some files.
    (This will be useful for testing the work with large RPM archives.)
  + Shortened the run time of apt-xxtra-heavy-load-checkinstall.
  + Shortened the run time of other *-checkinstall scripts
    (by shortening the the test with zillion packages).

- Fixed bad behavior when satisfying "Conflicts" or "Obsoletes" dependencies
  (discovered in ALT#42415), namely:
  + "Obsoletes" causing the deletion of its "virtual" targets (i.e.,
    providing packages) whereas this behavior is expected for Conflicts only;
  + versioned "Conflicts" (or "Obsoletes") causing the deletion of packages
    with non-matching version.
- Enriched the output of Debug::pkgMarkInstall with the versions of
  the dependency targets and the targets being considered during the search.
- Complemented it with a new option (Debug::pkgMarkAllCalls) -- to understand
  better why the result of resolving a broken dep is unexpected in some cases.

- Rewritten a tiny piece of code to fix build with lcc.
- Adapted to ALT's apt API a bit further by dropping an unused parameter of
  MarkInstall(). (In apt, it either has a default value or will be eliminated.)

- Fixed the new compilation warnings by simplifying the ALT-specific code
  in pkgcache.cc (AllTargets) from 0.5.15lorg2-alt31. (Discovered by gcc12.)

- Invalidate the in-memory cache of repositories when doing "update" or
  ListUpdate() to be able to detect updates without exiting the process,
  e.g., PackageKit or apt-shell (ALT#41816).
- Added apt-under-pkdirect-checkinstall subpackage
  and added a test for FileList() API via packagekit.

- Added an URL.

- Added an URL.

- Added an URL.

- Added an URL.

- Added an URL.

- Added an URL.

- Implemented generic callback system for packagekit, allowing to show progress
  during offline-update to user. (Thx Oleg Solovyov mcpain@)

- Show actual update percentage during offline update

- Adapted to the new API with custom callbacks introduced
  in libapt-0.5.15lorg2-alt74. (Surprisingly, the call that had to be adapted
  is a method "private" to the apt library used here by means of a hack;
  so libapt shouldn't really bother to provide a compatibility method.)

- Initial release for ALT Sisyphus (specially for
  testing packages on e2k instead of Girar's install checks).

- Initial release for ALT Sisyphus (specially for
  testing packages on e2k instead of Girar's install checks).

- Initial release for ALT Sisyphus (specially for
  testing packages on e2k instead of Girar's install checks).

- Initial release for ALT Sisyphus (specially for
  testing packages on e2k instead of Girar's install checks).

- (tests) Report if a test (marked XFAIL) uneXpectedly passes (XPASS).
- (tests) Run them not in %check, but in *-checkinstall subpkgs. (To break
  build-dep cycle with apt-repo-tools, whose features are required by the tests,
  but which needs to be recompiled to be linked with libapt.)
- (tests) Done more extensive testing of how apt works with "rpm" repos
  via any of the file, http(s), cdrom methods; including:
  + re-fetching if the saved complete or partial pkglist indices are corrupt
    (see https://bugzilla.altlinux.org/show_bug.cgi?id=40746#c9 );
  + the verification of the checksums of pkglist indices. (The verification
    is tested in two ways:
    * The verification of the checksum of a specific type is tested by faking
      it in the meta-data: for MD5, SHA1, SHA256, BLAKE2b and just the size.
    * Simply testing that a faked pkglist file of the same size is rejected--no
      matter which hashing algorithm is used.)
  + the verification of the checksums of rpm archives. (The verification
    is tested in two ways:
    * The verification of the checksum of a specific type is tested by faking
      it in the meta-data: for MD5, SHA1, SHA256, BLAKE2b.
    * Simply testing that a faked rpm file of the same size is rejected--no
      matter which hashing algorithm is used.)
- (source code; ABI) Reverted a lot of inessential optimizations
  from 0.5.15lorg2-alt72.
- (source code; ABI) Got rid of virtual methods with default parameters
  (because they are confusing for the programmer).
- (source code; ABI) Backported some pieces of the support for the multiplicity
  of checksum (and compression) types from apt-rpm (thx imz@):
  + the type of the compression for "pkglist" indices;
  + the type of the checksum for "pkglist" indices;
  + the type of the checksum for "rpm" archives.
- Added blake2b hash support (thx glebfm@).
- Changed file and copy download methods to always compute checksums
  (thx glebfm@).

- Some fixes in rpm-spec-mode:
  + its filename pattern (the .spec extenstion) now matches only at the end.
  + initialize a new spec-file with no trailing empty lines (in the changelog).
  + rpm-add-change-log-entry(): fixed inconsistency in the order of
    the items added to a single changelog entry.
  + in a new spec-file, initialize version, too
    (for correct initialisation of the changelog).

- Cleaned up the code (thx Dmitry V. Levin ldv@; including
  quite a few commits cherry-picked from http://apt-rpm.org/scm/apt.git):
  + to avoid compilation warnings altogether and some unreliable code;
  + to avoid using any old deprecated RPM API.
- Reverted (for a while) new features with unreliable implementation introduced
  in 0.5.15lorg2-alt70 (dynamic resizing of allocated memory; some support
  for large files). Updated how the other changes look in the history
  (thx darktemplar@). (The soname has been bumped again.)
- API changes:
  + Reverted inessential optimizations that caused incompatibilities with
    the Debian API (introduced in 0.5.15lorg2-alt70).
  + Made pkgCacheFile class lazy and immutable so that it better suits
    the expectations of modern libapt clients such as PackageKit
    and so that it is less prone to memory leaks and other programming errors.
  + And changed some other things (how functions return results) to avoid
    programming errors (which lead to the NULL dereference bugs listed below).
- Fixed some recently introduced and recently discovered bugs:
  + APT now can handle packages without ARCH tag (such as gpg-pubkey, brought
    by 3rd-party packages) without a crash (thx darktemplar@ et al)
    (ALT#38381, ALT#38642).
  + Some crashes with incomplete indices (after the old apt-cdrom or for
    incompatible arch).
- Increased the default APT::Cache-Limit (up to 192M)
  to make the "out of space" failure less probable for packagekit.

- Some changes from the previous release reverted or updated, since we recently
  reverted and updated some changes in the APT API in apt-0.5.15lorg2-alt72.

- Adapted to changed API in apt-0.5.15lorg2-alt72
  (pkgCacheFile class in RAII style).

- genbasedir/genpkglist: fixed generation of non-flat repositories.
- genbasedir (imz@):
  + Fixed SIGPIPE in the old non-patching mode. (It was due to a race.)
  + Fixed patching srclist in-place when there is no uncompressed srclist.
  + Fixed patching a basedir if there is a spurious {pkg,src}list.*.old file
    there, but no uncompressed {pkg,src}list.*.
  + Fixed a potential bug (due to a race) of choosing the overwritten
    output file as input for patching.
  + Report an error when patching and the previous list can't be found.
  + Do the search for the previous list the same way no matter whether
    --prev-basedir was specified or patching "in-place".

- Reverted some changes from the previous release, since we recently reverted
  some changes in the APT API in apt-0.5.15lorg2-alt72;
  and adapted to the current API (including pkgCacheFile class in RAII style).
- Fixed some unreliable/questionable code and added stricter compiler checks.

- Adapted to changed API in apt-0.5.15lorg2-alt72
  (pkgCacheFile class in RAII style).
- (.spec) Just added some compiler flags (which don't change anything)
  to be sure that APT API has not been used wrongly and will not be.
  (Namely, to be sure that if a method override was intended, it would
  actually be overriding and not hiding a virtual method and that it
  would be marked "override" for future.)

- Adapted to changed API in apt-0.5.15lorg2-alt72
  (pkgCacheFile class in RAII style).
- This also fixed a memory leak/dangling pointers to the caches.
  (The old implementation of pkgCacheFile in apt was "wrong".)

- Fixed /usr/lib/packagekit-direct (that didn't work, because
  it couldn't load the APT backend).

- Fixed a use-after-free bug (appearing as garbage being shown in the size
  columns, notably after recompilation with gcc10). (ALT#40010)
- Unset Synaptic::supported-text in the packaged configuration file. (That's
  the straight way to get "Package is supported" translated in the UI rather
  than the previous hack in the code.)
- Included the sr@latin translation in the package.
  (It got excluded before because of an outdated sr@Latn name.)

- Just added some compiler flags (which don't change anything):
  + to adapt the C++ standard for the upcoming APT API (increased to C++17);
  + to be sure that APT API has not been used wrongly and will not be.
    (Namely, to be sure that if a method override was intended, it would
    actually be overriding and not hiding a virtual method and that it
    would be marked "override" for future.)

- Fixed copying release information from cdrom (thx Aleksei Nikiforov).
  (Closes: #37531)

- Initial build for ALT Linux Sisyphus (from a gist by Jakub Jirutka).
- Enable creation of an unprotected key (if -p or $GPG_PASSPHRASE is empty).

[Restored some patches from 4.0.4-alt94]
- rpmrc.c: recognize new Intel CPUs (Dmitry V. Levin)
- rpmrc.c: classify SSE2-capable Intel CPUs as "pentium4" (Alexey Tourbin)
  [Restored a patch from 4.0.4-alt98.26]
- rpmrc.c (is_pentiumN): Added models with nonzero extended model
  (Dmitry V. Levin; reported by Alexander Sharapov).
  (Closes: #38708)

- Removed the buildtime nprocs value from the *-checkinstall package scriptlets:
  it would make no sense at runtime.

- Fixed a bug in ARM hardfloat detection (armv?h* archs), introduced
  in 4.13.0.1-alt22, which would prevent normal installation of armh
  packages on such machines. (Note that rpm-build-4.0.4-alt141 is still
  ignorant of armv?h* arches, so the only way to build packages installable
  on such machines is with --target armh.)

- Run additionally a test with a package built for system rpm's ARCH
  (i.e., with rpmbuild --target set to it).
- Ignore an arch_compat test on ARM (the one introduced in 1.1.3-alt5:
  with the default arch on this machine) until better times, because
  the ARM arch detection in rpm-build (<= 4.0.4-alt141) is not ideal yet.

- Improved machine detection & configuration for:
  + x86_64 on Darwin, armv5tl, Transmeta Crusoe as i686, parisc as hppa (on linux)
    (cherry-picked from upstream);
  + armv8[h]l (further refined by me on the base of changes from upstream and
    Peter Robinson).
  (Accompanying macros have been added, too.)

- Added /usr/lib/rpm/armv8l-alt-linux/macros for builds on armv8l machines;
  added armv8l to the %arm list and optflags for it for builds targeting it.
  (Fixes 4.0.4-alt108:
  - installplatform, rpmrc.in: made armv8l compatible with armh.)

- Also test the compatibility between `uname -m`, rpmbuild, and rpm -i.
  (A separate archcompat-checkinstall subpkg does this.)
- First, run a simple test (to fail fast).

- Collected new information (of types: fix, mention)
  from the tasks/changelogs (of branches: c7, c8).
  (gear-cronbuild-apply-hooks at your service!)

- Fixes:
  + OVE-20200317-0001 excessive/unoptimal memory allocation for hashtables due to read-beyond-end

    (Thanks svace & coverity static analysis,
    https://lists.gnu.org/archive/html/guile-devel/2011-07/msg00069.html,
    backported from 2.0.3.)

- Fixes:
  + OVE-20200311-0001 Normal update operations by root can corrupt/crash running processes through modifying libraries

    (Normally, the processes that can be affected are only the services
    already running in the chrooted environment being updated, if it is on
    a different filesystem than the system libraries, and hence they
    can't be hardlinked. Extraordinarily, if a file in a chrooted
    environment has been a hardlink to a system library, but ln hasn't
    succeeded this time for some reason, the affected processes are any
    already running processes system-wide.

    This could happen only as a result of actions initiated by root:
    for example, a package upgrade which included the old version of
    chrooted, or a package upgrade which included libraries in a system
    with the old version of chrooted, or an invocation of
    /sbin/update_chrooted.)

- initial build for ALT.

- Relocated the contents to /var/www/blog.

- Collected new information from the tasks/changelogs.

- Made BuildReqs more flexible. (So that it builds both in Sisyphus and p8;
  no matter whether "num" is included in core OCaml.)

- Made additional useful part of this font set (Liberation Narrow) be installed
  together with this package always, to make users happier. (By a Requires.
  It used to be part of this package until 2.00.1-alt1, but was separated into
  another package due to license, see https://lwn.net/Articles/502371/ ).
  This addition is a replacement for the Arial Narrow font, which is popular
  in official documents and helps LibreOffice render them nicer.

- Collected new information from the tasks/changelogs.

- Keep keys in /etc/xrdp/ on upgrade and removal. (Important on upgrade
  from <= 0.9.3-alt1, which used to own these files. They would be removed,
  but now this package owns them again, so that they are not removed on upgrade.
  It also marks them as config, so that they are not removed on removal.)

- Fixed build without check.

- genpkglist: added DistTag.
- genbasedir (by glebfm@):
  + added --compressedonly alias for --bz2only option;
  + added --{,no-}bz2 and --{,no-}xz to usage.
- Dropped Vendor tag from {pkg,src}list (by glebfm@).

- really adapt c flags for C++14 (used in the last source changes
  here & in APT in 0.3.12-alt4), so that they really have effect on e2k

- Introduced new function ListUpdate for improved packagekit support.

- Fixed support for refreshCache action. (Thx Aleksei Nikiforov darktemplar@)

- (.spec):
  + Avoided a nasty race in Makefile.
  + Sped up the build by force on non-intel.

- Made the treatment of the File Provides and the version ID of
  installed packages "stable": not affected negatively by sources.list.
  Also negative effects of different pkglists from sources.list on each other
  have been mitigated. (A pkglist from sources.list that lacked disttags could
  shadow the File Provides of packages from the database or other pkglists.)

- (sources) cleanup backup files

- %pre: corrected the old symlink path in 8.2.50.20180917-alt3.

- %pre: rm symlink to upgrade from gdb < 8.2.50.20180917-alt1. This works
  for rpm-4.0.4 (when upgrading from p8 to Sisyphus or p9), not for rpm-4.13.

- Packaged the forgtotten new rpmvercmp.h (whose use is discouraged
  in favor of rpmEVRDTCompare() or rpmRangesOverlap()).

- rpmrdups: honor disttag, buildtime.

- Use rpmquery-strictdep for the dependency on the native package,
  which gives an additional guarantee that the dependency is actually
  provided by the package. To get a strictdep with a beginning dot
  (a deprecated format not used in any ALT repo branches anymore except p8),
  RPM macro %_allow_deps_with_beginning_dot
  or env var $allow_deps_with_beginning_dot has to expand to 1.

- lib: introduced rpmEVRDTCompare() (useful for APT).
- Changes in what is considered "newer" by rpm -U  pertaining to disttag
  comparison. (On the whole, to determine which package is "newer", first,
  the EVRs are compared, then the branch prefixes of the disttags if the
  disttags are present, and then the buildtimes.) The comparison of the disttags:
  + Before the comparison of disttags, an optional initial padding
    (which is terminated by :) is skipped. (This will be useful for
    generating >,<-deps compatible with disttag-unaware rpm & apt.)
  + If a disttag contains no + separator (old format), the branch prefix is
    assumed to be empty (and hence "older" than any other branch prefix).
  + If the branch prefix of a disttag is equal to %_priority_distbranch
    (and it is not empty), then it is "newer" than any other ones.
  + The branch prefixes of disttags are ordered by rpmvercmp() rather than
    lexicographically. (For example, the numeric parts of "p8" or "p10" are
    compared as numbers. However, the first letter in "p7" or "c8.1" is more
    significant.)
- Give a default value to %_priority_distbranch based on the disttag
  when this package is built (the prefix before +),
  i.e., the current repo branch by default.
  (Useful for getting the rpm tool with good behavior in branches (like p9)
  forked off Sisyphus after disttags were introduced.)

- Add disttag to VerStrs (used by APT to identify package versions).

- update-kernel: fixed in the presence of disttags in package IDs
  printed by APT (apt-cache pkgnames).

- Provides: RPMQ(EPOCH) RPMQ(BUILDTIME) RPMQ(DISTTAG)
  (to be required by APT and other tools such as update-kernel, which need
  to query the RPM db by specifying Epoch, Buildtime, and Disttag).

- To catch more errors:
  + upgradable.mk strictly_newer_* helpers:
    a special reason for xfailing just the asymmetry test.
  (The tests failing for other reasons will just fail rather than xfail.)

- To catch more errors (evidenced by file conflicts during rpm -U),
  now the package template includes a file with a unique value.
  (This runs more slowly.) (Suggested by Vladimir Seleznev.)

(thx Vladimir D. Seleznev vseleznv@)
- Added support for dbi matching against DistTag and BuildTime (closes: #36375).
- Display NEVR with DistTag or BuildTime when label, specified to erase, is
  matching to multiple packages.
(thx Gleb F-Malinovskiy glebfm@)
- Fix non-LFS functions on i586 in rpm-plugins/systemd_inhibit.so,
  /usr/lib/rpm/sepdebugcrcfix (thx glebfm@).

- Add buildtime to VerStrs (used by APT to identify package versions).
  This data is used in several manners:
  * by CheckDep() (only when matching a dependency with a real package);
  * rpm_name_conversion() (only when making up an id for a duplicated package);
  * and by *CmpVersion().
  The latter needs buildtime to determine the correct upgrade direction and
  can be called through the API with some externally supplied versions.
  In order to honor buildtime without changing the API and its clients, we pass
  buildtime inside the existing argument. (Also fixes ALT#36528)
- Increase default APT::Cache-Limit in 1.5 times due to the extension of VerStrs
  (ALT#36775).

- build with debuginfo

- genpkglist: added DistTag.
- genbasedir (by glebfm@):
  + added --compressedonly alias for --bz2only option;
  + added --{,no-}bz2 and --{,no-}xz to usage.
- Dropped Vendor tag from {pkg,src}list (by glebfm@).

- (no user visible changes) Removed "interesting" dead code (related
  to Recommends, not present in ALT rpm), which didn't compile after
  we changed apt's API recently.

- Fixed crash during showing errors.
- Enabled verbose logging for packagekit service.

- rebuild with new perl 5.28.1

- NMU: fixed -Wreturn-type warnings.

- remove-old-kernels: APT invocation (to remove installed pkgs) rewritten in
  a more portable way. (A preparation for APT with support for disttags.)