Upstream changes:
1.20     2018-05-06

- Fix test failures on Windows. Reported by Adrian Devries. GH #23.

Add missing DEPENDS

Upstream changes:
0.84 Mon 16 Jul 2018 17:49:10 BST
     - Put back credential cache code lost in Signature 4 patch (Branislav Zahradník)

0.83 Tue 10 Jul 2018 22:30:23 BST
     - All patches by Branislav Zahradník
     - Use Signature 4 by default only for amazonaws host (issue #29)
     - Relaxing constraint on owner id, accepting any string (issue #18)
     - Enable secure by default (issue #23)
     - Sanity object uris to avoid invalid signature (issue #28)

0.82 Sat  7 Jul 2018 14:28:06 BST
     - Use HEAD request to determine bucket region (patch by Branislav Zahradník)

0.81 Thu 28 Jun 2018 20:27:44 GMT (TRIAL RELEASE)
     - Restore CHANGES files format
     - Change path to new github repo
     - Add optional delimiter parameter to
       Net::Amazon::S3::Client::Bucket->list (patch by Christian
       Lackas)
     - Add expected and received ETag value if upload is detected as
       being incorrect.
     - Add support for all location constraints when creating buckets
     - making DateTime coercion optional for last_modified (something that
       might happen million of times without ever been used, in a rather
       expensive operation) (patch by Christian Lackas)
     - avoiding MD5 calculation of large files (chunked uploads) that
       are then never used (patch by Christian Lackas)
     - add ability to set use_virtual_host to use virtual host method
       of making requests which eliminate having to set the region
       endpoint of a bucket. (patch by Christian Lackas)
     - Add V4 support (patch by Branislav Zahradník)

## [v0.4.16](https://github.com/chef/mixlib-archive/tree/v0.4.16) (2018-08-28)

#### Merged Pull Requests
- Fix read_tar_magic to properly return an array when a file has less than 264 characters [#30](chef/mixlib-archive#30) ([bdwyertech](https://github.com/bdwyertech))
<!-- latest_release -->

<!-- release_rollup since=0.4.13 -->
### Changes not yet released to rubygems.org

#### Merged Pull Requests
- Fix read_tar_magic to properly return an array when a file has less than 264 characters [#30](chef/mixlib-archive#30) ([bdwyertech](https://github.com/bdwyertech)) <!-- 0.4.16 -->
- Fix mixlib-archive on Windows, add Appveyor [#31](chef/mixlib-archive#31) ([stuartpreston](https://github.com/stuartpreston)) <!-- 0.4.15 -->
- Ensure that paths like foo..bar.baz aren&#39;t ignored [#9](chef/mixlib-archive#9) ([thommay](https://github.com/thommay)) <!-- 0.4.14 -->
<!-- release_rollup -->

<!-- latest_stable_release -->
## [v0.4.13](https://github.com/chef/mixlib-archive/tree/v0.4.13) (2018-08-03)

#### Merged Pull Requests
- add codeowners [#24](chef/mixlib-archive#24) ([thommay](https://github.com/thommay))
- Move deps to the Gemfile and expand Rake tasks [#25](chef/mixlib-archive#25) ([tas50](https://github.com/tas50))
- Add github issue template &amp; standardize expeditor config [#26](chef/mixlib-archive#26) ([tas50](https://github.com/tas50))
- Update codeowners and add PR template [#27](chef/mixlib-archive#27) ([tas50](https://github.com/tas50))
- bugfix to make dir.chdir threadsafe [#29](chef/mixlib-archive#29) ([stevenoneill](https://github.com/stevenoneill))
<!-- latest_stable_release -->

## [v0.4.8](https://github.com/chef/mixlib-archive/tree/v0.4.8) (2018-06-21)

#### Merged Pull Requests
- Use the correct entry size for content with multi-byte characters. [#23](chef/mixlib-archive#23) ([coderanger](https://github.com/coderanger))
<!-- latest_release -->

<!-- release_rollup since=0.4.7 -->
### Changes not yet released to rubygems.org

#### Merged Pull Requests
- Use the correct entry size for content with multi-byte characters. [#23](chef/mixlib-archive#23) ([coderanger](https://github.com/coderanger)) <!-- 0.4.8 -->
<!-- release_rollup -->

<!-- latest_stable_release -->
## [v0.4.7](https://github.com/chef/mixlib-archive/tree/v0.4.7) (2018-06-07)

#### Merged Pull Requests
- fixup strange edge case around tar magic reading [#22](chef/mixlib-archive#22) ([spion06](https://github.com/spion06))
<!-- latest_stable_release -->

## [v0.4.6](https://github.com/chef/mixlib-archive/tree/v0.4.6) (2018-05-08)

#### Merged Pull Requests
- update tar magic to identify oldgnu style tar headers [#21](chef/mixlib-archive#21) ([spion06](https://github.com/spion06))

## [v0.4.5](https://github.com/chef/mixlib-archive/tree/v0.4.5) (2018-05-04)

#### Merged Pull Requests
- Fix up creating archives [#18](chef/mixlib-archive#18) ([thommay](https://github.com/thommay))
- Fix up writing tar archives with the rubygems tar [#19](chef/mixlib-archive#19) ([thommay](https://github.com/thommay))

## [v0.4.2](https://github.com/chef/mixlib-archive/tree/v0.4.2) (2018-04-25)

#### Merged Pull Requests
- use libarchive by preference [#17](chef/mixlib-archive#17) ([thommay](https://github.com/thommay))

## 2.1.1 (2018-06-01)

* [#23](httprb/form_data#23)
  Allow override urlencoded form data encoder.
  [@FabienChaynes][]

Upstream changes:
(6.75 changelog not found)
6.74  2018-12-01
  -  Removed travis file from the tarball
        The tarball had a .travis.yml file in it that wasn't in the MANIFEST
        file. Since the MANIFEST file is automatically generated by the perl
        mkmanifest tool, since it didn't see the .travis.yml file, I have
        excluded it from the tarball. Reported by Mohammad S Anwar on GitHub
        (#23)

  -  Time zone fixes
        Newest zoneinfo data (tzdata 2018e)

Upstream changes:
2.66 2019-02-11
    - Full release of 2.65_001, with one additional definedness check (pajlpajl++)

2.65_001 2019-02-11
    - fix file_version method (it was calling the wrong key) (GH #23)
    - avoid repeated loading of files when getting headers
      and incorrect undef returns (GH #22)
    - both reported by pajlpajl

* Readd _OPENBSD_SOURCE explicitly to fix build on NetBSD 9

Changelog:
13 August 2019: Wouter
	- Fix error message for out of zone data to have more information.

12 August 2019: Wouter
	- Fix #33: Fix segfault in service of remaining streams on exit.

6 August 2019: Wouter
	- Tag for 4.2.2rc1.

5 August 2019: Wouter
	- PR #31: nsd-control: Add missing stdio header.
	- PR #32: tsig: Fix compilation without HAVE_SSL.
	- Cleanup tls context on xfrd exit.

31 July 2019: Wouter
	- Fix #29: SSHFP check NULL pointer dereference.
	- Fix #30: SSHFP check failure due to missing domain name.
	- Fix to timeval_add in minievent for remaining second in microseconds.

22 July 2019: Wouter
	- Set timeout for refetch immediately, only spread load when there
	  are retries.

19 July 2019: Wouter
	- Set no renegotiation on the SSL context to stop client
	  session renegotiation.

18 July 2019: Wouter
	- Fix #25: NSD doesn't refresh zones after extended downtime,
	  it refreshes the old zones, with a random delay of a couple of
	  seconds to spread the load.
	- Fix so that expired zones stay expired when server is down a
	  long time.

17 July 2019: Wouter
	- Fix that NSD warns for wrong length of the hash in SSHFP records.

15 July 2019: Wouter
	- PR #23: Fix typo in nsd.conf man-page.

4 July 2019: Wouter
	- Set version to 4.2.2 in development.
	- clean memory on exit of nsd-checkzone for memory debug.
	- Fix #20: CVE-2019-13207 Stack-based Buffer Overflow in the
	  dname_concatenate() function.  Reported by Frederic Cambus.
	  It causes the zone parser to crash on a malformed zone file,
	  with assertions enabled, an assertion catches it.
	- Fix #19: Out-of-bounds read caused by improper validation of
	  array index.  Reported by Frederic Cambus.  The zone parser
	  fails on type SIG because of mismatched definition with RRSIG.

2 July 2019: Wouter
	- Tag for 4.2.1rc1

27 June 2019: Wouter
	- Fix unit test for added options and no dot after zone updated
	  log message.
	- Fix compile without accept4.

21 June 2019: Wouter
	- Omit remaining tcp processing if the list is empty.
	- Fix output of nsd-checkconf -h.

20 June 2019: Wouter
	- Initialize event structures before event_set, to stop uninitialized
	  values from setting event library lists and assertions, that would
	  sometimes also show after event_del.
	- Added num.tls and num.tls6 stat counters.
	- PR #12: send-buffer-size, receive-buffer-size,
	  tcp-reject-overflow options for nsd.conf, from Jeroen Koekkoek.
	- Do not use symbol from libc, instead use own replacement, if not
	  available, for accept4.
	- Fix #14, tcp connections have 1/10 to be active and have to work
	  every second, and then they get time to complete during a reload,
	  this is a process that lingers with the old version during a version
	  update.

19 June 2019: Wouter
	- Fix tls handshake event callback function mistake, reported
	  by Mykhailo Danylenko.

18 June 2019: Wouter
	- Fix #15: crash in SSL library, initialize variables for TCP access
	  when TLS is configured.

14 June 2019: Wouter
	- Fix to init event not pointer, in reassignment.

12 June 2019: Wouter
	- Fix to init event structure for reassignment.

Changelog:

New Libraries

    Variant2: A never-valueless, strong guarantee implementation of std::variant, from Peter Dimov.

Updated Libraries

    Align:
        Support for any integral type in the integral versions of align_up and align_down.
        Revised the implementation of aligned_allocator to no longer require a specialization for void.
        Define propagate_on_container_move_assignment and is_always_equal in aligned_allocator.
    Asio:
        Improved performance slightly by eliminating a redundant move construction when completed handlers are dispatched.
        Eliminated a compiler warning by annotating a case fall-through in the free function connect() implementation.
        Fixed the is_*_buffer_sequence detection traits for user-defined sequence types.
        Fixed some Windows-specific warnings about an incompatible pointer cast when obtaining the CancelIoEx entry point.
        Changed to automatically set the defaults when opening a serial port on Windows.
        Changed the serial port get_option() member function to be const.
        Fixed a name hiding issue with the WinRT stream-oriented socket backend's shutdown function.
        Applied a minor fix to the documentation for is_dynamic_buffer.
        Added some support for Haiku OS.
        Added wolfSSL compatability.
        Changed to require C++17 or later for coroutines TS support with clang.
        Fixed a doxygen generation problem in the tutorial.
        Ensured example programs are correctly incorporated into the documentation.
    Any:
        Static initialization of the default constructed boost::any is now possible.
        Fixed performance regression in assignment on a compilers without move-semantics.
        Maintenance work.
    Beast:
        This version consists mostly of bug fixes and performance improvements.
        Substantial work included for the split compilation mode, to reduce compile time when defining BOOST_BEAST_SEPARATE_COMPILATION.
        We'd love to know how you or your company use Beast, consider adding an item to the Companies and Individuals Using Beast list.
        See the full Release Notes for a complete list of changes.
    Circular Buffer:
        Correct doxygen generated reference to no longer show internal members. (Glen Fernandes)
        Eliminate ubsan warning on add and sub. (Glen Fernandes)
        Fix incorrect check in is_uninitialized. (Niklas Fejes)
    Container:
        Fixed bugs:
            GitHub #47: "added alignment specification for small_vector".
            GitHub #88: "Implement C++17 MoveAssignable requirements for self-move assignments".
            GitHub #107: "Alignment ignored in resource_adaptor".
            GitHub #109: "Get rid of integer overflow in copy_move_algo.hpp (-fsanitize=integer)".
            GitHub #110: "Avoid gcc 9 deprecated copy warnings in new_allocator.hpp".
            GitHub #112: "vector::resize() compilation error with msvc-10..12: data is not a member of boost::detail::aligned_storage".
            GitHub #114: "Fix small_vector noexcept specification".
            GitHub #116: "MSVC + boost 1.70 compilation error when windows.h is already included (detail/thread_mutex.hpp)".
            GitHub #117: "flat_map/map::insert_or_assign with hint has wrong return types".
            GitHub #118: "Non-unique inplace_set_difference used in in flat_tree_merge_unique and iterator invalidation in insert_unique".
            GitHub #122: "Fix has_trivial_destructor_after_move".
            GitHub #123: "With heterogeneous lookup, equal_range can result in a range with length greater than 1".
        deque can now have options, using deque_options. The block size/bytes can be be specified.
        static_vector can now have options, using static_vector_options. Alignment and throwing behaviour can be be specified.
        small_vector can now have options, using small_vector_options. Alignment and growth factor can be be specified.
    Context:
        Add support for RISC-V LP64D
        #72: Fix ABI violation on ppc64 ELFv2
        #99: cleanup of stack if creating ucontext fails
    Conversion:
        Maintenance work.
    Core:
        Added functions alloc_construct, alloc_construct_n, alloc_destroy, and alloc_destroy_n in <boost/core/alloc_construct.hpp> for allocator aware and exception safe construction and destruction of objects and arrays. (Glen Fernandes)
        Added constexpr functions first_scalar in <boost/core/first_scalar.hpp> for obtaining a pointer to the first scalar element of an array. Given a pointer of type T* they return a pointer of type remove_all_extents_t<T>*. (Glen Fernandes)
        Added class template noinit_adaptor in <boost/core/noinit_adaptor.hpp> which is an allocator adaptor that converts any allocator into one whose construct(ptr) performs default initialization via placement new, and whose destroy(ptr) invokes the value_type destructor directly. (Glen Fernandes)
        Added class template default_allocator in <boost/core/default_allocator.hpp>, which can serve as a minimal default allocator that has interface similar to C++20 std::allocator, supports configurations with disabled exceptions and does not have std as an associated namespace. The allocator uses operator new and operator delete for allocation. (Glen Fernandes)
        In <boost/core/uncaught_exceptions.hpp> header, added workarounds for better compatibility with QNX SDP 7.0 when libc++/libc++abi libraries are used. (Andrey Semashev, #59)
        The <boost/detail/sp_typeinfo.hpp> header is now marked as deprecated and will be removed in a future release. <boost/core/typeinfo.hpp> should be used instead. (Peter Dimov)
    Dynamic Bitset:
        Enabled hardware-assisted popcount on MSVC (#38).
        Added support for boost::hash and std::hash (#45).
        Support copy-initialization with default constructor (#48).
    Endian:
        Clarified requirements on the value type template parameter
        Added support for float and double
        Added endian_load, endian_store
        Updated endian_reverse to correctly support all non-bool integral types
        Moved deprecated names to the deprecated header endian.hpp
    Fiber:
        documentation for shared_work updated
    Filesystem:
        Fixed incorrect error_code returned from directory iterator increment when readdir_r is used.
        For path, fixed rvalue-aware operator/ return type to return an rvalue instead of rvalue reference. This fixes leaving a dangling reference in the user's code if the result of operator/ is bound to a const reference. (#110)
        Fixes for better compatibility with Windows CE. (#24)
        Added minimal support for CMake. (#106)
    Flyweight:
        Maintenance work.
    Histogram:
        New features:
            Support for thread-safe storages and new thread-safe accumulators
            Support for compiling without exceptions/RTTI (increases performance by 10-20 %) (with Glen Fernandes)
            Performance improvements for 1D and 2D histograms
            boost::histogram::indexed now returns forward iterator range instead of input iterator range
            boost::histogram::indexed_range::accessor is now non-copyable and acts like reference to cell value, making more algorithms from the C++ stdlib work
            boost::histogram::algorithm::reduce with new slice option and option fusion
            boost::histogram::algorithm::project accepts runtime indices for static histograms
        Bug Fixes:
            boost::histogram::algorithm::reduce also works on histograms that have some axis types without reduction support
            boost::histogram::axis::traits::update now works correctly for boost::histogram::axis::variant
        Other:
            100 % test coverage
            Reduced internal Boost dependencies
            Improved documentation and examples
            Guaranteed no-throw moves for all builtin axis types
            Compile cleanly at higher warning levels
        See changelog for more details.
    IoStreams:
        Fixed processing of multi-stream files (#87).
        Added support for multi-threaded LZMA (#95).
    Interprocess:
        GitHub #85 ("warning: Implicit conversion loses integer precision").
        GitHub #86 ("warning: Possible misuse of comma operator").
    Intrusive:
        GitHub #42: Documentation does not describe treap priority_of_value changes
        GitHub #43: Fix tests with BOOST_INTRUSIVE_VARIADIC_TEMPLATES enabled
        GitHub #45: Disable variadic templates for MSVC-12 to avoid ICEs
    LexicalCast:
        Added tests for boost::filesystem::path conversions #25.
        Maintenance work, including #28.
    Log:
        New features:
            Improved support for C++17 std::string_view in basic_formatting_ostream. The string view can now participate in character code conversion on output.
            Added auto_newline formatter and stream manipulator. It can be used to ensure that formatted output always ends with a newline while avoiding duplicate newlines.
            In the output stream, text file and text multi-file sink backends added support for configuring behavior with regard to appending a trailing newline to every formatted log record. Use auto_newline_mode named parameter of the backend constructor or call the set_auto_newline_mode method on the sink backend.
            Note: The default behavior with regard to trailing newlines added by sink backends has changed slightly compared to the previous Boost.Log releases. The backends will now only add a trailing newline if there isn't one in the formatted log message string already. In previous releases a newline was added unconditionally.
            Added support for passing arbitrary function objects in the filter and format named parameters to sink constructors and convenience functions for initializing sinks. For example, it is now possible to specify C++11 lambda functions directly in these parameters. (#63)
            In the default filter and formatter factories used in filter and formatter parsers, added support for severity level attribute values of type boost::log::trivial::severity_level. For filters, the support is limited to attributes with "Severity" name.
        Bug fixes:
            Fixed incorrect parsing of components of the rotated file names while scanning for files in the text file sink backend. If the file name pattern ended with a placeholder (for example, a file counter), the scan_for_files method would not find files matching that pattern in the target storage, leaving them unmanaged. In particular, such files would not be deleted to free target storage. (#78)
            Updated basic_formatting_ostream and basic_record_ostream to make it possible to overload stream output operators for pointers to user-defined types. User-defined operator<< overloads taking std::basic_ostream and a pointer argument should now be picked up by the compiler when the pointer is being written to one of Boost.Log streams. (#84)
        See changelog for more details.
    Math:
        New features:
            Catmull-Rom interpolator now works in C++11
            Cardinal quadratic B-spline interpolation
            Domain of elliptic integrals extended
            sin_pi and cos_pi performance improvements
            Forward-mode automatic differentiation
            Vector valued barycentric rational interpolation
            Ooura's method for evaluation of Fourier integrals
        Bug fixes:
            Multiple compatibility issues with Multiprecision fixed
            Lambert-W fixed on a rare architecture
    Metaparse:
        New features:
            In C++11 variadic template support for the following: sequence, one_of_c, one_of, repeated_one_of, repeated_one_of1, one_char_except_c, one_char_except.
        Bug fixes:
            BOOST_METAPARSE_STRING does not use out of range character values as template arguments.
            any_of_c<> does not create empty array in C++14.
    Move:
        Git Issue #26: "Invalid iterator increment/decrement in the last iteration of adaptive_sort_combine_blocks".
    Multi Array:
        Simplify allocator support by using new alloc_construct_n and alloc_destroy_n facilities from Core. (Glen Fernandes)
    Multi-index Containers:
        Added variants of const_mem_fun and mem_fun for differently qualified member functions (issue #24).
        Terse key specification syntax now correctly handles noexcept-specified member functions (issue #24).
    Outcome:
        Enhancements:
            #184 As per request from Boost release managers, relocated version.hpp and revision.hpp into detail, and added the Boost licence boilerplate to the top of every source file which was missing one (I think). Also took the opportunity to run the licence restamping script over all Outcome, so copyright dates are now up to date.
            #185 Add FAQ item explaining issue #185, and why we will do nothing to fix it right now.
            #189 Refactored the OUTCOME_TRY implementation to use more clarified customisation points capable of accepting very foreign inputs. Removed the std::experimental::expected<T, E> specialisations, as those are no longer necessary. Fixed the documentation for the customisation points which previously claimed that they are ADL discovered, which they are not. Added a recipe describing how to add in support for foreign input types.
            #183 Added a separate motivation/plug_error_code specifically for Boost.
        Bug fixes:
            OUTCOME_VERSION_MINOR hadn't been updated to 1.
            #181 Fix issue #181 where Outcome didn't actually implement the strong swap guarantee, despite being documented as doing so.
            #190 Fix issue #190 in Boost edition where unit test suite was not runnable from the Boost release distro.
            #182 Fix issue #182 where trait::is_exception_ptr_available<T> was always true, thus causing much weirdness, like not printing diagnostics and trying to feed everything to make_exception_ptr().
            #192 Fix issue #192 where the std::basic_outcome_failure_exception_from_error() was being defined twice for translation units which combine standalone and Boost Outcome's.
    Parameter:
        Upgraded keyword generation macro BOOST_PARAMETER_TEMPLATE_KEYWORD (#15).
        Moved keyword generation macro BOOST_PARAMETER_NESTED_KEYWORD from Accumulators to this library (#28).
        Added support for std::reference_wrapper and std::ref() (#16).
        Moved boost::parameter::required, boost::parameter::optional, and boost::parameter::deduced metafunction definitions to their own header files in directory boost/parameter (#18).
        Added support for Boost.Parameter-enabled function call operators (#20).
        Added support for parameter category qualifiers "forward", "consume", and "move_from" (current qualifiers are "in", "out", and "in_out") (#21) (#23) based on http://www.modernescpp.com/index.php/c-core-guidelines-how-to-pass-function-parameters. Added new usage syntax BOOST_PARAMETER_NAME((object-name), namespace-name) qualifier(tag-name)) and BOOST_PARAMETER_NAME(qualifier(name)). (Existing code that uses qualifiers directly and correctly with BOOST_PARAMETER_FUNCTION and other code generation macros should remain unaffected for now, so no breaking changes.) The reason for the change in usage is to enable applying of parameter category constraints to Boost.Parameter-enabled functions and constructors invoked through argument composition. (Otherwise, it is currently possible to use argument composition to bypass parameter category constraints applied in BOOST_PARAMETER_FUNCTION et. al.)
        Added support for perfect forwarding (#23) (#26), so that parameter::parameters::operator() can accept non-const rvalues. As a positive side effect, Boost.Parameter-enabled functions and constructors are no longer bound by BOOST_PARAMETER_MAX_ARITY on compilers that support perfect forwarding. User code can now check for this support by detecting the configuration macro BOOST_PARAMETER_HAS_PERFECT_FORWARDING, or manually turn off this support by defining the configuration macro BOOST_PARAMETER_DISABLE_PERFECT_FORWARDING.
        Added metafunctions boost::parameter::is_argument_pack (#27), boost::parameter::are_tagged_arguments (#52), and boost::parameter::result_of::compose (#75).
        Added variadic function template boost::parameter::compose() which takes in named arguments and returns them in an argument pack (#52). For compilers that do not support perfect forwarding, the configuration macro BOOST_PARAMETER_COMPOSE_MAX_ARITY determines the maximum number of arguments that boost::parameter::compose() can take in (#61).
        Added code generation macros BOOST_PARAMETER_BASIC_FUNCTION_CALL_OPERATOR, BOOST_PARAMETER_BASIC_CONST_FUNCTION_CALL_OPERATOR, BOOST_PARAMETER_NO_SPEC_FUNCTION, BOOST_PARAMETER_NO_SPEC_MEMBER_FUNCTION, BOOST_PARAMETER_NO_SPEC_CONST_MEMBER_FUNCTION, BOOST_PARAMETER_NO_SPEC_FUNCTION_CALL_OPERATOR, BOOST_PARAMETER_NO_SPEC_CONST_FUNCTION_CALL_OPERATOR, BOOST_PARAMETER_NO_SPEC_CONSTRUCTOR, and BOOST_PARAMETER_NO_SPEC_NO_BASE_CONSTRUCTOR (#52).
        Added support for Boost.MP11 (#47) (#66) (#70). User code can now check for this support by detecting the configuration macro BOOST_PARAMETER_CAN_USE_MP11, or manually turn off this support by defining the configuration macro BOOST_PARAMETER_DISABLE_MP11_USAGE.
        Improved support for parameter-dependent return types via SFINAE (#73).
    PtrContainer:
        Fix a ptr_vector regression introduced in 1.66.0 (#24).
    PolyCollection:
        Maintenance work.
    SmartPtr:
        Added aliasing constructors to weak_ptr
        Added weak_ptr<T>::empty()
        Added enable_shared_from, shared_from, and weak_from
    Stacktrace:
        Fixed output of name(), source_location() and source_line() for the default constructed frame with thread sanitizer builds in C++98 mode.
        Fixed output of long strings from name() and source_location() on MSVC #78.
        Maintenance work.
    Test:
        Boost.test v3.11 see the Changes log for more details.
        Breaking changes:
            Boost.Test shows deprecation warnings if some very old headers as deprecated. If you encounter such warnings, please follow the indications: those headers will be removed in a future release.
        New feature:
            Now BOOST_TEST can be used to compare abstract types
        Bug fixes and pull requests:
            GitHub Issues: #209, #218
            GitHub Pull Requests: #219, #224
    Utility:
        Implemented function template ostream_string in <boost/utility/ostream_string.hpp> to optimally write any kind of string content to an output stream. It satisfies the requirements of [ostream.formatted.reqmts]. (Glen Fernandes)
        Optimized the stream output operators of basic_string_view and basic_string_ref to write directly to the rdbuf stream buffer. (Glen Fernandes)
    Uuid:
        Breaking change: MD5 name-based uuid generation was corrected to be identical on all endian systems. Define BOOST_UUID_COMPAT_PRE_1_71_MD5 to keep the result in a format compatible with 1.66 through 1.70. This does not affect the default name-based uuid generation which is based on SHA1. (#109)
    Variant:
        Fixed compilation of boost::apply_visitor with variants passed by non const reference in #68. Many thanks to Ed Catmur for providing the fix.
        Added support for std::hash (#49). Macro BOOST_VARIANT_DO_NOT_SPECIALIZE_STD_HASH could be defined to avoid those specializations.
        Added micro optimizations for binary size and performance by Nikita Kniazev in #63 and #66.
        Maintenance work, including #64 by Nikita Kniazev and #67 by Hans Dembinski.
    YAP:
        Added an example showing how to make self-evaluating YAP expressions (that is, expressions that don't need an explicit call to evaluate() or transform()).

pkgsrc-specific changes: we are now using upstream's formal release
archive, which means a configure script is provided, so we aren't
pulling in a bunch of tool dependencies just to generate such.

Change log:

Brian Bidulock <bidulock@openss7.org> (16):
      reformat TODO
      generate NEWS better
      update for gettext 0.20.1 to fix issue #23
      update release notes
      generate NOTES and TODO
      place filename last in sed command
      update build process
      have better defaults now
      gcc 9.1 too agressive on array-bounds warnings
      ignore release products
      generate release notes on dist
      handle annotated tags better when generating NEWS
      update po files
      ignore lz files
      update release files
      update release files

2019-08-23 Jerry Lundström

    Release 2.3.2

    This release fixes a buffer overflow when using TSIG and algorithms
    with digests larger then SHA256, reported by Mukund Sivaraman. Also
    fix build dependencies for `sqrt()`.

    e54aa58 Digest
    bca5d8d sqrt
    d9eaa5b Package

2019-07-24 Jerry Lundström

    Release 2.3.1

    After a report and additional confirming results the use of `poll()` in
    the network receive code for TCP and TLS has been removed. This `poll()`
    initially gave better results while testing in a docker container on
    it's loopback interface but when on physical networks it reduced
    performance to 1/12th, so it had to go.

    Thanks to Brian Wellington (Akamai/Nominum) for the initial report and
    testing, and to Jan Hák (CZ.NIC) for testing and confirming the results.

    Bugfix:
    - Fix check for having more DNS messages in the receive buffer for TCP
      and TLS

    670db9c TCP/TLS receive
    b8925b2 recvbuf have more

2019-07-17 Jerry Lundström

    Release 2.3.0

    This release adds support for DNS over TCP and TLS which can be selected
    by using the mode option for `dnsperf` and `resperf`. The default server
    port used is now determined by the transport mode, udp/tcp port 53 and
    tls port 853.

    Note that the mode option is different between the program because it was
    already taken for `resperf`.

    `dnsperf` changes:
    - Add `-m` for setting transport mode, `udp` (default), `tcp` or `tls`
    - Add verbose messages about network readiness and congestion

    `resperf` changes:
    - Add `-M` for setting transport mode, `udp` (default), `tcp` or `tls`
    - Add `-v` for verbose mode to report about network readiness and
      congestion

    ffa49cf LGTM, SonarCloud
    4cd5441 TLS
    35624d1 TCP send, socket ready loop
    fbf76aa TCP support
    5988b06 Funding

2019-01-28 Jerry Lundström

    Release 2.2.1

    The commit pulled from a fork that used `inttypes.h`, instead of ISC
    internal types, missed to remove the old conversion specifier.
    This was reported and fixed by Vladimír Čunát.

    9534ce1 remove visible "u" characters after numbers

2019-01-25 Jerry Lundström

    Release 2.2.0

    First release by DNS-OARC with a rework of the code to use autotools,
    semantic versioning 2.0 and bugfixes pulled from other's forks.

    Bugfixes:
    - Fix infinite loop in argument parsing
    - Fix min/max latency summing for multithreaded runs
    - Fix calculation of per_thread socket counts
    - Fixes to queryparse
      - Mark correctly end of file
      - Support python3
      - Stop looping on end of file undefinitely
    - Fix compilation issues and work around missing `dns_fixedname_initname()`
    - Clang `scan-build` fixes

    Other changes:
    - add "configure --with-bind" option
    - Handle bind library changes to HMAC (see #22) and other differences
      between versions
    - Workaround issue on FreeBSD (see #23)
    - Use `snprintf()` and OpenBSD's `strlcat()`
    - Add/update build dependencies for Debia, Ubuntu, CentOS, FreeBSD
      and OpenBSD

    ae9bc91 Clang format
    b9bb085 CI, buildbot
    b84e41b Autotools, README, changelog
    a2e1732 License
    9dcb661 Remove $Id markers, Principal Author and Reviewed tags from the
            full source tree
    0677bf0 Use dns_fixedname_initname() where possible
    d8d4696 [master] add "configure --with-bind" option to dnsperf
    b71a280 Add deb based distros dependencies
    439c614 Replace custom isc_boolean_t with C standard bool type
    407ae7c Replace custom isc_u?intNN_t types with C99 u?intNN_t types
    c27afd4 Replace ISC_PRINT_QUADFORMAT with inttypes.h format constants
    6fdb2f7 Fix queryparse
    4909b78 README
    2782d50 README.md: Rectify link to software
    e31ddf4 fix calculation of per_thread socket counts
    3bd7fb4 Fix min/max latency summing for multithreaded runs
    2207e27 Fix infinite loop in argument parsing.
    3bfe97a Include the github URL; remove the bug reports section.
    0cee04a Add note about bug reports.
    62c4b32 add .gitignore
    c45f0be Initial import.
    149172b Initial commit

pkgsrc-specific changes: we are now using upstream's formal release
archive, which means a configure script is provided, so we aren't
pulling in a bunch of tool dependencies just to generate such.

Change log:

Brian Bidulock <bidulock@openss7.org> (16):
      reformat TODO
      generate NEWS better
      update for gettext 0.20.1 to fix issue #23
      update release notes
      generate NOTES and TODO
      place filename last in sed command
      update build process
      have better defaults now
      gcc 9.1 too agressive on array-bounds warnings
      ignore release products
      generate release notes on dist
      handle annotated tags better when generating NEWS
      update po files
      ignore lz files
      update release files
      update release files

2019-08-23 Jerry Lundström

    Release 2.3.2

    This release fixes a buffer overflow when using TSIG and algorithms
    with digests larger then SHA256, reported by Mukund Sivaraman. Also
    fix build dependencies for `sqrt()`.

    e54aa58 Digest
    bca5d8d sqrt
    d9eaa5b Package

2019-07-24 Jerry Lundström

    Release 2.3.1

    After a report and additional confirming results the use of `poll()` in
    the network receive code for TCP and TLS has been removed. This `poll()`
    initially gave better results while testing in a docker container on
    it's loopback interface but when on physical networks it reduced
    performance to 1/12th, so it had to go.

    Thanks to Brian Wellington (Akamai/Nominum) for the initial report and
    testing, and to Jan Hák (CZ.NIC) for testing and confirming the results.

    Bugfix:
    - Fix check for having more DNS messages in the receive buffer for TCP
      and TLS

    670db9c TCP/TLS receive
    b8925b2 recvbuf have more

2019-07-17 Jerry Lundström

    Release 2.3.0

    This release adds support for DNS over TCP and TLS which can be selected
    by using the mode option for `dnsperf` and `resperf`. The default server
    port used is now determined by the transport mode, udp/tcp port 53 and
    tls port 853.

    Note that the mode option is different between the program because it was
    already taken for `resperf`.

    `dnsperf` changes:
    - Add `-m` for setting transport mode, `udp` (default), `tcp` or `tls`
    - Add verbose messages about network readiness and congestion

    `resperf` changes:
    - Add `-M` for setting transport mode, `udp` (default), `tcp` or `tls`
    - Add `-v` for verbose mode to report about network readiness and
      congestion

    ffa49cf LGTM, SonarCloud
    4cd5441 TLS
    35624d1 TCP send, socket ready loop
    fbf76aa TCP support
    5988b06 Funding

2019-01-28 Jerry Lundström

    Release 2.2.1

    The commit pulled from a fork that used `inttypes.h`, instead of ISC
    internal types, missed to remove the old conversion specifier.
    This was reported and fixed by Vladimír Čunát.

    9534ce1 remove visible "u" characters after numbers

2019-01-25 Jerry Lundström

    Release 2.2.0

    First release by DNS-OARC with a rework of the code to use autotools,
    semantic versioning 2.0 and bugfixes pulled from other's forks.

    Bugfixes:
    - Fix infinite loop in argument parsing
    - Fix min/max latency summing for multithreaded runs
    - Fix calculation of per_thread socket counts
    - Fixes to queryparse
      - Mark correctly end of file
      - Support python3
      - Stop looping on end of file undefinitely
    - Fix compilation issues and work around missing `dns_fixedname_initname()`
    - Clang `scan-build` fixes

    Other changes:
    - add "configure --with-bind" option
    - Handle bind library changes to HMAC (see #22) and other differences
      between versions
    - Workaround issue on FreeBSD (see #23)
    - Use `snprintf()` and OpenBSD's `strlcat()`
    - Add/update build dependencies for Debia, Ubuntu, CentOS, FreeBSD
      and OpenBSD

    ae9bc91 Clang format
    b9bb085 CI, buildbot
    b84e41b Autotools, README, changelog
    a2e1732 License
    9dcb661 Remove $Id markers, Principal Author and Reviewed tags from the
            full source tree
    0677bf0 Use dns_fixedname_initname() where possible
    d8d4696 [master] add "configure --with-bind" option to dnsperf
    b71a280 Add deb based distros dependencies
    439c614 Replace custom isc_boolean_t with C standard bool type
    407ae7c Replace custom isc_u?intNN_t types with C99 u?intNN_t types
    c27afd4 Replace ISC_PRINT_QUADFORMAT with inttypes.h format constants
    6fdb2f7 Fix queryparse
    4909b78 README
    2782d50 README.md: Rectify link to software
    e31ddf4 fix calculation of per_thread socket counts
    3bd7fb4 Fix min/max latency summing for multithreaded runs
    2207e27 Fix infinite loop in argument parsing.
    3bfe97a Include the github URL; remove the bug reports section.
    0cee04a Add note about bug reports.
    62c4b32 add .gitignore
    c45f0be Initial import.
    149172b Initial commit

Changes since 0.14.0:

Features

  * This release adds a new --diff/-d option that can be used to only show lines
    surrounding Git changes, i.e. added, removed or modified lines. The amount of
    additional context can be controlled with --diff-context=N. See #23 and #940

Bugfixes

  * Error message printed in the middle of the output for another file, see #946
  * Performance improvements when using custom caches (via bat cache --build): the
    bat startup time should now be twice as fast (@lzutao).

Themes

  * Updated version of the Solarized dark/light themes, see #941

bat as a library

  * There are a few changes in the "low level" API (the Config struct has changed and
    the error handler needs a new &mut dyn Write argument). The high-level API is not
    affected.

Patchlevel 7b (Oct 2019)

NEW FEATURES:
	o A X color database is not needed, but can be provided. The location of
	  the database can be given at compile time, default /etc/X11/rgb.txt.

BUGS FIXED:
	Ticket numbers refer to https://sourceforge.net/p/mcj/tickets/#.
	Debian bug numbers refer to https://bugs.debian.org/#.
	o Do not clip objects with line-thickness 0 having arrows. Ticket #53.
	o Do not segfault on circle/half circle arrowheads with a magnification
	  larger 42. Always draw circle arrowheads with 40 points. Ticket #52.
	o Allow circles or ellipses with negative radii. Ticket #49.
	o Avoid "dimension too large error" with tikz output by avoiding
	  coordinate values smaller than -16383.
	o Make tests (test1.c) work with -fsanitize=address compiler option.
	o Obey join-style of lines in tikz output.
	o Pass utf8-strings to svg output, escape some chars (<>&).
	o Accept inclined boxes and change them to polygons. Fixes ticket #43.
	o Make tests #27 and #33 work on Mac Darwin, failed due to whitespace
	  formatting differences. From Hanspeter Niederstrasser. Ticket #40.
	o Use only latex, neither etex or tex, to test tikz output. Usage of
	  etex, after hint from Roland Rosenfeld, closed debian bug 920368.
	o For tikz output, do not draw arrows on a single point line.
	o Omit spurious showpage when including jpg-file. From Rainer Buchty.
	o Correct a few memory leaks and corruptions. See commit d1c54f6.
	o Change negative color numbers to default color. Fixes ticket #30.
	o A spline with one point would cause segfault. Fixed, see ticket #29.
	o Allow one char without newline in the last line of an input file.
	  Fixes ticket #28.
	o Harden input, mainly against files in which an incomplete object would
	  be created and freeing the object would violate memory, i.e, it may
	  cause segfault. See, e.g., ticket #27.
	o Properly initalize line storage when reading fig files version 1.3.
	  Would segfault when reading incomplete line and trying to free it.
	  Fixes ticket #26, debian bug 906743.
	o Silently ignore the hundred-first and more comment lines. This
	  fixes ticket #25 and debian bug 906740.
	o Use SetFigFont, not SetFigFontNFSS in pictex output. Fixes
	  https://bugs.launchpad.net/ubuntu/+source/transfig/+bug/1359485 .
	o Accept blanks in color names (e.g., fig2dev -L eps -g"Misty Rose"..).
	o Correct typos in man-pages, debian 30_man_typo.patch.

-------------------------------------
Patchlevel 7a (May 2018)

NEW FEATURES:
	o Add option -w, wrap (create stand-alone perl file) for Perl/Tk output.
	o Distribute the X bitmaps files within fig2dev, no need to install
	  these files. The files were needed for Tk and Perl/Tk output.

BUGS FIXED:
	Ticket numbers refer to https://sourceforge.net/p/mcj/tickets/#.
	o Fix regression whereupon flipped ellipses were not read. Ticket #23.
	o Distribute i18n files ru_RU.CP1251.ps and uk_UA.KOI8-U.ps.
	o Make test "survive debian bug #890016" succeed on 32 bit systems.

-------------------------------------
Patchlevel 7 (April 2018)

OPTION LETTER CHANGES:
	o  Language         previous option         current option
	  ------------------------------------------------------------
	   cgm              -b dummy                -a
	   epic             -A scale                -d scale
	   eepic            -A scale                -d scale
	   eepicemu         -A scale                -d scale
	   gbx              -i on|off               -v
	   ibmgl            -m mag,xoff,yoff        -m mag -x xoff -y yoff
	   mp               -I file                 -d file
	   ps               -S dummy                -o

NEW FEATURES:
	o Print language-specific help text by using fig2dev -L lang -h.
	o Add option -M, multipage, for MetaPost output language.
	o Add option -P, pagemode, and -z to choose a pagesize for pdf output.
	o Add option -W (scaling of figures not possible) for tikz.
	o Add option -b, border width, for LaTeX output language.
	o Add option -f for pstex_t and pdftex_t output language.
	o Add uk_UA and ru_RU encodings for PostScript output. Ticket #12.

BUGS FIXED:
	o Update help text: Output help for dxf and textyl output language,
	  add description of -g option for Tk/Tcl and Perl/Tk output, allow -f
	  option for pstex_t and pdftex_t output language.

	Debian bug numbers refer to https://bugs.debian.org/#.
	Ticket numbers refer to https://sourceforge.net/p/mcj/tickets/#.
	o Sanitize input. Do not segfault on malformed input files. Fixes debian
	  bugs 881143, 881144, 881396, 890015, 890016, 882021 and also 882022.
	o Do not put an %%Orientation: comment into PostScript output. Some
	  viewers would rotate the resulting file, others not.
	o Fix build on NetBSD, which has a _setmode() function different from
	  _setmode() on Windows. Ticket #17. Also, avoid alloca(). Ticket #16.
	o tikz output: Omit the semicolon after \pgftext[..]{...};.
	o Define PostScript patterns with larger tiles, may render better. #13
	o Fix build in case libXpm is missing. Ticket #15.
	o Use netpbm programs instead of ghostscript, to produce smaller files.
	o Correctly embed eps files with binary preview (epsi, typically
	  found on Microsoft systems). Also, allow to embed ps-files. Fixes
	  debian bug 248807, ticket #8.
	o For compilation, do not depend on PATH_MAX being defined.

3.2.5
Added
  IMAP Daemon: added switch to control the diffential state reload
    (mailbox_update_strategy=2), more information in dbmail.conf,
    mailbox_update_strategy_2_max_iterations [#81]
  IMAP Daemon: added switch to control UNSEEN first message in SELECT commands

Changed
  IMAP Daemon: allow reporting UID COPY success in case of various failures
    (except quota), reporting issues are sent to error log as warnings [#87]

Optimizations
  optimizing differential state [#81]
  optimizing fetch message headers [#85]

Issues
  fixing issue related to copy message in regard to RFC 3501, section 6.4.8
  fixing issues related group_concat for PostgreSql [#75], [#78]
  fixing issue related to lastRowId [#71]
  fixing issues related with differential update [#70], [#73]
  fixing proc not being used in BSD systems [#74]
  IMAP Daemon: segmentation fault [#68]

3.2.4
Added
  IMAP Daemon: mailbox-update-strategy switch (see dbmail.conf), experimental
  support for application_name in database connection uri
  IMAP Daemon: mailbox_search_strategy switch (see dbmail.conf)

Changed
  systemd unit changed to type notify
  mailbox state is build using only valid messages [#39]

Optimizations
  IMAP Daemon: optimization of sql queries in relation to message headers
  libevent increased priority on accepting new connections
  libevent optimization on reading and writing to sockets
  simplify libzdb configuration (AC_CHECK_HEADERS)

Issues
  fix segmentation fault in imap_append_hash_as_string [#12]
  dbmail-users: sql issue on deleting alias user [#18]
  IMAP Daemon: generation of invalid BODYSTRUCTURE in Content-Type field [#23]
  fix support for jemalloc latest version [#35]
  IMAP Deamon: BYE Command now offers optional message even on normal operations
  IMAP Deamon: idle message now offers optional message (* OK Still Here)
  IMAP Daemon: random hangs when single user is connected [#37]
  fix fd leaks
  IMAP Daemon: fix MODIFIED keyword, too many '[' and ']'
  fix segmentation fault in find_end_of_header
  fix gcc 10 compilation issue, duplicated definition

Change to the OpenPrinting fork of apple/cups, e.g., that fork already contains

- dnssd patch patch-config-scripts_cups-dnssd.m4
- freebsd patch patch-cups_getifaddrs-internal.h
- libtool patch patch-af

apple/cups#4947 was fixed in both.

Changes in CUPS v2.3.3op1
-------------------------

- The automated test suite can now be activated using `make test` for
  consistency with other projects and CI environments - the old `make check`
  continues to work as well, and the previous test server behavior can be
  accessed by running `make testserver`.
- ippeveprinter now supports multiple icons and strings files.
- ippeveprinter now uses the system's FQDN with Avahi.
- ippeveprinter now supports Get-Printer-Attributes on "/".
- ippeveprinter now uses a deterministic "printer-uuid" value.
- ippeveprinter now uses system sounds on macOS for Identify-Printer.
- Updated ippfind to look for files in "~/Desktop" on Windows.
- Updated ippfind to honor `SKIP-XXX` directives with `PAUSE`.
- Updated IPP Everywhere support to work around printers that only advertise
  color raster support but really also support grayscale (Issue #1)
- ipptool now supports DNS-SD URIs like `ipps://My%20Printer._ipps._tcp.local`
  (Issue #5)
- The scheduler now allows root backends to have world read permissions but not
  world execute permissions (Issue #21)
- Failures to bind IPv6 listener sockets no longer cause errors if IPv6 is
  disabled on the host (Issue #25)
- The SNMP backend now supports the HP and Ricoh vendor MIBs (Issue #28)
- The scheduler no longer includes a timestamp in files it writes (Issue #29)
- The systemd service names are now "cups.service" and "cups-lpd.service"
  (Issue #30, Issue #31)
- The scheduler no longer adds the local hostname to the ServerAlias list
  (Issue #32)
- Added `LogFileGroup` directive in "cups-files.conf" to control the group
  owner of log files (Issue #34)
- Added `--with-max-log-size` configure option (Issue #35)
- Added `--enable-sync-on-close` configure option (Issue #37)
- Added `--with-error-policy` configure option (Issue #38)
- IPP Everywhere PPDs could have an "unknown" default InputSlot (Issue #44)
- The `httpAddrListen` function now uses a listen backlog of 128.
- Added USB quirks (Apple issue #5789, #5823, #5831)
- Fixed IPP Everywhere v1.1 conformance issues in ippeveprinter.
- Fixed DNS-SD name collision support in ippeveprinter.
- Fixed compiler and code analyzer warnings.
- Fixed TLS support on Windows.
- Fixed ippfind sub-type searches with Avahi.
- Fixed the default hostname used by ippeveprinter on macOS.
- Fixed resolution of local IPP-USB printers with Avahi.
- Fixed coverity issues (Issue #2)
- Fixed `httpAddrConnect` issues (Issue #3)
- Fixed web interface device URI issue (Issue #4)
- Fixed lp/lpr "printer/class not found" error reporting (Issue #6)
- Fixed xinetd support for LPD clients (Issue #7)
- Fixed libtool build issue (Issue #11)
- Fixed a memory leak in the scheduler (Issue #12)
- Fixed a potential integer overflow in the PPD hashing code (Issue #13)
- Fixed output-bin and print-quality handling issues (Issue #18)
- Fixed PPD options getting mapped to odd IPP values like "tray---4" (Issue #23)
- Fixed remote access to the cupsd.conf and log files (Issue #24)
- Fixed the automated test suite when running in certain build/CI environments
  (Issue #25)
- Fixed a logging regression caused by a previous change for Apple issue #5604
  (Issue #25)
- Fixed fax phone number handling with GNOME (Issue #40)
- Fixed potential rounding error in rastertopwg filter (Issue #41)
- Fixed the "uri-security-supported" value from the scheduler (Issue #42)
- Fixed IPP backend crash bug with "printer-alert" values (Issue #43)
- Removed old Solaris inetconv(1m) reference in cups-lpd man page (Issue #46)
- Fixed default options that incorrectly use the "custom" prefix (Issue #48)
- Fixed a memory leak when resolving DNS-SD URIs (Issue #49)
- Fixed systemd status reporting by adopting the notify interface (Issue #51)
- Fixed crash in rastertopwg (Apple issue #5773)
- Fixed cupsManualCopies values in IPP Everywhere PPDs (Apple issue #5807)

0.2.11 (2020-09-30)

Merged Pull Requests

* Remove redundant encoding comments #22 (tas50)
* Add winrm-shell-type option and winrm elevated shell #25 (catriona1)
* Add validation to winrm shell type option #28 (catriona1)
* Allow timeout option for WinRM commands #27 (james-stocks)
* Correct minor spelling mistakes #23 (tas50)

No release note for from 0.2.7 to 0.2.10 but these seems to various fixes.

3.0.2 (2022-11-11)

What's Changed

* Simplify charset parsing by @semaperepelitsa in #28

3.0.1 (2022-10-08)

What's Changed

* Fix ReDos/performance when receiving crafted response headers by
  @ooooooo-q in #27

3.0.0 (2022-08-08)

What's Changed

* Add support for the new Faraday streaming API by @iMacTia in #26

Why a 3.0 release? What's the breaking change?

Faraday 2.5 introduces a new streaming API and other changes that we want to
take advantage on.  However, releasing a new 2.x version of the gem would
cause it to be pulled in projects where Faraday has not been updated to
v2.5+ yet, causing errors.

For this reason, we're instead releasing this as v3.0, which only the next
(compatible) version of faraday will allow to use.

2.1.0 (2022-07-28)

What's Changed

* Handle verify_hostname ssl option by @kazarin in #23

2.0.3 (2022-05-17)

What's Changed

* Add Errno::EALREADY to list of Net::HTTP exceptions by @iMacTia in #21

2.0.2 (2022-04-10)

What's Changed

* Add Ruby 3.1 to CI by @petergoldstein in #15
* Anchor Encoding references to avoid faraday-encoding conflicts by @nbibler
  in #18

2.0.1 (2022-01-05)

Fixes

* Add back support for Faraday 1.0

2.0.0 (2022-01-04)

What's Changed

* Test on Ruby 3 by @tricknotes in #3
* Update gem to be compatible with Faraday 2.0 by @iMacTia in #9
* chore: Move development deps to Gemfile by @olleolleolle in #10
* refactor: CI: Inline scripts, cache gems by @olleolleolle in #11
* fix: gemspec metadata for changelog notes by @olleolleolle in #12
* Honor Content-Type charset by @xkwd in #13

NEWS/Changelog

1.6-0	(2022-07-05)
    o	Added text drawing using Harfbuzz and ICU. This significantly
	improves the quality of text rendeded by Cairo and enables
	support for right-to-left and bi-directional text. (#23)
	It also enables the use of ligatures. (#18)
	This feature is only enabled if `pkg-config` is present and
	knows how to use the `harfbuzz-icu` module.

    o	The `family=` argument in text drawing is now honored also for
	FreeType and Harfbuzz back-ends.

    o	configure gains two arguments: `--with-freetype` and
	`--with-harfbuzz`. They both default to `auto` which means
	FreeType and Harfbuzz will be used if present. If set to `yes`
	then the support is required. If set to `no` (or the
	`--without` form is used) then the corresponding support is
	explicitly disabled. Note that Harfbuzz requires FreeType.

    o	added `--enable-debug` argument to configure which will make
	Cairo extremely chatty on the console.

    o	By default `cairo-ft` will be ignored on Windows, becasue it
	it does not work without associated FontConfig configuration
	files and font cache. (#37)
	This behavior can be changed by using `-DUSE_CAIRO_FT=1` to
	force the use of `cairo-ft` unconditionally.

    o	Cairo.capabilities() now also reports two capabilities
	which are not format-related: "freetype" and "harfbuzz". The
	former indicates whether `cairo-ft` is used and the later
	indicates whether the newly introduced Harfbuzz + ICU
	bi-directional rendering support is present.


1.5-15	(2022-03-16)
    o	CairoX11 (xlib back-end): install an X11 I/O error handler
	while processing events such that broken connections are
	detected and don't kill R.

    o	added the option to license under GPL v3. All original
	authors explicitly agreed and included code based on R is
	licensed under GPL-2+ allowing this.


1.5-14	(2021-12-21)
    o	update configure/autoconf, some dependencies such as libtiff
	are now also detected using pkg-config. Tests are now correctly
	using CPPFLAGS from R.

    o	add entry points for GE version 13 (needed for R 4.1.0)
	Note that this is just a port of CRAN NMU which was posted as
	version 1.15-12.2 on CRAN, but was not an actual Cairo version.


1.5-13	(2020-04-15)
    o	add support for CairoFont(usePUA=FALSE) to support regular
	fonts without Adobe Symbol Encoding as symbol fonts in
	R 4.0.0 and higher. Setting usePUA=FALSE in older versions of
	R will issue a warning.

Release 2.3.0

Added

    Add no_color and force_color parameters to override env vars (#38) @hugovk
    Add support for Python 3.12 (#37) @hugovk

Changed

    Publish to PyPI with a Trusted Publisher (#45) @hugovk

Release 2.2.0

Added

    Add light shades, dark grey and black (#32) @hugovk

Release 2.1.1

Fixed

    Add __main__ to re-enable demo via python -m termcolor (#27) @hugovk

Release 2.1.0

Added

    Support FORCE_COLOR and detect tty (#25) @hugovk

Deprecated

    Deprecate __ALL__, use __all__ instead (#23) @hugovk

Release 2.0.1

Fixed

    Update source URL (#21) @felixonmars

Release 2.0.0

Added

    Add support for Python 3.11 and PyPy (#9) @hugovk
    Add support for Python 3.10 (#5) @hugovk
    Support NO_COLOR (#7) @hugovk
    Add type annotations to the project and run mypy on CI (#11) @jdufresne
    Add tests (#1) @hugovk

Changed

    Refer to GitHub Releases for release notes for 2.0.0+ (#20) @hugovk
    Autodeploy to TestPyPI and to PyPI for GH releases (#19) @hugovk
    Migrate from setuptools + setuptools_scm to hatchling + hatch-vcs (#17) @hugovk
    Replace deprecated license_file with license_files in setup.cfg (#14) @jdufresne
    Use declarative metadata in setup.cfg (#10) @hugovk
    Replace 3.9-dev with 3.9 in CI to use Python 3.9 final (#3) @CozyDoomer

Removed

    Drop support for EOL Python <= 3.6 (#9) @hugovk
    Remove hardcoded VERSION constant (#18) @hugovk

Fixed

    Fix docstring for colored function (#15) @tmetzl
    Fix typo: ANSII -> ANSI (#8) @hugovk

erdtree v2.0.0 introduces numerous breaking changes as well as a plethora of
new features. Most breaking changes are predicated on the fact that arguments
were either renamed, removed, or fundamentally modified. The following is a
list of all the PRs that document these changes and feature additions:

  - Regular expressions #130
  - Unix permissions #132
  - Force color #135
  - Show hardlinks without double counting file size #136
  - Regular expressions and globbing with file-type #137
  - Truncate output #138
  - Inverted tree layout #139

Perhaps the most important change to note is that the compiled binary has been
renamed from et to erd in order to address the following issue regarding name
collisions with other programs: #23

FeedGenerator 2.1.0

Modernize and improve tests (#32 & #34 — thanks to @venthur)
Drop support for Python 3.6 and test on 3.10 & 3.11 (#35 — thanks to @hugovk)
Exclude tests_feedgenerator/__pycache__ from distribution (#33 — thanks to @BenSturmfels)

FeedGenerator 2.0.0

Add preliminary support for adding images to feeds
Update code for Python 3.6+
Drop support for Python 2.7
Fix double subtitles if both description & subtitle are provided

FeedGenerator 1.9.2

Use description field as subtitle for Atom feeds, if provided (#23)
Update list of explicitly-supported Python versions to 3.6 – 3.9

0.2.0.2 - 2023-06-12, Alexey Radkov and Andreas Abel
* Support Haiku OS by including libnetwork in configure script. (PR #23.)

0.2.0.1 - 2023-03-31, Alexey Radkov and Andreas Abel
* Fix 0.2.0.0: Ship updated configure script.

0.2.0.0 - 2023-03-31, Alexey Radkov and Andreas Abel
* Bump bytestring to >= 0.10 for correct IsString ByteString instance. (PR
  #16.)
* Fix memory leaks due to missing res_nclose() after each res_ninit()
  call. (PR #12.)
* Check the value of h_errno on failures of res_nquery() and throw an
  appropriate exception of type DnsException built with one of new
  constructors DnsHostNotFound, DnsNoData, DnsNoRecovery, or
  DnsTryAgain. Note that previously such exceptions were thrown by fail and
  had type IOError. (PR #17.)
* Suppress configure warning on option --with-compiler passed by Cabal. (PR
  #21.)
* Tested with GHC 8.0 - 9.6.

1.5.6 (2023-11-18)

* Add support for "<" and ">" operator in "Requires:" (#23)

v1.3.1

Changes

    Drop 3.7 support by @bagel897 in #28
    do sys path hack by @bagel897 in #31

Misc

    Update deps by @bagel897 in #29
    Update docs by @bagel897 in #30

v.1.3.0

Features

    Enum Support by @bagel897 in #19

Misc

    Use trusted publishing by @bagel897 in #17
    deduplicate ci by @bagel897 in #20
    chore: Configure Renovate by @renovate in #21
    chore(deps): update pre-commit hook hadialqattan/pycln to v2.4.0 by @renovate in #23
    chore(deps): update pre-commit hook pre-commit/mirrors-prettier to v3.1.0 by @renovate in #25
    chore(deps): update pre-commit hook psf/black to v23.12.1 by @renovate in #26
    chore(deps): update pre-commit hook charliermarsh/ruff-pre-commit to v0.1.11 by @renovate in #22
    chore(deps): update actions/checkout action to v4 by @renovate in #27

2.05   2024-01-08

- Added explicit boolean overloading for trace objects. Without this, Perl
  will use the object's string overloading and then check the truthiness of
  the returned string, which is a lot of extra work. This can produce
  significant slowdowns in some cases, as seen in
  plack/Plack#697. Requested by Tatsuhiko
  Miyagawa. GH #23.

new minor version: 2.10.0
 - update dependencies
 - Methods for configuring color generation: 8ee3a77386
 - new cli flag: --generation/-g
 - new config optional variable: generation
 - new palettes (filters): a6741f19d6
    - darkcomp, darkcomp16
    - harddarkcomp, harddarkcomp16
    - softdarkcomp, softdarkcomp16
    - softlightcomp, softlightcomp16
    - lightcomp, lightcomp16
    - fixed contrast in softdark 4bfc945091
 - backends:
    - new backend: kmeans algo 80cd0a0a91
    - guess format, instead of using ext #32
 - template variables:
    - wallpaper will display a path (normal usage), a theme (when using theme
      subcommand) or the colorscheme file (when using cs subcommand) ecce329233
    - alpha_hex #23
    - wallpaper will always display the absolute path #33
 - colorspaces:
    - improved color picking 7a701e368f
    - remove some inconsistencies 1ea2b886ee
 - improved sample config generation 004127308e
 - Introduce new changes from v3.md
    - rename filter -> palette, in a backwards compatible way 7fa281f2d6
    - flag --filter/-f renamed to --palette/-p, in a backwards compatible way
      (aliasing)
    - wallust image.png -> wallust run image.png, the first one is still
      supported (backwards compatible)
    - added new config syntax, backwards compatible c3a2e00d9e

First version of the year, thanks everyone!

0.6.20240107 (2024-01-09)

* Update the eTLD database to 2024-01-07 11:25:45 UTC

Closed issues:

* git tag v0.6.20231109 is missing #25
* Unicode Normalization not appropriate for ASCII-8BIT #23

Merged pull requests:

* Fix regression: Allow DomainName.normalize to accept ASCII-only,
  non-unicode encoded input #27 (tisba)

v1.3.0
Released 2023-02-19
Features:

garden eval and garden expressions in general will now resolve variables defined
within environment blocks. Environment blocks were previously not considered when
resolving variables. Environment blocks are now resolved and checked for variables
when ${variable} expressions do not find the variable in scopes with higher
precedence. The precedence order, from strongest to weakest, is the variables
block in a garden's scope, the variables block in a tree's scope, the
variables block in global configuration scope, the environments block in
a garden's scope, the environments block in a tree's scope, the
environments block in global configuration scope and, lastly, OS environment
variables. The first entry found is used when expanding variable expressions.
(#23)

Evaluation cycles (i.e. circular variable dependencies) are now prevented when
evaluating garden variables. The evaluation engine will now return empty strings
when a variable with a cyclical expression is evaluated.
(#24)

When zsh is used as the garden.shell, which happens automatically when zsh
is installed, garden will now use zsh -o shwordsplit in order to enable
word-splitting of $variable expressions by default. This makes zsh behave
just like other shells by default, which improves the portability of commands.
Configure garden.shell-wordsplit to false or use the
garden <cmd> -z | --no-wordsplit option to opt-out of this behavior.
(#25)

garden.shell can now be configured to use arbitrary commands for executing
command strings. Garden uses the configured garden.shell as-is and does
not augment its options (e.g. -e  or -o shwordsplit) when a custom command
is used. Custom commands are identified as commands that expand to 2 or more
command-line arguments. Thus, python3 is not considered a custom command
and garden will run python3 -c <string> to run commands. On the other
hand, specifying ruby -e is considered a custom command because it
expands to ["ruby", "-e"] under the hood. If you need to use a custom
command  that takes no additional command-line arguments then you can
use env as an extra argument to have it be considered as a custom shell.
For example, env custom-shell will cause garden to run
env custom-shell <string>, which is equivalent to custom-shell <string>.
Using just custom-shell would have resulted in garden running
custom-shell -c <string> instead, which may not be desired.
(#26)

The garden shell command can now be configured to use an interactive command shell
that is distinct from the command specified in the garden.shell configuration by
configuring the garden.interactive-shell value.
(#26)

garden shell can now be run without any arguments. The tree query now defaults to
. so that the tree in the current directory is used when nothing is specified.
(#26)

Custom commands now have access to a ${GARDEN_CMD_VERBOSE} and ${GARDEN_CMD_QUIET}
variables which can be used to forward the --verbose and --quiet arguments
down into child garden invocations. ${GARDEN_CMD_VERBOSE} uses the short -v
flag in the value to support the case where the verbose option is specified
multiples times to increase the verbosity level (e.g. -vv).
(#27)

1.12.0 (2024-01-16)

* feature: add sqlite3 driver #18
* feature: add lastId() method on mysql, mysqli and sqlite3 #19 #21
* task: fix prototypes for PHP 8 #12
* task: Try and fix test by removing each #16
* task: PHP8.1 Compatability restore error handling to previous state whilst
  ... #13
* task: DB::apiVersion() should be declared as static #17
* task: Fix PHP8.2 Deprecated use of ${var} instead of {$var} #14
* task: PHP8.2 ready #21
* task: Mark mssql-Driver, mysql-Driver and sybase-Driver as Deprecated #21
* bug: PHP 8.0 testcase failure #23 #24

1.12.1 (2024-01-17)

* bug: Missing new file in package.xml #25
* bug: Minimal fix for sqlite3 #26

2.0.0 (2024-01-15)

* BugFix: Use CRLF for header separation, except sendmail, fixes #23 #24

1.16.0 (2024-03-01)

What's Changed

* Typo in meta_store_test.rb test description [ci skip] by @olleolleolle in
  #23

* Purge cache_key if response headers do not include content digest by
  @evman182 in #22

* Remove memcached gem. by @ioquatix in #24

New Contributors

* @evman182 made their first contribution in #22

### 1.8.10 - 15 June 2024

 * feature: new "`--output`" option to write to a file instead of standard output (pull request [#90](https://codeberg.org/a-j-wood/pv/pulls/90)) supplied by [xmort](https://codeberg.org/xmort)

### 1.8.9 - 21 April 2024

 * feature: new "`--si`" option to display and interpret size suffixes in multiples of 1000 rather than 1024 (pull request [#85](https://codeberg.org/a-j-wood/pv/pulls/85)) supplied by [kevinruddy](https://codeberg.org/kevinruddy)
 * fix: continue producing progress output when the output is blocking writes ([#34](https://codeberg.org/a-j-wood/pv/issues/34), [#86](https://codeberg.org/a-j-wood/pv/issues/86), [#87](https://codeberg.org/a-j-wood/pv/issues/87))
 * fix: honour the _TMPDIR_ / _TMP_ environment variables again, rather than hard-coding "`/tmp`", when using a terminal lock file (originally removed in 1.8.0) ([#88](https://codeberg.org/a-j-wood/pv/issues/88))
 * i18n: corrections and missing strings added to French translations (pull request [#83](https://codeberg.org/a-j-wood/pv/pulls/83)) supplied by [Thomas Bertels](https://codeberg.org/tbertels)

### 1.8.5 - 19 November 2023

 * fix: corrected percentage formatting so it doesn't jump from 2 to 3 characters wide at 100% ([#80](https://codeberg.org/a-j-wood/pv/issues/80))
 * fix: replaced "`--remote`" mechanism, using a temporary file instead of SysV IPC, so it can work reliably even when there are multiple PV instances
 * fix: corrected compilation failure when without IPC support
 * security: addressed all issues highlighted by the software auditing tools "`splint`" and "`flawfinder`" (see "`make analyse`") ([#77](https://codeberg.org/a-j-wood/pv/issues/77))
 * cleanup: compilation warnings fixed on non-IPC and MacOS systems

### 1.8.0 - 24 September 2023

#### Features

 * feature: new "`--discard`" option to discard input as if writing to */dev/null* ([#42](https://codeberg.org/a-j-wood/pv/issues/42))
 * feature: new "`--error-skip-block`" option to make "`--skip-errors`" skip whole blocks ([#37](https://codeberg.org/a-j-wood/pv/issues/37))
 * feature: use `posix_fadvise()` like `cat`(1) does, to improve efficiency ([#39](https://codeberg.org/a-j-wood/pv/issues/39))
 * feature: new "`--enable-static`" option to "`configure`" for static builds ([#75](https://codeberg.org/a-j-wood/pv/pull/75))

#### Security

 * security: with "`--pidfile`", write to a temporary file and rename it into place, to improve security
 * security: keep self-contained copies of name and format string in PV internal state for memory safety
 * security: ignore _TMP_ / _TMPDIR_ environment variables when using a terminal lock file

#### Fixes

 * fix: only report errors about missing files when starting to transfer from them, not while calculating size, and behave more like `cat`(1) by skipping them and moving on
 * fix: auto-calculate total line count with "`--line-mode`" when all inputs are regular files
 * fix: use `clock_gettime()` in ETA calculation to cope with machine suspend/resume ([#13](https://codeberg.org/a-j-wood/pv/issues/13))
 * fix: if "`--width`" or "`--height`" were provided, do not change them when the window size changes ([#36](https://codeberg.org/a-j-wood/pv/issues/36))
 * fix: when a file descriptor position in "`--watchfd`" moves backwards, show the rate using the correct prefix ([#41](https://codeberg.org/a-j-wood/pv/issues/41))
 * fix: rewrite terminal state save/restore so state is not intermittently garbled on exit when using "`--cursor`" ([#20](https://codeberg.org/a-j-wood/pv/issues/20)), ([#24](https://codeberg.org/a-j-wood/pv/issues/24))

#### Cleanups

 * cleanup: addressed many potential issues highlighted by the software auditing tools "`splint`" and "`flawfinder`" (see new target "`make analyse`")
 * cleanup: switched the build system to GNU Automake
 * cleanup: replaced the test harness with the one native to GNU Automake
 * cleanup: added a test for terminal width detection to "`make check`"
 * cleanup: added a test to "`make check`" to ensure that "`make install`" installs everything expected
 * cleanup: replaced *AC_HEADER_TIOCGWINSZ* with *AC_CHECK_HEADERS(sys/ioctl.h)* for better MacOS compatibility ([#74](https://codeberg.org/a-j-wood/pv/issues/74))
 * cleanup: with "`--sync`", call `fsync()` instead of `fdatasync()` on incapable systems ([#73](https://codeberg.org/a-j-wood/pv/issues/73))
 * cleanup: the manual is now a static file instead of needing to be built with "`configure`"

#### Dropped items

 * dropped: dropped support for "`--enable-static-nls`"
 * dropped: removed the Linux Software Map file, as the LSM project appears to be long dead
 * dropped: will no longer publish to SourceForge as it has a chequered history and is unnecessary
 * dropped: removed project from GitHub and moved to Codeberg - see "[Give Up GitHub](https://giveupgithub.org/)"

#### Other items

 * licensing change from Artistic 2.0 to GPLv3+

### 1.7.24 - 30 July 2023

 * fix: correct terminal size detection, broken in 1.7.17 by the configuration script rewrite ([#72](https://codeberg.org/a-j-wood/pv/issues/72))
 * security: removed *DEBUG* environment variable in debug mode, added "`--debug`" instead
 * cleanup: added "`make analyse`" to run "`splint`" and "`flawfinder`" on all source files
 * cleanup: corrected detection of boolean capability
 * cleanup: word wrapping of "`--help`" output is now multi-byte locale aware
 * cleanup: adjusted "`indent`" rules to line length of 120 and reformatted code

### 1.7.18 - 28 July 2023

 * fix: language file installation had been broken by the configuration script rewrite

### 1.7.17 - 27 July 2023

 * feature: new "`--sync`" option to flush cache to disk after every write (related to [#6](https://codeberg.org/a-j-wood/pv/issues/6), to improve accuracy when writing to slow disks)
 * feature: new "`--direct-io`" option to bypass cache - implements [#29 "Option to enable *O_DIRECT*"](https://codeberg.org/a-j-wood/pv/issues/29) - requested by Romain Kang, Jacek Wielemborek
 * fix: correct byte prefix size to 2 spaces in rate display, so progress display size remains constant at low transfer rates
 * cleanup: rewrote `configure.in` as per suggestions in newer "`autoconf`" manuals
 * cleanup: replaced `header.in` with one generated by "`autoheader`", moving custom logic to a separate header file "`config-aux.h`"
 * cleanup: added copyright notice to all source files as per GNU standards
 * cleanup: changed "`--version`" output to conform to GNU standards
 * cleanup: replaced backticks with `$()` in all shell scripts that did not come from elsewhere, as backticks are deprecated and harder to read
 * cleanup: improved the output formatting of "`make test`"
 * cleanup: extended the "`make test`" mechanism to allow certain tests to be skipped on platforms that cannot support them
 * cleanup: skip the "pipe" test (for *SIGPIPE*) if GNU "`head`" is not available, so that "`make test`" on stock OpenBSD 7.3 works
 * cleanup: added a lot more tests to "`make test`"
 * cleanup: replace all calls to `sprintf()` and `snprintf()` with a new wrapper function `pv_snprintf()` to improve security and compatibility
 * cleanup: replace all calls to `strcat()` with a wrapper `pv_strlcat()` to improve security and compatibility
 * cleanup: replace all `write()` calls to the terminal with a wrapper `pv_write_retry()` for consistency
 * cleanup: tidy up and fix compilation warning in "`--watchfd`" code
 * cleanup: rewrote all local shell scripts to pass analysis by [ShellCheck](https://www.shellcheck.net)

### 1.7.0 - 17 July 2023

 * dropped: support for Red Hat Enterprise Linux and its derivatives has been dropped; removed the RPM spec file, and will no longer build binaries
 * feature: the "`--size`" option now accepts "`@filename`" to use the size of another file (pull request [#57](https://codeberg.org/a-j-wood/pv/pull/57) supplied by [Dave Beckett](https://github.com/dajobe))
 * feature: the "`--watchfd`" option is now available on OS X (pull request [#60](https://codeberg.org/a-j-wood/pv/pull/60) supplied by [christoph-zededa](https://github.com/christoph-zededa))
 * feature: new "`--bits`" option to show bit count instead of byte count (adapted from pull request [#63](https://codeberg.org/a-j-wood/pv/pull/63) supplied by [Nick Black](https://nick-black.com))
 * feature: new "`--average-rate-window`" option, to set the window over which the average rate is calculated, also used for ETA (modified from pull request [#65](https://codeberg.org/a-j-wood/pv/pull/65) supplied by [lemonsqueeze](https://github.com/lemonsqueeze))
 * feature: the "`--watchfd`" option will now show relative filenames, if they are under the current directory (pull request [#66](https://codeberg.org/a-j-wood/pv/pull/66) supplied by [ikasty](https://github.com/ikasty))
 * fix: correction to `pv_in_foreground()` to behave as its comment block says it should, when not on a terminal - corrects [#19 "No output in Arch Linux initcpio after 1.6.6"](https://codeberg.org/a-j-wood/pv/issues/19), [#31 "No output written from inside zsh <() construct"](https://codeberg.org/a-j-wood/pv/issues/31), [#55 "pv Stopped Working in the Background"](https://codeberg.org/a-j-wood/pv/issues/55) (pull request [#64](https://codeberg.org/a-j-wood/pv/pull/64) supplied by [Michael Weiß](https://github.com/quitschbo))
 * fix: workaround for OS X 11 behaviour in configure script regarding stat64 at compile time (pull request [#57](https://codeberg.org/a-j-wood/pv/pull/57) supplied by [Dave Beckett](https://github.com/dajobe))
 * fix: workaround for macOS equivalence of stat to stat64 - patches from [Filippo Valsorda](https://github.com/FiloSottile) and [Demitri Muna](https://github.com/demitri), correcting [#33 "Fix compilation problems due to `stat64()` on Apple Silicon"](https://codeberg.org/a-j-wood/pv/issues/33)
 * fix: add burst rate limit to transfer, so rate limits are not broken by bursty traffic (pull request [#62](https://codeberg.org/a-j-wood/pv/pull/62) supplied by [Volodymyr Bychkovyak](https://github.com/vbychkoviak))
 * fix: corrected "`--force`" option so it will still output progress when not in the same process group as the owner of the terminal - corrects [#23 "No output with "`-f`" when run in background after 1.6.6"](https://codeberg.org/a-j-wood/pv/issues/23) and helps to correct [#31 "No output written from inside zsh <() construct"](https://codeberg.org/a-j-wood/pv/issues/31)
 * fix: corrected elapsed time display to show as D:HH:MM:SS after 1 day, like the ETA does - corrects [#16 "Show days in same format in ETA as in elapsed time"](https://codeberg.org/a-j-wood/pv/issues/16)
 * fix: corrected bug where percentages went down after 100% when in "`--numeric`" mode with a "`--size`" that was too small - corrects [#26 "Correct "`-n`" behaviour when going past 100% of "`-s`" size"](https://codeberg.org/a-j-wood/pv/issues/26)
 * i18n: recoded Polish translation file to UTF-8
 * i18n: removed inaccurate fuzzy translation matches
 * docs: moved all open issues into GitHub and updated the TODO list
 * docs: renamed README to README.md and altered it to Markdown format
 * docs: moved contributors from the README to docs/ACKNOWLEDGEMENTS.md
 * docs: moved TODO to TODO.md and altered it to Markdown format
 * docs: moved NEWS to NEWS.md, converted it to UTF-8, and altered it to Markdown format

Changes since 1.1.0:

1.2.0

What's Changed

  * Fix OpenBSD's host info by @danielhejduk in #22
  * radula is now called rad by @firasuke in #23
  * Fix a broken command that led to some slightly broken output
  * Fix broken output from disk info in some conditions
  * Add a fast path for apk package count
  * Add tiny option to PF_PACKAGE_MANAGERS
  * Fix WSL2 detection
  * Fix an issue when unable to determine distro
  * Fix chocolatey package count
  * Add macOS Sequoia support

New Contributors

  * @danielhejduk made their first contribution in #22
  * @firasuke made their first contribution in #23

Full Changelog: Un1q32/pfetch@1.1.0...1.2.0

This release is a bit earlier then usual because I'm going on vacation
and won't be available on the first of July to do the release, I will
return on July 3rd.

Revision history for App-ClusterSSH

4.16 2020-06-20 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Further fix for 'resolve_names' error when adding hosts via the UI
- Fix missing space separator for ssh_args (thanks to Petr Vorel)

4.15 2020-05-18 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Include all utilies within each man page
- Add in 'command_pre' and 'command_post' configuration options
- Fix 'Add Host' menu error finding 'resolved_names'
- Ensure lib path is added to range tests to find the libraries
- Mark permission test as TODO as it appears to be inconsistent

4.14 2019-08-21 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Include README within the repository, not just created tar.gz files
- Add 'autoquit' setting to 'File' menu (Github issue #114)
- Correct macro_hostname to be the FQDN of the server where cssh is
  being run (Github issue #116)
- Add in user defined macros

4.13.2_02 2019-01-14 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Fix Getopt-Long minimum version
- Fix excess test output when Sort::Naturally isn't installed

4.13.2_01 2018-11-24 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Move all Tk code into its own module as-is
- Fix for 'bad pad value "3m"' error when using Tk 804.034
- Update to Perl::Tidy 20181117
- Convert to using Dist::Zilla

4.13.2 2018-03-14 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Fix for running builds in parallel
- Improvements to SUPPORT and REPORTING BUGS sections in documentation

4.13.1 2018-03-05 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Minor update to fix failing tests due to 3rd party perltidy changes

4.13 2017-12-27 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Ensure ssh_args is keep unset if it is emptied in the configuration file
- Obey configured console position (Debian bug 758215) (Github issue #100)

4.12 2017-12-23 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Fix 'undefined value' error

4.11 2017-12-22 Duncan Ferguson <duncan_ferguson@user.sf.net>

- Fix for multiple range expansion, as in 'h{a,b}{1,2}' (Github issue
  #97) (Thanks to lazyfrosch)
- Upgrade Perl::Tidy requirement to version 20171214 (Github issue
  #99) (Thanks to eserte)
- Add in 'external command pipe' to allow for some commands being
  passed in from the command line

4.10_02 2017-08-08 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Include coverage tests in the resources
- Include the version of cssh in the utility documentation and README
- Fix dashes (-) not being accepted in hostname range expansion (Github issue #89)
- Amend ranges to work on ports, FQDN's and IP addresses
- Fix bug tracker links in the main documentation (Github issue #92)
- New options to specify --rows, --columns and --fillscreen (Github
  pull request #88) (Thanks to AsharLohmar)

4.10_01 2017-04-12 Duncan Ferguson <duncan_ferguson@user.sf.net>

- Allow 'include' directives when reading SSH configuration files
  (Github issue #77) (thanks to Azenet)
- Generate README when creating the distribution from cssh man page so
  www.cpan.org and www.metacpan.org can display documentation

4.09 2017-03-11 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Add perl-5.24 Travis-CI automated testing config
- Correct a logic bug around the --debug option (Github issue #75)
- Fix 'Re-add closed windows' not using the correct username (Github issue #72)
- Update copyright year
- Make WM decorations algorithm configurable as causes problems on
  some systems (Debian bug 842965, re Github pull request #66) (thanks
  to Tony Mancill)

4.08 2016-10-18 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Add perl-5.8, 5.10 and 5.12 to Travis-CI automated testing
- Fix building and testing on perl-5.8.9
- Improve testing on systems that do not have xterm installed
- Take into account WM decorations when tiling (Github pull request
  #66) (thanks to Andrew Stevenson)
- Add option in the config file to hide the menu (Github issue #69)
- Add 'unique_servers' into the configuration file to match command
  line option (Github issue #70)

4.07 2016-04-30 Duncan Ferguson <duncan_ferguson@user.sf.net>

- Fixed tests on systems where bash is not installed in /bin/bash
  (Github issue #60)
- Include link to travis-ci site in release emails for automated build
  and test reports
- Rework hostname expansion to be pure-perl rather than relying on the
  bash shell (Github issue #53)

4.06 2016-03-26 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Failure to find the terminal binary should not be fatal
- Fix processing of '--extra_tag_file' and its configuration item
  (Github issue #51)
- Add bash shell expansion on host names containing a '{' character
  (Github issue #53)
- Fix tests when running on a server without xterm installed (such as
  Travis CI via GitHub)
- Expand $HOME and ~ correctly when looking for files (thanks to
  Andrew Stevenson)
- Typo correction in README (thanks to Ankit Vadehra)

4.05 2015-11-28 Duncan Ferguson <duncan_ferguson@user.sf.net>

- Change default key_quit from 'Control-q' to 'Alt-q' (Github issue
  #50)
- Amend tests to always use C locale as some error messages are
  hardcoded in English (Github issue #49)

4.04_01 2015-11-21 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Ensure documentation is generated using same perl as the build
  (Github issue #45)
- Pass '--action' through macro parsing (Github issue #42)
- Workaround for glitch in KDE where windows can become unmoveable
  (Github issue #46) (thanks to Brandon Perkins)
- Add in '--quiet | -Q ' option to reduce output in certian scenarios
- Add in 'csftp' command

4.04 2015-11-03 Duncan Ferguson <duncan_ferguson@user.sf.net>
- Include bash completion script in distribution (Github issue #29)
- Allow re-adding closed session (Github issue #27 - thanks to Andrew
  Stevenson)
- Allow sorting windows in natural order (Github isue #28 - thanks to
  Andrew Stevenson)
- Fix links in metadata files to trackers (Github issue #41)
- Fix ctel and ccon not working correctly (Github issue #35)
- Amend t/10host.t to use a random hostname to prevent clashes (Github
  issue #23)
- Amend coght message in README to match all other files for the
  perl license

2.2.0 (2024-09-16)

What's Changed

New Features

* Add a #close implementation, calling #shutdown on any cached_connection by
  @olleolleolle in #23
* Support verify_hostname SSL option by @stefanmb in #25

Others

* Adds Ruby 3.2 to the CI matrix by @petergoldstein in #16
* Fix linting on CI by @bquorning in #21
* CI: Add Ruby 3.3 by @olleolleolle in #20
* CI: use ruby/setup-ruby in the publish workflow by @olleolleolle in #22

New Contributors

* @petergoldstein made their first contribution in #16
* @bquorning made their first contribution in #21
* @stefanmb made their first contribution in #25

2.3.0 (2024-10-24)

* Update rubocop requirement from ~> 1.66.0 to ~> 1.67.0 (#22)
* bump to excon 1.0 (#23)

v0.2.9
What's Changed

    Migrate to cargo workspaces and introduce github workflows by @triyanox in #14
    feat: add directory indicator #11 by @triyanox in #15
    chore: fix release workflow by @triyanox in #16
    Merge pull request #16 from triyanox/chore/fix-release-workflow by @triyanox in #17
    fix release workflow by @triyanox in #18
    Merge pull request #18 from triyanox/chore/fix-release-workflow by @triyanox in #19
    fix release workflow by @triyanox in #20
    Merge pull request #20 from triyanox/chore/fix-release-workflow by @triyanox in #21
    Fix release workflow by @triyanox in #22
    chore: version crates by @triyanox in #23
    chore: fix the syncing issue on the release workflow by @triyanox in #24
    chore: fix version check by @triyanox in #25
    chore: add explicit version for lla_plugin_interface by @triyanox in #26
    chore: fix path for lla_plugin_interface by @triyanox in #27

v0.2.7
What's Changed

    Add the config --set command to update the lla config by @triyanox in #13

# rnaturalearth 1.0.1

# rnaturalearth 1.0.0

## Breaking changes

This is a breaking changes release that ends support to `sp` object in
favour of more modern interfaces (`sf` and `terra`). Although that
`sp` is still available on CRAN, it is no longer being actively
developed (https://geocompx.org/post/2023/rgdal-retirement/). This is
the main reason that motivated the choice to transition toward `sf`
(the default) and `terra`.

Users can choose either get an `sf` or `SpatVector` using the
`returnclass` argument:

```
ne_countries(returnclass = "sf")
ne_countries(returnclass = "sv")
```

Affected functions are `ne_countries()`, `ne_coastline()`,
`ne_states()`, `ne_load()` and `ne_download()`.

If changing the return type to `sf` creates too many problems to your
existing code, you can still convert it back to `sp` :

```
countries <- ne_countries(returnclass = "sf")

# option 1
sf::as_Spatial(countries)


# option 2
as(countries, "Spatial")
```

More information about the retirement of `rgdal`, `rgeos` and
`maptools`: https://r-spatial.org/r/2022/04/12/evolution.html

## Bugfix

- Correctly downloading and reading raster object (#96, closes #52).

# rnaturalearth 0.3.4

- This is a maintenance release that document/use the new special sentinel "\_PACKAGE".

# rnaturalearth 0.3.3

- Using `lifecycle` to indicate that support of `sp` object will be
  eventually dropped. Users should now use `ne_download(returnclass =
  "sf")`, instead of `ne_download(returnclass = "sp")`.

- `terra` is now included in the Imports section.

# rnaturalearth 0.3.2

- Added new maintainer and contributors ([#62](ropensci/rnaturalearth#62)).
- Using terra over raster ([#63](ropensci/rnaturalearth#63))

  - See
    <https://r-spatial.org/r/2022/04/12/evolution.html#packages-depending-on-sp-and-raster>
    and
    <https://r-spatial.org/r/2022/12/14/evolution2.html#deprecations-in-retiring-packages>

- Fixes broken data download links.  Now using Amazon hosted URLs as
  - the primary source for downloading data
  - ([#48](ropensci/rnaturalearth#48),
  - [#64](ropensci/rnaturalearth#64)).

# rnaturalearth 0.3.0 2021-10-11

- fix rnaturalearthhires installation #47 thank you Ian Taylor for #43

# rnaturalearth 0.2.0

- add to river options in ne_download() by adding to
  data_list_physical.csv fixing
  [#23](ropensci/rnaturalearth#23)

- update data to new version [Natural Earth
  v4.1](https://www.naturalearthdata.com/blog/miscellaneous/natural-earth-v4-1-0-release-notes/)
  released May 2018.

# rosm 0.3.0

* Silent deprecation the entire previous API (#20).
* Drop rgdal dependency (#21).
* Add new API based on wk and the curl package's multi
  download interface (#23).
* Update test and CI infrastructure (#27).
* Added NEWS.md to track changes in this package.

# rosm 0.2.5

* Updates to ensure compliance with CRAN policies.

# farver 2.1.2

* make sure output channels are always named (#23)
* farver can now decode short version hex-code colours (#41)
* Fix a bug in colour comparison that affected hcl input (#44)

# bit64 4.6.0-1

## NOTICE OF PLANNED BREAKING CHANGES

1. {bit64} exports many S3 methods directly. Calling S3 methods
directly is generally bad form; we should rely on the S3 dispatch
system for this. Needing to export an S3 method is usually indicative
of some deep issue that's otherwise hard to work around.

  I plan to un-export most if not all S3 methods in future
  versions. In this release, there will be no change in behavior
  besides this notice in the NEWS. Going forward, I see two types of
  S3 exports: (1) exports that have no discoverable direct usage (that
  is, a global GitHub search, which includes the CRAN mirror, turned
  up _no_ R code calling them directly, except perhaps in `:::` form,
  which would be unaffected by un-export); and (2) exports that _are_
  observed to be called directly by some number of downstreams. With
  the former, I am more comfortable un-exporting more aggressively;
  with the latter, I will take a more gradual approach.

  Here are the S3 methods that are currently exported, for which I
  found no record of them being called directly:

  `-.integer64`, `:.default`, `:.integer64`, `!.integer64`,
  `!=.integer64`, `[.integer64`, `[[.integer64`, `[[<-.integer64`,
  `*.integer64`, `/.integer64`, `&.integer64`, `%/%.integer64`,
  `%%.integer64`, `%in%.default`, `%in%.integer64`, `^.integer64`,
  `+.integer64`, `<.integer64`, `<=.integer64`, `==.integer64`,
  `>.integer64`, `>=.integer64`, `|.integer64`, `all.equal.integer64`,
  `as.bitstring.integer64`, `as.integer64.factor`,
  `as.integer64.integer64`, `as.integer64.NULL`, `as.list.integer64`,
  `as.logical.integer64`, `cbind.integer64`, `ceiling.integer64`,
  `cummax.integer64`, `cummin.integer64`, `cumprod.integer64`,
  `cumsum.integer64`, `diff.integer64`, `duplicated.integer64`,
  `floor.integer64`, `hashdup.cache_integer64`,
  `hashfin.cache_integer64`, `hashfun.integer64`, `hashmap.integer64`,
  `hashmaptab.integer64`, `hashmapuni.integer64`,
  `hashmapupo.integer64`, `hashpos.cache_integer64`,
  `hashrev.cache_integer64`, `hashrin.cache_integer64`,
  `hashtab.cache_integer64`, `hashuni.cache_integer64`,
  `hashupo.cache_integer64`, `is.double.default`,
  `is.double.integer64`, `is.finite.integer64`,
  `is.infinite.integer64`, `is.nan.integer64`, `is.sorted.integer64`,
  `is.vector.integer64`, `keypos.integer64`, `length<-.integer64`,
  `log10.integer64`, `log2.integer64`, `match.default`,
  `match.integer64`, `mean.integer64`, `median.integer64`,
  `mergeorder.integer64`, `mergesort.integer64`,
  `mergesortorder.integer64`, `na.count.integer64`, `nties.integer64`,
  `nunique.integer64`, `nvalid.integer64`, `order.default`,
  `order.integer64`, `orderdup.integer64`, `orderfin.integer64`,
  `orderkey.integer64`, `ordernut.integer64`, `orderpos.integer64`,
  `orderqtl.integer64`, `orderrnk.integer64`, `ordertab.integer64`,
  `ordertie.integer64`, `orderuni.integer64`, `orderupo.integer64`,
  `prank.integer64`, `print.bitstring`, `prod.integer64`,
  `qtile.integer64`, `quantile.integer64`, `quickorder.integer64`,
  `quicksort.integer64`, `quicksortorder.integer64`,
  `radixorder.integer64`, `radixsort.integer64`,
  `radixsortorder.integer64`, `ramorder.integer64`,
  `ramsort.integer64`, `ramsortorder.integer64`, `range.integer64`,
  `rank.default`, `rbind.integer64`, `round.integer64`,
  `scale.integer64`, `shellorder.integer64`, `shellsort.integer64`,
  `shellsortorder.integer64`, `sign.integer64`, `signif.integer64`,
  `sort.integer64`, `sortfin.integer64`, `sortnut.integer64`,
  `sortorderdup.integer64`, `sortorderkey.integer64`,
  `sortorderpos.integer64`, `sortorderrnk.integer64`,
  `sortordertab.integer64`, `sortordertie.integer64`,
  `sortorderuni.integer64`, `sortorderupo.integer64`,
  `sortql.integer64`, `sorttab.integer64`, `sortuni.integer64`,
  `sqrt.integer64`, `summary.integer64`, `table.integer64`,
  `tiepos.integer64`, `trunc.integer64`, `unipos.integer64`

  Here are the S3 methods that are currently exported for which I _do_
  find record of them being called directly:

  `abs.integer64`, `as.character.integer64`,
  `as.data.frame.integer64`, `as.double.integer64`,
  `as.integer.integer64`, `as.integer64.bitstring`,
  `as.integer64.character`, `as.integer64.double`,
  `as.integer64.integer`, `as.integer64.logical`, `c.integer64`,
  `format.integer64`, `identical.integer64`, `is.na.integer64`,
  `lim.integer64`, `max.integer64`, `min.integer64`,
  `print.integer64`, `rank.integer64`, `seq.integer64`,
  `str.integer64`, `sum.integer64`, `unique.integer64`

  In the next release (provisionally, 4.7.0), I will add a `warning()`
  to any S3 method in the former classification, while nothing will
  change for the latter classification. I may reach out to authors
  observed to call the methods directly.

  In the subsequent release (provisionally, 4.8.0), I will un-export
  any S3 method in the former classification, and add a `warning()` to
  any S3 method in the latter classification.

  In the sub-subsequent release (provisionally, 4.9.0), I will
  un-export any S3 method in the latter classification.

  Please reach out (e.g., the GitHub log for #76) if you have any
  concerns about this plan.

  1. {bit64} lists {bit} as `Depends:`. IMO this form of dependency
  should be deprecated by R now that `Imports:` is widely available and
  well-supported for many years.

  In the next release (provisionally, 4.7.0), I will move bit to
  Imports. The practical implication is that currently,
  `library(bit64)` will make {bit} objects like `is.bit()` available
  for use without namespace-qualification. This practice makes code
  harder to read and maintain.

  Users relying on this in scripts can (1) write `library(bit)` to
  attach {bit} explicitly or (2) namespace-qualify all {bit} calls
  with `bit::`.

  Package authors relying on this can (1) add `import(bit)` to make
  the full {bit} namespace available or (2) namespace-qualify all
  {bit} calls with `bit::`; adding {bit} to `Imports:` or `Suggests:`
  will also be necessary.

  I will reach out to CRAN authors with any required
  changes. Depending on the impact size, I might make this transition
  more gradual (e.g. starting by re-exporting some or all {bit}
  functions from {bit64}, with warning, before un-exporting them in a
  subsequent release).

## NEW FEATURES

1. Implemented S3 methods for `rowSums()` and `colSums()`. Importantly
they handle `NA` values correctly, #38. Thanks @vlulla for the
request. Note that these are implemented as wrappers to `apply()`
calls, so they may not be as efficient. PRs welcome for implementing
the efficient equivalents.

  Note that by necessity, this grows the set of base exports
  overwritten to include `rowSums()` and `colSums()`, which are
  exported as S3 generics dispatching to `base::rowSums()` and
  `base::colSums()` by default.

1. Partially powering this is a new `aperm()` method for integer64
which allows `apply()` to work as intended. Using `apply()` directly
may still strip the integer64 class; that may be supported later (see
#87).

1. `is.na()` is supported for long vector input (more than `2^31`
elements), #30. Thanks @ilia-kats for the request. Long vector support
will be added on an as-needed basis as I don't have a great machine
for testing these features -- PRs welcome!

## BUG FIXES

1. `all.equal.integer64()` gets the same fix for vector `scale=` to
work as intended that `all.equal.numeric()` got in R 4.1.3, #23.

1. Made edits to `match()` to handle `is.integer64(table)` better for
older versions of R, including a new `mtfrm()` method for integer64
objects in R>=4.2.0, #85 and #111.

## NOTES

1. After creating, developing, and maintaining {bit64} for about 13
years, Jens Oehlschlägel has decided to step down as maintainer of the
package. Michael Chirico will take over in this duty. Thank you Jens
for creating such a wonderful & important part of the R ecosystem!

  I don't have any major plans for new features, and mostly hope to
  keep the package running and up to date. Contributors most welcome!
  I am also trying to freshen up the code base to make contribution
  easier.

1. The R version dependency has increased from 3.0.1 (May 2013) to
3.4.0 (April 2017). We plan to keep roughly the same R dependency as
{data.table}, i.e., as old as possibly for as long as possible, with
some bias towards gradually bringing in new R features to reduce the
maintenance overhead of a growing nest of workarounds to keep the
package "fresh" for users of the latest R versions.

  Required package {bit} already requires R 3.4.0, so the old 3.0.1
  requirement was effectively impossible anyway.

1. Default packages {methods}, {stats}, and {utils} are now
`Imports:`, not `Depends:`, dependencies. `Depends:` is an out-dated
mode of dependency in R. This will only affect the small audience of
users that run R with `R_DEFAULT_PACKAGES=NULL` (or some other subset
excluding some of these three), _and_ who are relying (perhaps
implicitly) on {bit64} being responsible for attaching those packages.

  It is my intention to move {bit} from `Depends:` to `Imports:` as
  well, but this migration will be done more gingerly -- it is more
  conceivable that this will constitute a breaking change for some use
  cases, therefore it will be done in phases. Nothing is done in this
  release, but here is your earliest warning that from the next
  release, it will be a warning to rely on {bit64} to attach {bit}
  functions for you.

1. Package documentation is now managed with {roxygen2}, #61. I tried
to retain everything in the original documentation, but the diff
required to do so was quite unmanageable (5,000+ lines), so please
alert me if anything looks amiss. Most importantly, I ensured the
NAMESPACE remains unchanged.

1. The signature of `identical.integer64()` loses `extptr.as.ref=`,
which is unavailable for R<4.2.0, but gains `...` to allow this
argument in newer versions, #37. This retains the transparency of
having all arguments named in the signature (and thus in
`?identical.integer64` as well as available for tab-completion) while
also retaining the old R version dependency R 3.3.0.

# bit64 NEWS for versions 0.8-3 through 4.5.2 are now in
  [NEWS.0](https://github.com/r-lib/bit64/blob/master/NEWS.0)

## Changes in version 0.23

Released on March, 29, 2025

* Issue #101: fix release tarballs by using `make dist` instead of
  alternative ad hoc method.

* Issue #76: `atf_check`: fix `std::length_error` thrown from `temp_file`.

* #87: Update documentation/build files to point to FreeBSD repo/mailing lists.

* #86: Improve diagnostics when paths cannot be opened

* #85: various build-related changes. Highlights:
    * Remove leftover references of `auto_array` (bugfix).
    * Remove the `atf-*-api` manpages (enhancement).
    * Require -std=C++14 with ATF (enhancement).
    * Expose `WCOREDUMP(..)` on platforms that support it (bugfix).
    * Add `__attribute__((nonnull))` support (enhancement).
    * Restore `AC_PROG_CPP` (bugfix).

### General fixes

* Fix various style issues which were blocking `make dist` from completing
  cleanly.

* Fix `atf_utils_file_exists` declaration in atf-c(3).

* Fix various memory leaks and logic errors.

* `atf-test-case(4)`: update reference to the "descr" metadata property so it's
  marked Optional.

* `atf-check.cpp`: remove unnecessary data copy in `temp_file`.

## Changes in version 0.22

Released on November 25th, 2024.

* Issue #23: Fix double-free triggered by atf_map_insert in low memory
  scenarios, caused by an overlook in the atf_list code.

* Issue #31: Added require.progs metadata properties to the tests that
  need a compiler to run.

* Added the atf_check_not_equal function to atf-sh to check for
  unequal values.

* Add `-r timeout` flag to `atf-check`.

* Open results files before executing tests to fix an issue ATF tests that
  adjust the processes' Capsicum rights as part of the testcase(s)
  executed.

* Add Cirrus CI integration for FreeBSD CI/CD support.

* Address compilation issues on OpenSolaris distributions.

* Replace `auto_array` with `std::vector` (fixes modern C++ compliance).

* Replace `auto_ptr` with `std::shared_ptr` (fixes modern C++ compliance).

* Update autotools idioms and requirements. The minimum required version of
  autoconf is now 2.68.

* Modify `ATF_CHECK*` and `ATF_REQUIRE*` macros to avoid double
  evaluation/allow the gcc toolchain to detect accidental single
  assignment typos.

* Always define CPP to fix use of ATF_BUILD_CPP when the user did not
  define CPP when invoking the configure script.

### General fixes

* Fix various typos and formatting errors in manual pages and markdown
  documents.

## Changes in version 0.14.1

**Released on March 29th, 2025.**

* Issue #263: fix release tarball by relying on `make dist`.

* Document ATF to Kyua mappings for metadata properties.

* Support `require.diskspace` ATF metadata property.

* Support backwards compatible `unprivileged_user` metadata property for ease
  of use dealing with ATF vs Kyua.

* Fix building from source by tracking `utils/process/Kyuafile.in` and
  `engine/execenv/execenv_fwd.hpp` with automake.

## Changes in version 0.14

**Released on December 31st, 2024.**

* Replace `--with-atf` with `--enable-atf`. The latter must be specified
  to enable ATF support and the Kyua tests by proxy.

* Require newer versions of ATF (0.21) and Lutok (0.5). This is being done
  as part of a multi-phase deprecation of pre-C++11 environments.

* Vendor the ATF m4 files from the freebsd/atf repo.

* Translate `is_exclusive` metadata to its ATF counterpart.

* Issue #236: fix accidental test forkbomb in specific user environments
  with `integration` tests by avoiding infinite recursion calling kyua.

* Fix the build on Ubuntu 24.04.

- execenv support added. This FreeBSD-specific feature allows tests to be
  executed in jails.

* Issue #169: Remove deprecated `readdir_r` use.

* Fix build issue on more recent versions of MacOS.

* Fix build issue on IllumOS.

* Fix issues with PID reuse in testcases.

* Escape characters captured from stderr and stdout in HTML output.

* `integration/cmd_report_test:verbose`: fix regular expression used in
  tests which would break in certain user environments.

* Add Cirrus CI support for testing Kyua on FreeBSD.

* Remove `UTILS_UNUSED_PARAM` macro.

* Fix document rendering with newer versions of doxygen.

* Issue #152: explicitly require C++11 language features when compiling Kyua.
  Remove `std::shared_ptr` compatibility header, `utils/shared_ptr.hpp`.

* Fix `igor` and `mandoc -Tlint` reported issues with generated manpages.

* Issue #167: fix typos in kyua(1) manpage.

## Changes in version 0.13

**Released on August 26th, 2016.**

* Fixed execution of test cases as an unprivileged user, at least under
  NetBSD 7.0.  Kyua-level failures were probably a regression introduced
  in Kyua 0.12, but the underlying may have existed for much longer:
  test cases might have previously failed for mysterious reasons when
  running under an unprivileged user.

* Issue #134: Fixed metadata test broken on 32-bit platforms.

* Issue #139: Added per-test case start/end timestamps to all reports.

* Issue #156: Fixed crashes due to the invalid handling of cleanup
  routine data and triggered by the reuse of PIDs in long-running Kyua
  instances.

* Issue #159: Fixed TAP parser to ignore case while matching `TODO` and
  `SKIP` directives, and to also recognize `Skipped`.

* Fixed potential crash due to a race condition in the unprogramming of
  timers to control test deadlines.


## Changes in version 0.12

**Released on November 22nd, 2015.**

This is a huge release and marks a major milestone for Kyua as it finally
implements a long-standing feature request: the ability to execute test
cases in parallel.  This is a big deal because test cases are rarely
CPU-bound: running them in parallel yields much faster execution times for
large test suites, allowing faster iteration of changes during development.

As an example: the FreeBSD test suite as of this date contains 3285 test
cases.  With sequential execution, a full test suite run takes around 12
minutes to complete, whereas on a 4-core machine with a high level of
parallelism it takes a little over 1 minute.

Implementing parallel execution required rewriting most of Kyua's core and
partly explains explains why there has not been a new release for over a
year.  The current implementation is purely subprocess-based, which works
but has some limitations and has resulted in a core that is really complex
and difficult to understand.  Future versions will investigate the use of
threads instead for a simplified programming model and additional
parallelization possibilities.

* Issue #2: Implemented support to execute test cases in parallel when
  invoking `kyua test`.  Parallel execution is *only* enabled when the new
  `parallelism` configuration variable is set to a value greater than `1`.
  The default behavior is still to run tests sequentially because some test
  suites contain test cases with side-effects that might fail when run in
  parallel.  To resolve this, the new metadata property `is_exclusive` can
  be set to `true` on a test basis to indicate that the test must be run on
  its own.

* Known regression: Running `kyua debug` on a TAP-based test program does
  not currently report the output in real time.  The output will only be
  displayed once the test program completes.  This is a shortcoming of
  the new parallel execution engine and will be resolved.

* Removed the external C-based testers code in favor of the new built-in
  implementations.  The new approach feels significantly faster than the
  previous one.

* Fixed the handling of relative paths in the `fs.*` functions available
  in `Kyuafile`s.  All paths are now resolved relative to the location of
  the caller `Kyuafile`.  `Kyuafile.top` has been updated with these
  changes and you should update custom copies of this file with the new
  version.

* Changed temporary directory creation to always grant search
  permissions on temporary directories.  This is to prevent potential
  problems when running Kyua as root and executing test cases that require
  dropping privileges (as they may later be unable to use absolute paths
  that point inside their work directory).

* The cleanup of work directories does not longer attempt to deal with
  mount points.  If a test case mounts a file system and forgets to unmount
  it, the mount point will be left behind.  It is now the responsibility of
  the test case to clean after itself.  The reasons for this change are
  simplicity and clarity: there are many more things that a test case can
  do that have side-effects on the system and Kyua cannot protect against
  them all, so it is better to just have the test undo anything it might
  have done.

* Improved `kyua report --verbose` to properly handle environment
  variables with continuation lines in them, and fixed the integration
  tests for this command to avoid false negatives.

* Changed the configuration file format to accept the definition of
  unknown variables without declaring them local.  The syntax version
  number remains at 2.  This is to allow configuration files for newer Kyua
  versions to work on older Kyua versions, as there is no reason to forbid
  this.

* Fixed stacktrace gathering with FreeBSD's ancient version of GDB.
  GDB 6.1.1 (circa 2004) does not have the `-ex` flag so we need to
  generate a temporary GDB script and feed it to GDB with `-x` instead.

* Issue #136: Fixed the XML escaping in the JUnit output so that
  non-printable characters are properly handled when they appear in the
  process's stdout or stderr.

* Issue #141: Improved reporting of errors triggered by sqlite3.  In
  particular, all error messages are now tagged with their corresponding
  database filename and, if they are API-level errors, the name of the
  sqlite3 function that caused them.

* Issue #144: Improved documentation on the support for custom properties
  in the test metadata.

* Converted the `INSTALL`, `NEWS`, and `README` distribution documents to
  Markdown for better formatting online.


## Changes in version 0.11

**Released on October 23rd, 2014.**

* Added support to print the details of all test cases (metadata and
  their output) to `report`.  This is via a new `--verbose` flag which
  replaces the previous `--show-context`.

* Added support to specify the amount of physical disk space required
  by a test case.  This is in the form of a new `required_disk_space`
  metadata property, which can also be provided by ATF test cases as
  `require.diskspace`.

* Assimilated the contents of all the `kyua-*-tester(1)` and
  `kyua-*-interface(7)` manual pages into more relevant places.  In
  particular, added more details on test program registration and their
  metadata to `kyuafile(5)`, and added `kyua-test-isolation(7)`
  describing the isolation features of the test execution.

* Assimilated the contents of all auxiliary manual pages, including
  `kyua-build-root(7)`, `kyua-results-files(7)`, `kyua-test-filters(7)`
  and `kyua-test-isolation(7)`, into the relevant command-specific
  manual pages.  This is for easier discoverability of relevant
  information when reading how specific Kyua commands work.

* Issue #30: Plumbed through support to query configuration variables
  from ATF's test case heads.  This resolves the confusing situation
  where test cases could only do this from their body and cleanup
  routines.

* Issue #49: Extended `report` to support test case filters as
  command-line arguments.  Combined with `--verbose`, this allows
  inspecting the details of a test case failure after execution.

* Issue #55: Deprecated support for specifying `test_suite` overrides on
  a test program basis.  This idiom should not be used but support for
  it remains in place.

* Issue #72: Added caching support to the `getcwd(3)` test in configure
  so that the result can be overriden for cross-compilation purposes.

* Issue #83: Changed manual page headings to include a `kyua` prefix in
  their name.  This prevents some possible confusion when displaying,
  for example, the `kyua-test` manual page with a plain name of `test`.

* Issue #84: Started passing test-suite configuration variables to plain
  and TAP test programs via the environment.  The name of the
  environment variables set this way is prefixed by `TEST_ENV_`, so a
  configuration variable of the form
  `test_suites.some_name.allow_unsafe_ops=yes` in `kyua.conf` becomes
  `TEST_ENV_allow_unsafe_ops=YES` in the environment.

* Issues #97 and #116: Fixed the build on Illumos.

* Issue #102: Set `TMPDIR` to the test case's work directory when running
  the test case.  If the test case happens to use the `mktemp(3)` family
  of functions (due to misunderstandings on how Kyua works or due to
  the reuse of legacy test code), we don't want it to easily escape the
  automanaged work directory.

* Issue #103: Started being more liberal in the parsing of TAP test
  results by treating the number in `ok` and `not ok` lines as optional.

* Issue #105: Started using tmpfs instead of md as a temporary file
  system for tests in FreeBSD so that we do not leak `md(4)` devices.

* Issue #109: Changed the privilege dropping code to start properly
  dropping group privileges when `unprivileged_user` is set.  Also fixes
  `testers/run_test:fork_wait__unprivileged_group`.

* Issue #110: Changed `help` to display version information and clarified
  the purpose of the `about` command in its documentation.

* Issue #111: Fixed crash when defining a test program in a `Kyuafile`
  that has not yet specified the test suite name.

* Issue #114: Improved the `kyuafile(5)` manual page by clarifying the
  restrictions of the `include()` directive and by adding abundant
  examples.


## Changes in version 0.10

**Experimental version released on August 14th, 2014.**

* Merged `kyua-cli` and `kyua-testers` into a single `kyua` package.

* Dropped the `kyua-atf-compat` package.

* Issue #100: Do not try to drop privileges to `unprivileged_user` when we
  are already running as an unprivileged user.  Doing so is not possible
  and thus causes spurious test failures when the current user is not
  root and the current user and `unprivileged_user` do not match.

* Issue #79: Mention `kyua.conf(5)` in the *See also* section of `kyua(1)`.

* Issue #75: Change the `rewrite__expected_signal__bad_arg` test in
  `testers/atf_result_test` to use a different signal value.  This is to
  prevent triggering a core dump that made the test fail in some platforms.


## Changes in kyua-cli version 0.9

**Experimental version released on August 8th, 2014.**

Major changes:

The internal architecture of Kyua to record the results of test suite
runs has completely changed in this release.  Kyua no longer stores all
the different test suite run results as different "actions" within the
single `store.db` database.  Instead, Kyua now generates a separate
results file inside `~/.kyua/store/` for every test suite run.

Due to the complexity involved in the migration process and the little
need for it, this is probably going to be the only release where the
`db-migrate` command is able to convert an old `store.db` file to the
new scheme.

Changes in more detail:

* Added the `report-junit` command to generate JUnit XML result files.
  The output has been verified to work within Jenkins.

* Switched to results files specific to their corresponding test suite
  run.  The unified `store.db` file is now gone: `kyua test` creates a
  new results file for every invocation under `~/.kyua/store/` and the
  `kyua report*` commands are able to locate the latest file for a
  corresponding test suite automatically.

* The `db-migrate` command takes an old `store.db` file and generates
  one results file for every previously-recorded action, later deleting
  the `store.db` file.

* The `--action` flag has been removed from all commands that accepted
  it.  This has been superseded by the tests results files.

* The `--store` flag that many commands took has been renamed to
  `--results-file` in line with the semantical changes.

* The `db-exec` command no longer creates an empty database when none
  is found.  This command is now intended to run only over existing
  files.


## Changes in kyua-testers version 0.3

**Experimental version released on August 8th, 2014.**

* Made the testers set a "sanitized" value for the `HOME` environment
  variable where, for example, consecutive and trailing slashes have
  been cleared.  Mac OS X has a tendency to append a trailing slash to
  the value of `TMPDIR`, which can cause third-party tests to fail if
  they compare `${HOME}` with `$(pwd)`.

* Issues #85, #86, #90 and #92: Made the TAP parser more complete: mark
  test cases reported as `TODO` or `SKIP` as passed; handle skip plans;
  ignore lines that look like `ok` and `not ok` but aren't results; and
  handle test programs that report a pass but exit with a non-zero code.


## Changes in kyua-cli version 0.8

**Experimental version released on December 7th, 2013.**

* Added support for Lutok 0.4.

* Issue #24: Plug the bootstrap tests back into the test suite.  Fixes
  in `kyua-testers` 0.2 to isolate test cases into their own sessions
  should allow these to run fine.

* Issue #74: Changed the `kyuafile(5)` parser to automatically discover
  existing tester interfaces.  The various `*_test_program()` functions
  will now exist (or not) based on tester availability, which simplifies
  the addition of new testers or the selective installation of them.


## Changes in kyua-testers version 0.2

**Experimental version released on December 7th, 2013.**

* Issue #74: Added the `kyua-tap-tester`, a new backend to interact with
  test programs that comply with the Test Anything Protocol.

* Issue #69: Cope with the lack of `AM_PROG_AR` in `configure.ac`, which
  first appeared in Automake 1.11.2.  Fixes a problem in Ubuntu 10.04
  LTS, which appears stuck in 1.11.1.

* Issue #24: Improve test case isolation by confining the tests to their
  own session instead of just to their own process group.


## Changes in kyua-cli version 0.7

**Experimental version released on October 18th, 2013.**

* Made failures from testers more resilent.  If a tester fails, the
  corresponding test case will be marked as broken instead of causing
  kyua to exit.

* Added the `--results-filter` option to the `report-html` command and
  set its default value to skip passed results from HTML reports.  This
  is to keep these reports more succint and to avoid generating tons of
  detail files that will be, in general, useless.

* Switched to use Lutok 0.3 to gain compatibility with Lua 5.2.

* Issue #69: Cope with the lack of `AM_PROG_AR` in `configure.ac`, which
  first appeared in Automake 1.11.2.  Fixes a problem in Ubuntu 10.04
  LTS, which appears stuck in 1.11.1.


## Changes in kyua-cli version 0.6

**Experimental version released on February 22nd, 2013.**

* Issue #36: Changed `kyua help` to not fail when the configuration file
  is bogus.  Help should always work.

* Issue #37: Simplified the `syntax()` calls in configuration and
  `Kyuafile` files to only specify the requested version instead of also
  the format name.  The format name is implied by the file being loaded, so
  there is no use in the caller having to specify it.  The version number
  of these file formats has been bumped to 2.

* Issue #39: Added per-test-case metadata values to the HTML reports.

* Issue #40: Rewrote the documentation as manual pages and removed the
  previous GNU Info document.

* Issue #47: Started using the independent testers in the `kyua-testers`
  package to run the test cases.  Kyua does not implement the logic to
  invoke test cases any more, which provides for better modularity,
  extensibility and robustness.

* Issue #57: Added support to specify arbitrary metadata properties for
  test programs right from the `Kyuafile`.  This is to make plain test
  programs more versatile, by allowing them to specify any of the
  requirements (allowed architectures, required files, etc.) supported
  by Kyua.

* Reduced automatic screen line wrapping of messages to the `help`
  command and the output of tables by `db-exec`.  Wrapping any other
  messages (specially anything going to stderr) was very annoying
  because it prevented natural copy/pasting of text.

* Increased the granularity of the error codes returned by `kyua(1)` to
  denote different error conditions.  This avoids the overload of `1` to
  indicate both "expected" errors from specific subcommands and
  unexpected errors caused by the internals of the code.  The manual now
  correctly explain how the exit codes behave on a command basis.

* Optimized the database schema to make report generation almost
  instantaneous.

* Bumped the database schema to 2.  The database now records the
  metadata of both test programs and test cases generically, without
  knowledge of their interface.

* Added the `db-migrate` command to provide a mechanism to upgrade a
  database with an old schema to the current schema.

* Removed the GDB build-time configuration variable.  This is now part
  of the `kyua-testers` package.

* Issue #31: Rewrote the `Kyuafile` parsing code in C++, which results in
  a much simpler implementation.  As a side-effect, this gets rid of the
  external Lua files required by `kyua`, which in turn make the tool
  self-contained.

* Added caching of various configure test results (particularly in those
  tests that need to execute a test program) so that cross-compilers can
  predefine the results of the tests without having to run the
  executables.


## Changes in kyua-testers version 0.1

**Experimental version released on February 19th, 2013.**

This is the first public release of the `kyua-testers` package.

The goal of this first release is to adopt all the test case execution
code of `kyua-cli` 0.5 and ship it as a collection of independent tester
binaries.  The `kyua-cli` package will rely on these binaries to run the
tests, which provides better modularity and simplicity to the
architecture of Kyua.

The code in this package is all C as opposed to the current C++ codebase
of `kyua-cli`, which means that the overall build times of Kyua are now
reduced.


## Changes in kyua-cli version 0.5

**Experimental version released on July 10th, 2012.**

* Issue #15: Added automatic stacktrace gathering of crashing test cases.
  This relies on GDB and is a best-effort operation.

* Issue #32: Added the `--build-root` option to the debug, list and test
  commands.  This allows executing test programs from a different
  directory than where the `Kyuafile` scripts live.  See the *Build roots*
  section in the manual for more details.

* Issue #33: Removed the `kyuaify.sh` script.  This has been renamed to
  atf2kyua and moved to the `kyua-atf-compat` module, where it ships as a
  first-class utility (with a manual page and tests).

* Issue #34: Changed the HTML reports to include the stdout and stderr of
  every test case.

* Fixed the build when using a "build directory" and a clean source tree
  from the repository.


## Changes in kyua-cli version 0.4

**Experimental version released on June 6th, 2012.**

* Added the `report-html` command to generate HTML reports of the
  execution of any recorded action.

* Changed the `--output` flag of the `report` command to only take a
  path to the target file, not its format.  Different formats are better
  supported by implementing different subcommands, as the options they
  may receive will vary from format to format.

* Added a `--with-atf` flag to the configure script to control whether
  the ATF tests get built or not.  May be useful for packaging systems
  that do not have ATF in them yet.  Disabling ATF also cuts down the
  build time of Kyua significantly, but with the obvious drawbacks.

* Grouped `kyua` subcommands by topic both in the output of `help` and
  in the documentation.  In general, the user needs to be aware of
  commands that rely on a current project and those commands that rely
  purely on the database to generate reports.

* Made `help` print the descriptions of options and commands properly
  tabulated.

* Changed most informational messages to automatically wrap on screen
  boundaries.

* Rewrote the configuration file parsing module for extensibility.  This
  will allow future versions of Kyua to provide additional user-facing
  options in the configuration file.

  No syntax changes have been made, so existing configuration files
  (version 1) will continue to be parsed without problems.  There is one
  little exception though: all variables under the top-level
  `test_suites` tree must be declared as strings.

  Similarly, the `-v` and `--variable` flags to the command line must
  now carry a `test_suites.` prefix when referencing any variables under
  such tree.


## Changes in kyua-cli version 0.3

**Experimental version released on February 24th, 2012.**

* Made the `test` command record the results of the executed test
  cases into a SQLite database.  As a side effect, `test` now supports a
  `--store` option to indicate where the database lives.

* Added the `report` command to generate plain-text reports of the
  test results stored in the database.  The interface of this command is
  certainly subject to change at this point.

* Added the `db-exec` command to directly interact with the store
  database.

* Issue #28: Added support for the `require.memory` test case property
  introduced in ATF 0.15.

* Renamed the user-specific configuration file from `~/.kyuarc` to
  `~/.kyua/kyua.conf` for consistency with other files stored in the
  `~/.kyua/` subdirectory.

* Switched to use Lutok instead of our own wrappers over the Lua C
  library.  Lutok is just what used to be our own utils::lua module, but
  is now distributed separately.

* Removed the `Atffile`s from the source tree.  Kyua is stable enough
  to generate trustworthy reports, and we do not want to give the
  impression that atf-run / atf-report are still supported.

* Enabled logging to stderr for our own test programs.  This makes it
  slightly easier to debug problems in our own code when we get a
  failing test.


## Changes in kyua-cli version 0.2

**Experimental version released on August 24th, 2011.**

The biggest change in this release is the ability for Kyua to run test
programs implemented using different frameworks.  What this means is
that, now, a Kyua test suite can include not only ATF-based test
programs, but also "legacy" (aka plain) test programs that do not use
any framework.  I.e. if you have tests that are simple programs that
exit with 0 on success and 1 on failure, you can plug them in into a
Kyua test suite.

Other than this, there have been several user-visible changes.  The most
important are the addition of the new `config` and `debug` subcommands
to the `kyua` binary.  The former can be used to inspect the runtime
configuration of Kyua after parsing, and the latter is useful to
interact with failing tests cases in order to get more data about the
failure itself.

Without further ado, here comes the itemized list of changes:

* Generalized the run-time engine to support executing test programs
  that implement different interfaces.  Test programs that use the ATF
  libraries are just a special case of this.  (Issue #18.)

* Added support to the engine to run `plain` test programs: i.e. test
  programs that do not use any framework and report their pass/fail
  status as an exit code.  This is to simplify the integration of legacy
  test programs into a test suite, and also to demonstrate that the
  run-time engine is generic enough to support different test
  interfaces.  (Issue #18.)

* Added the `debug` subcommand.  This command allows end users to tweak
  the execution of a specific test case and to poke into the behavior of
  its execution.  At the moment, all this command allows is to view the
  stdout and stderr of the command in real time (which the `test`
  command currently completely hides).

* Added the `config` subcommand.  This command allows the end user to
  inspect the current configuration variables after evaluation, without
  having to read through configuration files.  (Issue #11.)

* Removed the `test_suites_var` function from configuration files.  This
  was used to set the value of test-suite-sepecific variables, but it
  was ugly-looking.  It is now possible to use the more natural syntax
  `test_suites.<test-suite-name>.<variable> = <value>`.  (Issue #11.)

* Added a mechanism to disable the loading of configuration files
  altogether.  Needed for testing purposes and for scriptability.
  Available by passing the `--config=none` flag.

* Enabled detection of unused parameters and variables in the code and
  fixed all warnings.  (Issue #23.)

* Changed the behavior of "developer mode".  Compiler warnings are now
  enabled unconditionally regardless of whether we are in developer mode
  or not; developer mode is now only used to perform strict warning
  checks and to enable assertions.  Additionally, developer mode is now
  only automatically enabled when building from the repository, not for
  formal releases.  (Issue #22.)

* Fixed many build and portability problems to Debian sid with GCC 4.6.3
  and Ubuntu 10.04.1 LTS.  (Issues #20, #21, #26.)


## Changes in kyua-cli version 0.1

**Experimental version released on June 23rd, 2011.**

This is the first public release of the `kyua-cli` package.

The scope of this release is to provide functional replacement for the
`atf-run` utility included in the atf package.  At this point, `kyua`
can reliably run the NetBSD 5.99.53 test suite delivering the same
results as `atf-run`.

The reporting facilities of this release are quite limited.  There is
no replacement for `atf-report` yet, and there is no easy way of
debugging failing test programs other than running them by hand.  These
features will mark future milestones and therefore be part of other
releases.

Be aware that this release has suffered very limited field testing.
The test suite for `kyua-cli` is quite comprehensive, but some bugs may
be left in any place.

Upstream changes:
v6.0.2 - 2024-07-14
 - Fix removal of scripts/pod2man and scripts/pod2text by make realclean,
   broken in the v6.0.0 release.  Thanks to James E Keenan for the report.
v6.0.1 - 2024-07-12
 - Remove autodie from the module build process.  When built as part of
   Perl core, podlators is built before autodie is available.  Thanks to
   James E Keenan for the report and a draft patch.  (GitHub #33)
v6.0.0 - 2024-07-10
 - Drop support for Perl 5.10.  podlators now requires Perl 5.12 or later.
 - podlators now uses semantic versioning for the package and module
   versions, with a v prefix to work with Perl's packaging system.
 - Pod::Man now translates all "-" characters in the input into *roff "\-"
   escapes (normally rendered as an ASCII hyphen-minus, U+002D) rather
   than using fragile heuristics to decide which characters represent true
   hyphens and which represent ASCII hyphen-minus.  The previous
   heuristics misrendered command names such as apt-get, causing search
   and cut-and-paste issues.  This change may cause line-break issues with
   long hyphenated phrases.  In cases where the intent is a true hyphen,
   consider using UTF-8 as the POD character set (declared with =encoding)
   and using true Unicode hyphens instead of the ASCII "-" character.
 - Pod::Man now disables the special *roff interpretation of "`" and "'"
   characters as paired quotes everywhere, not just in verbatim text, thus
   forcing them to be interpreted as the regular ASCII characters.  This
   also disables the use of "``" and "''" for paired double-quotes.  The
   rationale is similar to that for hyphens: there is no way to tell from
   the POD source that the special interpretation as quotes is intended.
   To produce paired typographic quotes in the output, use UTF-8 and
   Unicode paired quote characters.
 - Man page references in L<> that are detected as such by Pod::Simple are
   now always formatted as man page references even if our normal
   heuristic would not detect them.  This fixes the formatting of
   constructions such as @@RXVT_NAME@@Perl(3), which are used by packages
   that format a man page with POD and then substitute variables into it
   at build time.  Thanks to Marco Sirabella for the analysis and an
   initial patch.  (GitHub #21)
 - Add a workaround to Pod::Man to force persistent ragged-right
   justification under nroff with groff 1.23.0.  Thanks to Guillem Jover
   for the report and G. Branden Robinson for the analysis.  (GitHub #23)
 - Fix wrapping of text with S<> markup in all subclasses of Pod::Text.
   Thanks to Jim Avera for the report.  (GitHub #24)
 - Pod::Man now forces a blank line after a nested list contains only
   =item tags without bodies.  In previous versions, the blank line before
   the next item in the surrounding =over block was not included.  Thanks
   to Julien ÉLIE for the report.  (GitHub #26)
 - Import PerlIO before checking for layers so that PerlIO::F_UTF8 is
   available, which fixes double-encoding of output when a :utf8 layer is
   in place and PerlIO is not imported.  Thanks to youpong for the bug
   report, James Keenan for the elaboration, and Graham Knop for the fix.
   (GitHub #25)
 - pod2text --help now exits with status 0, not 1, matching normal UNIX
   command behavior and the behavior of pod2man.  (GitHub #19)
 - Fix tests when NO_COLOR is set in the environment.  (GitHub #20)

0.51     2025-06-19

- Made it possible to force Specio to only use pure Perl dependencies by setting the
  SPECIO_IMPLEMENTATION environment variable to "PP". Requested by @arodland (Andrew Rodland). GH
  #23.

0.52     2025-08-09

- Really made it possible to force Specio to use only pure Perl dependencies. The changes in 0.51
  were incomplete, so lots of places still loaded the Clone (XS) module. Fixed by @arodland (Andrew
  Rodland). GH #23.