~ubuntu-branches/ubuntu/vivid/pike7.8/vivid

Changes since Pike 7.8.352 (third 7.8 release):
----------------------------------------------------------------------

Extensions and New Functions
----------------------------

o Added module ZXID

  This module provides an interface to the ZXID Identity Management
  toolkit, which implements (among other tasks) the SAML 2.0, Liberty
  ID-WSF 2.0, and XACML 2.0 protocols.

o Added module Search.

  This module implements a free text search database. It needs some
  indexed storage, and there is a storage backend for use with MySQL.
  There is support for indexers with filters for different document
  types. The query support includes metadata and date range searches.

  It is the core of the search functionality that is used in the
  commercial Roxen products, so it has been in production use for
  quite some time already.

o Typed query result support in Sql.Sql and in the MySQL module.

  Traditionally the SQL glue has returned all types of data as
  formatted strings with their display representations. Now there are
  new so-called "typed" variants of the query functions (typed_query,
  big_typed_query, and streaming_typed_query in the Sql.Sql class)
  that instead return the data as the appropriate pike native types,
  e.g. pike integers for SQL integers, pike floats for SQL floats,
  etc.

  This requires support by the DB server specific backends, which is
  currently limited to the MySQL glue.

o has_prefix now works on objects.

o Standards.JSON

  New module for encoding and decoding JSON data. Implemented in C.

o Val

  New module to contain some global special objects that are shared
  between other modules. There is e.g. a Val.null object which is used
  both by the Sql module to represent SQL NULLs in typed mode, and by
  Standards.JSON to represent the null value.

o Protocols.HTTP.Session will no longer override an existing
  content-type header in async_do_method_url.

o Protocols.HTTP

  Allow directly specifying the body of a POST request (needed for the Google
  Apps API).

  A more complete list of response codes and descriptions is included and used.

  Query includes proxy support for GET (http/https) and POST (https) in both
  sync and async mode.

  Query exposes close() so you do not have to wait for garbage collection.

  Server adds Basic Range support.

o Regexp.PCRE.replace() adds support for capture groups.

o Updated timezone data to tzdata2010o.

o String.normalize_space

  New and efficient C implementation; sanitises whitespace, even in wide
  strings.

o String.trim_all_whites

  Extended to the entire range of Unicode white spaces. (And is faster.)

o Tools.PV and pike -x pv now work with GTK2 as well as GTK.

o Integer constants exported by modules can now be used in cpp expressions.

o Modules/programs that have the constant "dont_dump_program" will not
  be dumped on installation. This is used to prevent dumping of
  programs that check for external dependencies at compile time such
  as the MySQL module.

o Added Thread.Queue.peek_array.

o Fixed stack overrun bug in Nettle.CBC and Nettle.Proxy. This is not
  believed to have been a security issue.

o System

  Added getloadavg() on platforms that support the getloadavg() call.

  Added cross-platform get_home() and get_user() wrappers.
  
  Added GetUserName() for Windows systems.

o SSL

  Added support for AES to SSL3, and fixed bug that prevented some
  handshake failure alerts to be sent to the client.

  Implemented RFC 5746 "TLS Renegotiation Indication Extension".

  Support parsing of Hello Extensions (RFC 5246 7.4.1.4).

o Stdio

  New methods in Stdio.File:

    set_callbacks() and query_callbacks() for handling several callbacks at
    once.

    send_fd() and receive_fd() allow file descriptors to be sent and received
    over UNIX-domain sockets.

  The type for Stdio.Fd.create() is now as documented.

  Stdio.Terminfo contains a new class, MetaTerminfoDB, that merges several
  terminfo directories.

  Stdio.File()->query_address() now knows about IPv6-mapped IPv4 addresses.

  Added some extra magic to Stdio.File()->connect() when mixing IPv4 and IPv6
  addresses.

o ADT.Heap->values and ->num_values can now be overloaded.

o Added support for binding a Protocols.DNS server to multiple port/interface
  combinations.

o Support for asynchronous DNS query replies in Protocols.DNS.

o Calendar

    dwim_time() should know how to parse ISO 8601 timestamps.

    Added Calendar.Greek.

o Process
  
  Process.spawn(), popen() and system() et al now use Process.Process.

  Process.Process adds support for spawning processes via forkd.

  Process.spawn_pike() can now optionally pass predefines as well as 
  program and include path arguments.

  Process.spawn_pike() should also be more reliable on Windows.
 
o Added functions to access the numerator and the denominator of a Gmp.mpq.

o Added lower level functions Tools.X509.[dr]sa_sign_key(). This makes it
  easier to create hierarchial certificates.
  make_selfsigned_[dr]sa_certificate() now use the corresponding
  [dr]sa_sign_key().

o Added SSL3.sslfile.{set,query}_callbacks() as per Stdio.File.

o Added the !-modifier for sscanf().

o Added Int.reflect().

o Image.Dims adds support for Photoshop files.

o Added Gmp.encode_json().

o Added Thread.Queue.peek_array.

o Add System.getloadavg() on platforms that support the getloadavg() call.

o Pike.Backend has reinstated call_out memory usage, and a new function,
  Pike.DefaultBackend.get_stats() for viewing this data.

o Locale.Charset now supports encoding and decoding UTF32.

o Added Sql.Sql()->is_open() and Sql.Sql()->ping(). Note that
  your chosen database module must support these (Mysql and pgsql
  currently do).

o MIME.ext_to_type() now knows about most all popular mime types (including 
  ~700 new entries).

Optimizations
-------------

o MIME should now use 1/3 less memory when decoding big messages.

o Nettle.hash

  Avoid allowing threads during hashing of small strings (current threshold
  is 1 MB). This change improves ApacheBench numbers ~20% for a Roxen
  benchmark on a quad-core Core i5.

o String.trim_all_whites is 10% faster and includes all Unicode.

o Improved performance of Process.run when threads aren't supported.

o Locale.Charset improves performance of encoders when replacement is active
  by a few magnitudes.

o pike -x pmar_install can install from an http/s url, supports gzipped
  archives, md5 hash checking and better uninstall support on Windows.

o Monger

  Monger (pike -x monger) can now install pmar (pre-built module archives) 
  from the module repository, when available (particularly useful on Windows).

  Monger can also download and build from a module's source repository (svn, 
  hg, git) when specified in the Pike module repository (modules.gotpike.org).

Deprecations
------------

o The libpq based old Postgres driver is now deprecated. The new pgsql
  driver is faster, better tested, more stable, maintained, has more
  features, does not depend on external libraries and is less filling.

o Stdio.File: << is deprecated.


Incompatible changes
--------------------

o Signal handlers in GTK2 do no longer get their arguments in the
  form of an array. The prototype for a signal handler which was
  previously e.g.

    int query_tooltip(GTK2.Widget w, array(mixed) params, mixed callback_arg)

  now becomes

    int query_tooltip(GTK2.Widget w, int x, int y, int keyboard_tooltip,
		      GTK2.GObject tooltip, mixed callback_arg)


Bug fixes
---------

o Architecture/OS specific:

  - Fixed segfault in combine_path_nt on windows when the first char
    of an appended path is wide.

  - Stdio.cp should now work on directory trees on Windows.

  - Fixed bug that caused file_stat on Windows NT to not return a result 
    on certain non-FAT filesystems. Fixes bug 6432.

  - Fixed a root directory check in Stdio on Windows NT. Fixes bug 3759.

o Calendar module:

  - Fixed backward compatibility for Calendar.iso_name() and
    Calendar.iso_short_name().

  - Fix evaluation order for Calendar.YMD.`+().

  - Fixed problem with Calendar.month_from_yday() when applied on week
    53. [bug 5303]

o Compiler:

  - Fixed issue where a stale pointer into the identifier table could
    sometimes be dereferenced when defining getters and setters.

  - Fixed NULL-deref for compilation errors in range expressions.

  - Fixed bug that prevented the literal zero from being considered a constant.

  - Fixed fatal error in the constant expression evaluator of the
    compiler for certain code.

o Documentation:

  - Added Autodoc support for new-style getter and setter syntax.

  - Fixed support for documenting arguments to implicit create().

o Graphics:

  - Fixed leak in the PNG decoder for images with tRNS chunk.

  - Fixed several issues related to lables on the X and Y axises
    and the calculation of xmaxvalue for the bars/line/linear case
    in Graphics.Graph.

  - Fix integer underflow in Image.JPEG. Fixes [bug 6413].

o I/O:

  - Support undocumented OS behaviour on out-of-band data returning
    EOPNOTSUPP and ECONNRESET when a stream is half closed.
    The undocumented behaviour is found on (at least) 2.6.x Linux
    and FreeBSD 7.x and 8.x.

  - Stdio.{write,append}_file() now throw errors in ALL failure modes.

  - Stdio.File()->query_address() should return a correct errno when given
    a socket with an invalid protocol family.

  - Stdio.File()->file_peek() now refrains from releasing the interpreter
    lock if there's no timeout.

o Protocol modules:

  - Fixed backtraces when Protocols.HTTP.Query objects where garbage
    collected. [bug 5730]

  - Multiple multipart file submissions in Protocols.HTTP.Server will be
    preserved as suggested by HTML5.

  - Protocols.HTTP.Query fixes broken state when open_socket() has thrown
    an error.

  - Tools.Standalone.httpserver has fixed entity encoding. Makes files
    with spaces in them accessible.

  - Protocols.HTTP Sets the content-length header for zero-length data.
    Fixes [bug 5936].

  - Protocols.HTTP.Query uses content-length instead of content_length.
    The latter will be corrected in the encode method, but is problematic
    when that function is overridden.

  - Protocols.HTTP.Query filters weak SSL-ciphers rather than explicitly
    listing strong ones.

  - Protocols.HTTP includes some close, state reset and keep-alive handling
    fixes.

  - Fixed so than an existing content-type is not overriden in
    async_do_method_url.

o Runtime:

  - Use nanosleep() when available instead of poll or select hacks for
    the Pike-level sleep() function.

  - Fixed a segfault occuring if exit() is used from a thread when
    the Pike runtime has been compiled with cleanup-on-exit enabled.

  - Fixed bug in new_check_call(). Fixes [bug 6442].

  - Prevent short disablements of the gc from affecting gc intervals.


o sprintf:

  - Fix a buffer overrun when formatting Gmp.mpz as floats if the precision
    exceeds the number of digits.

  - Ensure that decimal points are used when floats are formatted
    in casts etc. Previously the float formatting in casts and in
    sprintf("%O",...) were susceptible to locale settings and could
    therefore use something other than a '.' as decimal separator.
    Note that sprintf("%f",...) is still locale sensitive.

o Sql database modules:

  - Fixed memory leak on out of memory in Mysql.fetch_json_result().

  - Fixed locking problem in Mysql.

  - Fixed dangling cache reference in prepared statements when using CREATE
    statements via Sql.pgsql.

  - Fixed broken (auto-re-) connect logic in Sql.pgsql.

  - Support all text queries in Sql.pgsql.

  - Fixed rare string corruption in Sql.pgsql.

o SSL/TLS:

  - Fixed bugs in handling of abrupt SSL3 remote close. sslfile now also
    triggers EPIPE errors if the user attempts to read or write to a
    connection that has been abruptly closed (to detect truncation
    attacks better).

  - Fixed issue where SSL3 async_client objects became garbage even
    after close().

  - Tools.X509 uses a unique serial number for each selfsigned certificate.
    Firefox otherwise complains about serial number clashes.

o Misc:

  - System.normalize_path no long fails for files with 8-bit chars in
    their names.

  - Improved widestring support for Parser.Tabular.

  - Fixed a problem in "pike -x module" that prevented auto tools from
    running in the correct directory.

  - Fixed issue where GTK objects weren't destroyed properly.
    http://pike.ida.liu.se/docs/tutorial/hello/window.xml now
    works again.

  - Fixed encoding/decoding bug in "pike -x httpserver" that prevented
    files with space in the name from being accessable.

  - Parser.XML.Tree contains fixes for add_child_before() and add_child_after().

  - Parser.Tabular fixes support for Stdio.File objects. For example:

    Pike v7.8 release 468 running Hilfe v3.5 (Incremental Pike Frontend)
    > Parser.Tabular(Stdio.File("data.csv"), Stdio.File("format.csv"));

  - Fixed return value from Regexp.PCRE.exec to always have entries for all
    submatches.

  - Fuse has fixed support for readlink().

  - Fixed infinite loop in Audio.Format when reading MP3 frames.

  - Added path to pike binary in pike -x module used by verify targets. 
    Should also fix problems installing modules using pike -x monger.

  - Reduce optimization in gethostby{addr,name} that could cause 
    overzealous caching or results.

Building, installing and developer tools
----------------------------------------

o Fixes to allow Regexp.PCRE to compile against recent versions of PCRE lib.

o Fix Gz module build on recent Gentoo Linux.

o Fixed some library searchpath issues.

o Some potential fixes for --enable-pikelib mode on platforms had
  problems with the final linking.

o Some issues with the support for building modules needing
  a C++ compiler have been fixed.

o Made it possible to override $(OS) with an environment variable
  $PIKE_BUILD_OS.

o Check for configure scripts built with pre-propagated_variables pike.

o Added Nettle/config.guess that knows about MaxOS X and x86_64.

o Fix broken detection of libjpeg lossless transformation support.

o Installer will no longer dump sql driver files that depend on
  system libraries.

o Fixed support for --without-threads.

o Fix to make dump.pike dump itself just like any other module class.

o Improved support for LIBPIKE.

o Added PIKE_CORE define to detect when not compiling a module.

o Improved support for autoconf 2.6x.

o Ensure that --with-rtldebug enables assertions.

o Configure script now attempts to set up the PKG_CONFIG_PATH.

o Updated the ABI selector for MacOS X. Default to attempting the compiler
  default ABI (instead of forcing 32-bit ABI).

o CMOD Enhancements

    Tools.Standalone.precompile supports ranges for strings and negative
    int(low..high) bounds.

    Fixed bug in generate_overload_func_for() when there are alternatives
    accepting zero arguments.

o Tools.Testsuite:

    Added some functions to make pretty logging in tests easier.

    Tools.Standalone.test_pike passes the verbosity level in an environment
    variable TEST_VERBOSITY to subtests.

    Tools.Testsuite accumulates multiple result reports.

Changes since Pike 7.8.316 (second 7.8 release):
----------------------------------------------------------------------

General
-------

o ADT.Heap
  The compat functions top() and size() are now marked as deprecated.


Extensions and New Functions
----------------------------

o Image.Image

  Added support for sending Color objects to the color() function.

o Sql.Sql

  Added support for fetching result sets as JSON-encoded strings.

o Calendar

  Updated timezone data to 2009j.

o Parser.XML.Tree

  Output from render_xml() is now on canonical form.

o Added "pike -x httpserver" that implements a minimal HTTP-server
  exporting the current directory.


Optimizations
-------------

o Lowered startup time by extending encode_value() and MasterCodec so
  that master.pike can be precompiled.

o Lowered startup time by reducing the number of MEMSET() calls done
  on startup.

o Allow threads while calling SDL_GL_SwapBuffers.

o Added optimization of has_value() and search() for single-character strings.

o Added optimization when combining two arrays where the second array
  contains a single item.


Bug fixes
---------

o Fixed casting of floats to strings to again work more like in Pike
  7.6 and earlier: The string will always contain either a "." or an
  "e" to signify that it is a float, and it will no longer display so
  many digits that the binary/decimal conversion errors become
  visible.

o Protocols.HTTP.Server.HeaderParser will now discard parts of a
  header value if the header value contains a newline (like "Host:
  google.com\n.hax.net\r\n") and headers without colon.

o file_open_socket() now attempts to set SO_REUSEPORT on the socket.
  Potentially fixes issues on FreeBSD 7.x where ports aren't reused.

o Calendar.TimeRanges now knows that `+() and `-() can get any number
  of arguments.

o The experimental extended ZFS support is disabled since it pulled in
  unwanted dependencies.

o Fixed propagation of changed module directories for joinnodes by
  zapping the joinnode cache.

o Fixed memory leak in Image.WBF.

o Fixed potential NULL-deref in Image.XWD.

o Fixed potential NULL-deref in Image.PNG.

Building and installing
-----------------------

o Nettle 2.0 is now supported.


Changes since Pike 7.8.116 (initial 7.8 release):
----------------------------------------------------------------------

General
-------

o Sql.pgsql:
  - Substantial documentation improvements.

o Stdio.Fd_ref is deprecated and in the process of being eliminated.
  You have never used it before, so don't start using it now.


Bugfixes
--------

o Fixed bug in AGGR_ARR_EPILOGUE() (aka END_AGGREGATE_ARRAY()) that
  caused the empty array to always be returned if there were no
  unaggregated elements on the stack at END_AGGREGATE_ARRAY.

  You are encouraged to upgrade just for this fix since the bug
  creates very rare but (seemingly) random occurances of empty arrays.

o Fixed multiset_delete_node to work with destructed indices. This
  fixes occasional bugs in various functions when there are destructed
  objects in indices.

  You are encouraged to upgrade just for this fix since the bug creates
  (seemingly) random crashes that are hard to reproduce and track down.

o Fixed overzealous filtering of the value returned from Stdio.File()->mode().
  This bug caused Stdio.nb_sendfile to always run in blocking mode.

  You are encouraged to upgrade just for this fix since the bug can
  cause supposedly nonblocking actions to block and hang the script.

o Fixed bug in the Odbc module on WIN32 where it would fail for some
  field-types with the error "22003:Numeric value out of range".

o Sql.pgsql:
  - Various small fixes and performance improvements.
  - Fixed fully transparent UTF8/Unicode/wide string/character set support.
  - Fixed formatting errors in error().

o Fixed bug when seeking backwards in Gz.File.

o Fixed bug in fd_stat on windows when used on a file system with a
  long name.

o Fixed off-by-one bug in Stdio.Readline that caused a space in the
  last column to disappear instead of being converted to a linefeed
  when word wrapping is used.

o Fix bug in FakeFile.read()
  Replaced the broken-in-multiple-ways implementation by something that
  actually works (hence the original enlightnening description: "fixed a bug",
  there is really no concise way to classify this level of brokenness).

o Image.TIFF: Fixed memory leak on OOM.

o Fixed crash bug in yyexplain_not_implements().

o LysKOM.ProtocolTypes: Bugfix of LocalToGlobalBlock

o Protected Pike_fatal() against the risk of hanging on mutexes.

o Added 7.6 compat versions for RegGet{Value{,s},KeyNames}().

o Fixed memory leak in Image.Colortable.

o Parser.SGML: Reinstated the class as Parser.SGML(). It inadvertedly
  became Parser.SGML()->SGML() in the move to a separate file.
  Parser.SGML.SGML is kept as a compatibility alias.

o Added Pike 7.6 and earlier compat for store_constant(UNDEFINED).

o Fixed race-condition causing "Lost track of child" on pikes compiled
  --without-threads.

o Regexp.PCRE: Do not match the empty string past the end of the input.

o Reduced the risk of throwing out-of-stack errors while formatting an
  out-of-stack backtrace.

o Fixed bug that would cause sprintf's %nH to refuse all strings as too
  long for some values of "n" on some architectures.

o Made Process.search_path() aware if Win32 and AmigaOS specifics.

o Fixed bug in Protocols.HTTP.Query that dropped the last byte of a
  request if the server did not provide a Content-Length header.

o Fixed bug in Protocols.HTTP.Query that dropped the first argument to
  the async_fetch() callbacks (ie the query object) in some circumstances.

o Fixed bug in Protocols.HTTP.AsyncClient that made all data decoding fail.

o Fixed fatal "Invalid formatting method: 'p' 0x70." in COM.

o low_get_function_line() now knows about prototypes.
  Fixes [bug 4855] ("Bad offset" when generating backtrace).

o Removed spurious call to the fail callback after a successful
  Protocols.HTTP.Query()->timed_async_fetch(). Fixes [bug 4773].

o Fixed issue where symbols could be hidden by a later inherit of the
  same, but protected, symbol.

o Fixed NULL dereference on failure to resolve the function return
  type when functions are used as types.


Optimizations
-------------

o Fixed Filesystem._Tar.ReadFile to not gobble the whole tar entries
  into memory at once.

o Improve performance of get_cpu_time() on OS X by avoiding a kernel
  trap to get the ID of the underlying system thread.

o Throttle calls to task_info() on OS X since they are quite
  expensive. This gives a significant performance boost (>10%) on
  CPU-intensive applications.


Extensions and New Functions
----------------------------

o Added support for specifying modifiers via a mapping to '*' in sprintf().

  See the documentation for details.

  This support is detectable by checking for
  String.__HAVE_SPRINTF_STAR_MAPPING__

o Added support for generating little-endian floats with %F in sprintf().

  This support is detectable by checking for
  String.__HAVE_SPRINTF_NEGATIVE_F__

o Sql.Sql now supports NULL values for %s-parameters.

o Timeout support in win32 threads.

o Made FakeFile() support getchar/gets/unread.

o Added option to chown and utime that makes it possible to change the
  symlink instead of the target of the link.

o function_name() now knows how to adjust the names of named lambdas.

o Changed handling of POLLHUP for poll devices, it is now signalled on
  all of the callbacks.

o Updated timezone data to tzdata2009c.

o Added some support for the TIMESTAMP and INTERVAL datatypes to
  Oracle module.


New modules / classes / methods added
-------------------------------------

o Protocols.IPv6

  Helpers for IPv6 related things. This currently only contains tools
  to parse and normalize addresses. (Not much more is necessary since
  the usual I/O interfaces support IPv6 as-is.)

o Added Filesystem._Tar.extract to be able to extract many/all files
  in a reasonably efficient way.

o Added Parser.Tabular
  It provides a flexible yet concise record-description language to parse
  character/column/delimiter-organised records.

o Calendar.YMD.http_time() that knows how to handle dates according to
  HTTP 1.1 (RFC 2616).

o Added support for GTK2 get_window() and get_pixmap().


Building and installing
-----------------------

o Use the #-character as prefix and suffix for placeholders in
  master.pike.in instead of the currency sign, to avoid 8-bit character
  encoding issues.

o Improved support for version bumping via both Subversion and git.

o A recent stable Pike (7.8.116+) is now required to do a "make
  export".

o The win32 installer now removes the generated master.pike on uninstall.



Changes since Pike 7.6:
----------------------------------------------------------------------

This is a high level list of changes between Pike 7.6 and Pike 7.8.
General bug fixes, build fixes and optimizations are not mentioned
here. For a complete list of changes, please consult the CVS changelog
either directly or through Code Librarian.


New / improved language functionality
-------------------------------------

o New syntax to index from the end in range operations.

  A "<" can be added before an index in the [..] operator to count
  from the end instead, beginning with 0 for the last element. This is
  convenient to e.g. chop off the last element in an array: a[..<1].

o New `[..] operator function.

  Range operations have been separated from the `[] operator function
  and are now handled by the new `[..] which provides greater control
  for how the range bounds are specified. For compatibility, if there
  is no `[..] then `[] is still called for range operations.

  The `[..] callback will get four arguments, start value, start type,
  end value and end type. The type arguments is any of
  Pike.INDEX_FROM_BEG, Pike.INDEX_FROM_END or Pike.OPEN_BOUND. Here
  are a few examples of what input arguments different calls would
  generate.

    [..]      0   OPEN_BOUND       0   OPEN_BOUND
    [1..]     1   INDEX_FROM_BEG   0   OPEN_BOUND
    [..2]     0   OPEN_BOUND       2   INDEX_FROM_BEG
    [3..4]    3   INDEX_FROM_BEG   4   INDEX_FROM_BEG
    [<5..]    5   INDEX_FROM_END   0   OPEN_BOUND
    [6..<7]   6   INDEX_FROM_BEG   7   INDEX_FROM_END

o Generalized this and this_program.

  It is now possible to refer to inherits in objects. Example:

    class A {
      int a;
      void foo() { werror("A\n"); }
    }
    class B {
      inherit A;
      int b;
      void foo() { werror("B\n"); }
      A::this_program get_a() { return A::this; }
    }

  In the above B()->get_a() will return an object with two symbols,
  'a' and 'foo', but B()->get_a()->foo() will still write "B\n".

o Added support for getters and setters.

  It is now possible to simulate variables with functions. Example:

    class A {
      private int a;
      int `b() { return a; }		// Getter for the symbol b.
      void `b=(int c) { a = c; }	// Setter for the symbol b.
      int c()
      {
        return b;			// Calls `b().
      }
    }

    object a = A();
    a->b = 17;				// Calls `b=(17).
    werror("%d\n", a->b);		// Calls `b().

o Casting to derived types.

  It is now possible to call a value of type type to perform the
  corresponding value cast. eg:

    typedef string MyString;

    return MyString(17);		// Casts 17 to "17".

o Unicode escapes.

  Pike now understands the common way to escape unicode chars, using
  \uxxxx and \Uxxxxxxxx escapes. These escapes works both in string
  and character literals and in the preprocessor. The latter means
  that unicode escapes can be used in identifiers, which is useful if
  they contain characters that can't be represented raw in the source
  code charset.

o Stricter type checker for function calls.

  The type checker for function calls is now based on the concept of
  currification. This should provide for error messages that are more
  easily understood. It also is much better att typechecking function
  calls utilizing the splice (@) operator. The mechanisms used by the
  typechecker are also made available as Pike.get_first_arg_type(),
  Pike.low_check_call() and Pike.get_return_type().

o Stricter typing of strings.

  The string type may now have an optional value range.

    string(0..255) bytes;

o Support for having multiple different active backend implementations.

  In Pike 7.6 and earlier, there could only be one active backend
  implementation at a time, which was the one selected for best
  performance with lots of files (cf Pike.Backend below). This led
  to problems when Pike attempted to run poll-device based backends
  on older versions of operating systems where poll devices aren't
  available, and with extra overhead when the backend was only used
  with a very small set of files.

  Basic backend implementations:

  - Pike.PollDeviceBackend
    This is a backend that is implemented based on keeping the state in
    the operating system kernel and thus eliminating the need to send the
    state in every system call. This is however not available on all
    operating systems. Currently supported are /dev/poll (Solaris, etc),
    epoll(2) (Linux) and kqueue(2) (FreeBSD, MacOS X).

  - Pike.PollBackend
    This is a backend that is implemented based on the poll(2) system call.
    This is typically available on all System V or Linux based systems.

  - Pike.SelectBackend
    This is a backend that is implmented based on the BSD select(2) system
    call. This backend is available on all operating systems supported by
    Pike.

  Derived backend implementations:

  - Pike.Backend
    This is the backend selected among the basic backend implementations,
    which is likely to have the best performance when there are lots of
    files in the backend.

  - Pike.SmallBackend
    This is the backend selected among the basic backend implementations,
    which is likely to have the best performance when there are very few
    files in the backend.

  Note that there is also a default backend object:

  - Pike.DefaultBackend
    This is the Pike.Backend object that is used for file callbacks and
    call outs if no other backend object has been specified.

o cpp

  The preprocessor now supports macro expansion in the #include and #string
  directives.

    #include USER_SETTINGS

o Destruct reason passed to lfun::destroy.

  lfun::destroy now receives an integer flag that tells why the object
  is being destructed, e.g. if it was an explicit call to destroy(),
  running out of references, or reaped by the garbage collector.

  These integers are defined in the new Object module as
  DESTRUCT_EXPLICIT, DESTRUCT_NO_REFS, DESTRUCT_GC and
  DESTRUCT_CLEANUP.

o Improved support for mixin.

  The Pike compiler now supports mixin for symbols that have been
  declared protected. Mixin is the concept of overloading symbols
  via multiple inheritance. In previous versions of Pike the mixin
  paradigm was only supported for public symbols. For more information
  about mixin see eg http://en.wikipedia.org/wiki/Mixin .

o Implicit and explicit create().

  The compiler now supports defining classes with both an implicit
  and an explicit create().

    class A (int i)
    {
      protected string j;
      protected void create(string j)
      {
        A::j = reverse(j);
      }
    }

o Warnings for unused private symbols.

  The compiler now checks that all symbols that have been declared
  private actually are used.

o Warnings for unused local variables.

  The compiler now checks that all local variables are used.

o Unicode

  Case information and the Unicode module are updated to Unicode
  5.1.0.

o The keyword protected

  The modifier protected is now an alias for the modifier static.
  NOTE: In the next release static will be deprecated.

o extern declared variables

  Variables can now be declared as 'extern'. This means that
  inheriting classes must have them. They can be used like normal
  variables in the base class.

  Example:
    class A
    {
	extern int a;
	int dummy() { return ++a; }
    }

o __attribute__ and __deprecated__

  It's now possible to set custom attributes on types, so that
  it is possible to make custom type checking. This is currently
  used to improve the argument checking for sprintf() and related
  functions, and for marking symbols as deprecated.
  eg:

    __deprecated__ mixed obsolete_function();
    __deprecated__(mixed) obsolete_return_value();
    mixed fun(__deprecated__(mixed) obsolete_arg);
    __deprecated__(mixed) obsolete_variable;

  The deprecated type flag using __deprecated__ is a convenience
  syntax to use instead of e.g.

    void f(void|__attribute__("deprecated",int) timeout)

  Other uses of __attribute__ in type declarations can be seen in e.g.
  the type for werror():

    > typeof(werror);
    (1) Result: scope(0,function(string : int) |
        function(__attribute__("sprintf_format", string),
                 __attribute__("sprintf_args", mixed) ... : int) |
        function(array(string), mixed ... : int))

o __func__

  The symbol __func__ now evaluates to the name of the current
  function. Note that this name currently can differ from the
  declared name in case of local functions (i.e. lambdas). Note
  also that __func__ behaves like a literal string, so implicit
  string concatenation is supported. eg:

    error("Error in " __func__ ".\n");

o __DIR__

  __DIR__ is a new preprocessor symbol that resolves to the directory
  that the current file is placed in. Similar to how __FILE__ points
  out the file the code is placed in.

o #pragma {no_,}deprecation_warnings

  Warnings for use of deprecated symbols can be turned off for a
  segment of code with

    #pragma no_deprecation_warnings

  and turned on again with

    #pragma deprecation_warnings

o Compatibility name spaces

  Older versions of a function can be reached through its version name
  space. For example the 7.4 version of the hash function can be
  called through 7.4::hash().

o Iterator API

  The iterator API method Iterator->next() is no longer optional.

o IPv6

  Support for IPv6 has been improved.


Extensions and New Functions
----------------------------

o exit()

  Exit now takes optional arguments to act as a werror in addition to
  exiting the process.

    exit(1, "Error while opening file %s: %s\n", path, strerror(errno()));

o getenv()/putenv()

  getenv() and putenv() are now accessing and modifying the real
  environment.

o get_dir()

  Calling get_dir() with no arguments will now return the directory
  listing for the current directory.

o undefinedp()/destructedp()

  undefinedp() and destructedp() have been added as more readable
  alternatives to zero_type().

o limit()

  The new toplevel function limit(a, x, b) is a convenience function
  that works like min(max(a,x),b).

o listxattr(), getxattr(), setxattr(), removexattr()

  The new xattr functions listxattr(), getxattr(), setxattr and
  removexattr() allows extended file attributes to be modified from
  within Pike.

o sprintf() and sscanf()

  - sprintf() now attempts to lookup the name of the program when
    formatting types of objects and programs.

  - The new formatting directive %H can be used to format a string as
    a binary hollerith string.

      > sprintf("%2H", "Hello");
      (1) Result: "\0\5Hello"

  - The new formatting directive %q can be used to format a atring as
    a quoted string, quoting all control character and non-8-bit
    characters as well as quote characters. Like %O but always on one
    line.

      > sprintf("%q", "abc \x00 \" \' \12345");
      (1) Result: "\"abc \\0 \\\" ' \\u14e5\""

  - Ranges in sscanf sets can now start with ^, even at the beginning
    of the set specifier. Example: %[^-^] matches a set with only ^ in
    it. To negate a set with - in it, the - character must now be the
    last character in the set specifier.

o encode/decode value and programs

   --with-portable-bytecode is now the default. Pike programs that have
   been dumped on one architecture now can be decoded on another.

o gethrtime, gethrvtime, gauge

  Added support for POSIX style timers using clock_gettime(3). Notably
  this fixes nice high resolution thread local cpu time and monotonic
  real time on reasonably modern Linux systems.

  There are new constants CPU_TIME_* and REAL_TIME_* in the System
  module to allow pike code to query various properties of the CPU and
  real time clocks in use.

o ADT.BitBuffer

  Added read() method that reads whole bytes from the buffer and
  returns as a string.

o ADT.Queue

  It is now possible to use sizeof() and values() on a Queue object to
  get the size of the queue and all elements in the queue as an array.

o ADT.Stack

  Stack objects can now be cast to arrays to get all elements on the
  stack.

o ADT.Struct

  - New Item class SByte, SWord and SLong represents a signed byte,
    word and longword respectively.

  - The Item classes int8, uint8, int16, uint16, int32, uint32, int64
    and uint64 are aliases for already existing Item classes.

  - sizeof() now works for empty Struct objects.

  - sizeof() can be used on individual Item objects.

  - Struct objects can now be used as Items in other Structs.

    class Pair
    {
      inherit ADT.Struct;
      Item id = uint8();
      Item val = uint64();
    }

    class File
    {
      inherit ADT.Struct;
      Item magic = Chars(4);
      Item one = Pair();
      Item two = Pair();
    }

o Array

  - New function combinations() returns all combinations of a
    specified length with elements from a given array.

  - Added push(), pop(), shift() and unshift() functions for Perl
    weirdos.

o Calendar

  - Added new calendar Badi, used in the Baha'i religion.

  - Fixed bugs in discordian, where the year() was off by 1166 and
    now() was off 5 month per year.

  - Time objects now handle * and / with floats. A split() method has
    been added for more advanced splitting where the preferred time
    quanta can be given as an argument.

  - A new formatting method format_ext_time_short() has been added to
    Time objects.

  - Timezones are now read from /etc/localtime, if available.

  - Cleaned up year-day (yd) handling so that it never goes below 1.
    Instead be careful to use either year (y) or week-year (wy)
    depending on context.

  - Fixed inconsistent range handling in year() and years() that made
    them almost but not quite zero-based. Now they are one-based just
    like day()/days(), month()/months() etc.

  - Cleaned up conversion between weeks and years: E.g. if a week has
    days in two years then converting it to years will produce a range
    of both years. The attempt to always map a week to a single year
    is gone since it's inconsistent with how other overlapping time
    ranges are handled. If the user wants to convert a week to the
    year it "unambiguously" belongs to, (s)he can do
    Calendar.ISO.Year(week->year_no()).

  - Did away with the attempt to map leap days between February 24th
    and 29th before and after year 2000, since doing so breaks date
    arithmetic.

  - The last four changes above are not entirely compatible.
    Compatibility with 7.6 and older is retained with #pike 7.6.

o CompilerEnvironment & CompilerEnvironment()->PikeCompiler

  The Pike compiler has been refactored to be more object-oriented and
  more transparent. It is now possible to customize the compiler by
  overloading functions in the above two classes. The compiler object
  used by Pike internally is available through the global constant
  DefaultCompilerEnvironment.

o Debug

  The new function count_objects() will return the different kinds of
  objects existing within the Pike process. Useful when trying to
  pinpoint a leak of Pike objects.

o Error

  The new function mkerror() will normalize any thrown value into a
  proper error object (or 0).

o Filesystem

  Traversion has been extended with two new create arguments. The
  first one is a flag to supress errors and the other is a sorting
  function which will be applied to the entries in every directory.
  This callback can also be used to filter the directory entries.

o Float

  The function isnan() can be used to check if a float is Not a
  Number.

    > Float.isnan(Math.inf/Math.inf);
    (1) Result: 1

o Gdbm

  Gdbm databases can be opened in synchronous mode or locking mode by
  adding "s" or "l" respectively in the mode parameter to the create
  method.

o Geography.Position

  It is now possible to encode Position objects with encode_value().

o GLUE

  - The default event callback will now exit both on Exit and Escape
    events.

  - The texture, list and light id generators will now be created
    after we are sure that GL has been initialized, to prevent crashes
    on several systems.

  - BaseTexture and BaseDWIM now supports clamped textures.

o Gmp

  Many small changes to fix Gmp build issues on different platforms.
  The most visible being that the Gmp module now is statically linked
  to work around Mac OS X 10.5 linker issues.

o Gz

  - Added compress() and uncompress() functions as a simpler and more
    efficient but non-streaming interface.

  - Support for RLE And FIXED compression method, if supported by
    zlib. Give Gz.RLE or Gz.FIXED to the strategy argument of
    compress() or the deflate constructor.

  - Added support for configurable compression window size. This is
    the last argument to compress() and the constructor for
    inflate/deflate.

o Image.Colortable

  - The new method greyp() can be used to query if the color in the
    color object is grey.

  - Partial support for serializing color objects. Dithering type and
    lookup mode is not currently saved.

o Image.Dims

  Support for parsing out the dimensions of TIFF files has been added.

o Image.FreeType

  - Added support for handling monochrome (bitmap) fonts.

o Image.Image

  - Image object can now be serialized and deserialized with
    encode_value() and decode_value().

  - It is possible to convert the colors in an image object to and
    from YUV (YCrCb) with the new rgb_to_yuv() and yuv_to_rgb()
    methods.

o Image.Layer

  - It is now possible to get the raw bitmap data of a layer by
    casting the layer object to a string.

o Image.PNG

  - Properly decode cHRM (chrome), sBIT (sbit), gAMA (gamma), pHYs
    (physical), oFFs (offset), tIME (time) chunks.

  - The compression level and strategy when compressing PNG images can
    be controlled by passing "zlevel" and "zstrategy" options to the
    encode() method. Available strategies are filtered, huffman, rle
    and fixed.

      Image.PNG.encode( img, ([ "zlevel":9, "zstrategy":Gz.RLE ]) );

o Image.TIFF

  Added support for little endian TIFF files.

o Int

  Int.inf is an object that can be used as an infinitly large integer.

o Java

  - If possible, Pike now uses libffi instead of creating our own
    trampolines. Adding libffi as a bundle is supported.

  - Added built-in support for SPARC64 trampolines.

  - The method signature fuzzy-matcher now only considers Pike strings
    to match formal parameters of type String, Object, Comparable,
    CharSequence or java.io.Serializable.

  - When the fuzzy-matcher can not decide on a single method
    signature, all candidates are now listed in the error message.

o Locale.Charset

  - The character name normalizer now recognizes Unicode prefixes,
    e.g. "unicode-1-1-utf-7", though the Unicode version itself is
    ignored.

  - Added support for the following character sets

    GB18030/GBK (CP936)
    UTF-EBCDIC
    DIN-31624 (ISO-IR-38)
    ISO-5426:1980 (ISO-IR-53)
    ISO-6438:1996 (ISO-IR-39, ISO-IR-216)
    ISO-6937:2001 (ISO-IR-156)
    GSM 03.38
    Mysql Latin 1

  - Added typed encode and decode error objects,
    Locale.Charset.EncodeError and Locale.Charset.DecodeError, to make
    it possible to catch such errors in a better way.

  - ISO-IR non-spacers are now converted into combiners.

o Math

  - Matrix multiplication was bugged and gave B*A instead of A*B,
    which is now fixed.

  - Matrix objects now have xsize() and ysize() methods to query their
    dimensions.

  - To ease your logarithmic needs log2() and logn() have been added.

o MIME

  - Added remapping variants of the encode words functions with
    encode_word_text_remapped(), encode_word_quoted(),
    encode_words_quoted_remapped() and
    encode_words_quoted_labled_remapped().

  - Added workaround for a bug in Microsoft Internet Explorer where it
    forgets to properly quote backslashes in the Content-Disposition
    field.

  - Fixed a bug that could occur when casting MIME.Message objects to
    strings.

o Mysql

  - Two functions set_charset() and get_charset() have been added to
    set and query the connection charset. The create() function also
    takes a "charset" setting in the options mapping.

  - Improved Unicode support. The MySQL driver now supports
    (possibly wide) Unicode queries and text return values, handling
    the connection charset and all encoding/decoding internally. This
    is enabled by setting the charset to "unicode", or through the new
    functions set_unicode_encode_mode() and set_unicode_decode_mode().
    See their documentation for further details.

o Odbc

  The Odbc module has been updated to support the UnixODBC library, and
  several issues with Unicode and FreeTDS handling have been fixed.

o Oracle

  - The module has been updated to work with Oracle 10.

  - An approximation of the number of rows in a result object can be
    queried from the new function num_rows().

o Parser.HTML

  - Allow string and array as argument to _set_*_callback(). Those
    variants work just like a function only returning the string or
    array.

o Parser.Pike and Parser.C

  - Parser.Pike and Parser.C have been rewritten in C for increased
    performance.

  - The #string directives should be handled correctly now.

o Parser.RCS

  - The RCS parser has been rewritten to be more robust with regards
    to broken RCS data.

o Parser.XML.NSTree

  - Added add_child_before() and add_child_after() methods to the
    NSNode object.

o Parser.XML.Simple

  - The autoconvert() function, used to autodetect the character
    encoding of an XML file and decode it, has been moved from being a
    method of the Parser.XML.Simple object to being a function in the
    Parser.XML module.

o Parser.XML.SloppyDOM

  Yet another DOM-like module aimed at quickly and conveniently parse
  an xml snippet (or whole document) to access its parts. Nodes can be
  selected using a subset of XPath.

  Footnote: This module was previously part of Roxen WebServer.

o Parser.XML.Tree

  The module code has been refactored and a second "simple" interface
  has been added. The (new) SimpleNode interface implements a node tree
  interface similar to the (old) Node interface, with two major
  differences:

  - The SimpleNodes do not have parent pointers, this means that they
    do not generate cyclic data structures (and thus less garbage, and
    no need for zap_tree()), but also that it is no longer possible to
    find the root of a tree from any node in the tree.

  - Some methods in SimpleNode return different values than the
    corresponding method in Node; notably SimpleNode()->add_child(),
    which returns the current node rather than the argument.

  The refactoring also added individual classes for all of the XML node
  types (both for Nodes and for SimpleNodes). This allows for stricter
  typing of code involving XML tree manipulation.

  Several new functions added to manipulate and insert nodes in the
  XML tree.

  The module now also has a much better knowledge of DTDs and DOCTYPEs.

o Parser.XML.Validating

  - get_external_entity() now gets the callback info mapping as
    the third argument rather than a constant zero.

  - SYSTEM URIs are now normalized during parsing if possible.

  - parse() and parse_dtd() now accept getting the base URI for the
    parsed data as the second argument.

  - Improved error reporting.

  - Several issues with the parsing of DTDs have been fixed.

o Postgres

  - Extended the SQL query interface to include affected_rows() and
    streaming_query() as well as variable bindings.

  - Automatic binary or text transfer for queryarguments and resultrows.

  Note: A new native Postgres driver has also been added; see
  Sql.pgsql.

o Pike

  - A new function count_memory() has been added which can calculate
    the memory consumed by arbitrary data structures. Useful when
    implementing memory caches.

  - A new function get_runtime_info() has been added which returns
    information about current ABI, if automatic bignums are enabled,
    what bytecode method is used, the size of native floats and
    integers and the native byte order.

  - The constants INDEX_FROM_BEG, INDEX_FROM_END and OPEN_BOUND has
    been added for use with the `[..] operator API.

  - The functions low_check_call(), get_return_type() and
    get_first_arg_type() allows for inspection of attributes and
    return values of functions.

o Pike.Backend

  Besides the new multiple backend implementations described earlier,
  backends now support two new callbacks: before_callback and
  after_callback are two new variables in the backend objects. They
  can be set to functions that gets called just before and after a
  backend is waiting for events.

o Process

  - The new function spawn_pike() will spawn a Pike process similar to
    the current one, using the same binary file, master and module
    paths.

  - The new function run() is an easy interface that will run a
    process and return a mapping with all the outputs and exit code.

  - Process.popen is now able to run in nonblocking mode. If a second
    argument is provided a file object will be opened with that mode
    and return, enabling two way communication with the new process.

  - The system() function has been extended to be able to pass stdin,
    stdout and stderr arguments to the spawn() call it performs.

o Protocols.DNS

  - Added support for NAPTR (RFC 3403) and SPF (RFC 4408) records.

  - The gethostbyname() function now returns IPv6 addresses too, if
    available.

  - Fixed bugs in IPv6 record parsing.

o Protocols.Bittorrent

  - Support for gzipped and compact tracker responses.

  - Many performance and bug fixes, such as 30% faster hashing of
    files.

o Protocols.HTTP

  - Added support for httpu and httpmu requests.

  - Queries will now throw an exception in case of an errno condition.

  - A new function, do_async_method(), has been added to allow access
    to low level asynchronous HTTP calls.

  - The http_encode_string() function now knows how to encode UCS-2
    characters.

o Protocols.HTTP.Server

  - If accept() fails on the open port, the Port object will continue
    trying, to support servers with high load.

o Protocols.HTTP.Query

  - Added unicode_data() method that will return the payload decoded
    according to the charset described in the Content-Type header.

  - Many fixes for bugs in asynchronous mode.

  - A query will not silently downgrade to http from https anymore if
    there is no crypto support.

  - Fixes for keep alive.

  - IPv6 fixes in DNS handling.

o Protocols.LDAP

  - Enabled support for paged queries.

  - Added more resilience to UTF-8 encode errors.
    Locale.Charset.DecodeError is thrown for UTF-8 decode exceptions.

  - Added a connection pool for connection reuse. It is used through
    get_connection() and return_connection().

  - Added some schema handling and use it to fix the UTF-8 conversion
    to only affect attributes that actually are UTF-8 encoded.

  - Added client.read(), client.read_attr(),
    client.get_root_dse_attr(), client.get_basedn(),
    client.get_scope(), client.get_attr_type_descr(),
    get_constant_name(), and a bunch of constants for various object
    oids, attributes, well-known object guids and other things.

  - Rewrote the search filter parser to handle LDAPv3 extensible
    matches. It now also throw errors on all syntactic errors (using a
    new FilterError object), instead of sending a partial filter to
    the server. It is also possible to compile filters separately
    through make_filter(), and there is a very simple cache for
    compiled filters through get_cached_filter().

  - Added ldap_encode_string(), ldap_decode_string(),
    encode_dn_value(), canonicalize_dn(), parse_ldap_url(),
    client.get_parsed_url(), and client.get_bound_dn().

  - Added client.result.fetch_all().

  - Added new flag field to client.search() to control various aspects
    of how results are returned: SEARCH_LOWER_ATTRS lowercases all
    attribute names. SEARCH_MULTIVAL_ARRAYS_ONLY uses arrays for
    attribute values only for attributes that might return more than
    one value. SEARCH_RETURN_DECODE_ERRORS may be used to avoid
    throwing exceptions on decode errors.

  - Added client.get_protocol_version(),
    client.get_supported_controls(), and the possibility to specify
    controls in client.search().

  - Made the result iterator somewhat more convenient: next() now
    advances one step past the end so the next fetch() returns zero.

  - Added client.error_number(), client.error_string(), and
    client.server_error_string() to make it possible to query errors
    when no result object is returned.

o Protocols.SNMP

  The readmsg() method in the protocol object now takes an optional
  timout argument.

o Protocols.TELNET.Readline

  Several issues where callback functions could be lost have been fixed.

o Protocols.XMLRPC

  The new AsyncClient class implements an asynchronous XMLRPC client.

o Regexp.PCRE.Widestring

  - Replace matches in a string, with support for backreferences, now
    possible from replace_positional().

      > Regexp.PCRE.Plain("my name is ([a-zA-Z]+)")->
      >> replace_positional("hello, my name is john.",
      >> "%[0]s is my name");
      (1) Result: "hello, john is my name."

  - Regexp.PCRE.Widestring is now reported in the basic feature list
    (pike --features).

o Sql

  - Bugfixes in listing Postgres fields.

  - If ENABLE_SPAWN_RSQLD is defined, rsqld will be spawned when
    needed to complete rsql queries.

  - Added streaming_query() method to Sql objects which enables larger
    result sets than available memory.

  - It is possible to iterate over the result object from big_query()
    queries directly in foreach.

  - Support UNDEFINED to designate NULL in emulated bindings.

  - Support for ODBC DSN files.

      Sql.Sql db = Sql.Sql("dsn://user:pass@host/database");

  - Some native support for the TDS protocol, used by Sybase and
    Microsoft SQL server.

      Sql.Sql db = Sql.Sql("tds://user:pass@host/database");

  - Support for the SQLite database added. A raw interface is
    available through the SQLite module.

      Sql.Sql db = Sql.Sql("sqlite://relative/path/to/file");
      Sql.Sql db = Sql.Sql("sqlite:///absolute/path/to/file");

  - Sql.pgsql.  New driver for native PostgreSQL network protocol support.
    It implements a superset of the existing Postgres driver.
    Current features: no library dependencies (no libpq), native binding
    support, streaming support, NOTIFY/LISTEN support (fully eventdriven,
    no polling), binary support for integer, float and string datatypes
    through big_typed_query(), native support for 64-bit ints and doubles,
    COPY FROM/TO STDIN/STDOUT support, multiple simultaneous streaming
    queries on the same connection (i.e. multiple PostgreSQL-portal-
    support), automatic precompilation and caching of often-used
    long-compile-time-needing queries, extended columndescriptions,
    accurate error messages under all circumstances, SSL-support,
    SQL-injection protection since it will ignore everything after the
    first semicolon delimiting the first command in the query, integrated
    statistics, _reconnect callback for sessions that use temptables.

    Performance tuned, with the helperclass _PGsql.PGsql it currently
    is around 21% faster than the old libpq based Postgres driver
    for smaller queries; for large bytea blobs and large texts, it
    speeds up even more.

    This driver serves URLs of the form:
				    pgsql:// (plain) and pgsqls:// (SSL).
    In case the old Postgres driver is disabled, this driver takes
    over postgres:// transparently as well.

o SSL

   It is now possible to set certificates to SSL connections. Example:

   SSL.sslfile ssl_connection(Stdio.File conn, string my_key,
                              string my_certificate)
   {
     ctx->client_rsa = Standards.PKCS.RSA.parse_private_key(my_key);

     // client_certificates is an array holding arrays of certificate
     // chains. since we've got a self-signed cert, our cert array has
     // only 1 element.
     ctx->client_certificates += ({ ({ my_certificate }) });

     return SSL.sslfile(conn, ctx, 1, 1);
   }

o Standards.IIM

  Some bugfixes in parsing Photoshop headers and DOS EPS Binary
  Headers.

o Standards.ISO639_2

  Updated with the latest ISO639-2 languages.

o Standards.URI

  - Updated to conform to RFC 3986.

  - Added methods get_query_variables(), set_query_variables(),
    add_query_variable() and add_query_variables() to give a better
    API to to handle query variables.

  - The new method get_http_query() returns the query part and
    get_http_path_query() returns both the path and the query, both
    coded according to RFC 1738.

o Standards.UUID

  - Added support for UUID version 5; name based with SHA hash, which
    can be generated from the make_version5 function.

  - An UUID object can now be used as namespace in the second argument
    to make_version3 as well as the new make_version5.

o Standards.XML.Wix

  - Updated to support a more recent version of the Wix tool chain.

  - Improved generation of 8.3-style filenames.

  - Added support for Shortcut and IniFile-nodes.

  - Added support for overriding the language and installer version.

  - Improved support for TTF-files.

o Stdio

  - Stdio.cp can now work recursively in a directory tree. It will
    also keep the permissions of files it copies.

  - Added Stdio.recursive_mv which works on every OS and also when the
    destination isn't on the same filesystem as the source.

  - Added more symbolc default termcap/terminfo bindings to
    Stdio.Readline.

  - Improved support for Stdio.Terminfo on NetBSD.

  - read_file(), read_bytes(), write_file() and append_file() will may
    now throw exceptions on uncommon errors such as when write_file is
    unable to write all its data.

  - Stdio.File->openat(), statat() and unlinkat() opens, stats and
    removes a file or directory relative to an open directory.

  - Stdio.FILE->unread() allows pushing back binary strings into the
    input stream, as opposed to ungets() which pushes back lines.

  - Stdio.UDP has had enable_multicast(), set_multicast_ttl(),
    add_membership() and drop_membership() added to make real
    multicast use possible.

o String

  - The function int2size has been rewritten to fixpoint as well as
    using the more common abbreviation of "B" for byte.

  - String.secure marks a string as "secure" which currently only
    means that the memory is cleared before it is freed.

o System

  - resolvepath() is now enabled on more OSes and falls back to
    realpath(3C) if resolvepath(2) doesn't exists.

  - On systems that support it, setproctitle() can be used to set the
    title of the running application.

  - Added support for POSIX style timers using clock_gettime(3) to
    allow for high resolution thread local cpu time and monotonic real
    time on reasonable modern Linux systems for gethrvtime() and
    gauge(). Added CPU_TIME_RESOLUTION, CPU_TIME_IMPLEMENTATION,
    REAL_TIME_IS_MONOTONIC, REAL_TIME_RESOLUTION and
    REAL_TIME_IMPLEMENTATION constants to tell the system
    capabilities.

o Tools.Hilfe

  - Added support for tab-completion on modules, global and local
    symbols and operators.

  - Added support for file/directory completion within strings

  - Added doc command and F1 key to print documentation on an item if
    available (currently only works for modules and classes written in
    pike).

o Tools.Standalone

  - "pike -x cgrep" now tries to parse Java code.

  - "pike -x features" now tests for several more features.

o Web.Crawler

  - Bugfix to support robots.txt created on windows.

  - User Agent change to "Mozilla 4.0 (PikeCrawler)"

o Web.RDF

  - Added add_statement() method which allows new relations to be
    added to an RDF set.


New modules / classes / methods added
-------------------------------------

o Fuse

  FUSE (Filesystem in USErspace) provides a simple interface for
  userspace programs to export a virtual filesystem to the Linux
  kernel (and some other OS:es). FUSE also aims to provide a secure
  method for non privileged users to create and mount their own
  filesystem implementations.

  This module implements the needed interfaces to make it possible to
  write a FUSE filesystem in Pike.

o ADT.List

  A simple doubly linked list of values.

    ADT.List l = ADT.List(1, 2, 3);
    l->insert(-1, 0);
    l->append(4, 5);
    foreach(l; int index; int value) {
      werror("  %d: value: %d\n", index, value);
    }

o ADT.Set

  ADT.Set implements a datatype for sets. These sets behave much
  like multisets, except that they are restricted to containing only
  one instance of each member value.

  From a performance viewpoint, it is probably more efficient for a
  Pike program to use mappings to serve as sets, rather than using an
  ADT.Set, so ADT.Set is mainly provided for the sake of completeness
  and code readability.

o Arg

  The new argument parser module allows for Getopt style argument
  parsing, but with a much simpler and object oriented API.

    class Parser
    {
      inherit Arg.Options;
      Opt verbose = NoOpt("-v")|NoOpt("--verbose")|Env("VERBOSE");
      Opt name = HasOpt("-f")|HasOpt("--file")|Default("out");
      Opt debug = MaybeOpt("-d")|MaybeOpt("--debug");
    }

    void main(int argc, array(string) argv)
    {
      Parser p = Parser(argv);
      werror("name: %O, verbose: %O, debug: %O\n",
             p->name, p->verbose, p->debug);
    }

  A more simplistic interface is also available for smaller hacks and
  programs.

    void main(int argc, array(string) argv)
    {
      mapping opts = Arg.parse(argv);
      argv = opts[Arg.REST];
    }

o GSSAPI

  Implements Pike access to GSS-API v2 as specified in RFC 2743. This
  API is used to authenticate users and servers, and optionally also
  to encrypt communication between them. The API is generic and can be
  used without any knowledge of the actual implementation of these
  security services, which is typically provided by the operating
  system.

  The most common implementation is Kerberos, which means that the
  main benefit of this API is to allow clients and servers to
  authenticate each other using Kerberos, thereby making single
  sign-on possible in a Kerberized environment.

o GTK2

  Wrapper for the GTK2 library. Not yet 100% completed, but usable.

o Protocols.DNS_SD

  This module provides an interface to DNS Service Discovery. The
  functionality of DNS-SD is described at <http://www.dns-sd.org/>.

  Using the Proctocols.DNS_SD.Service class a Pike program can
  announce services, for example a web site or a database server, to
  computers on the local network.

  When registering a service you need to provide the service name.
  service type, domain and port number. You can also optionally
  specify a TXT record. The contents of the TXT record varies between
  different services; for example, a web server can announce a path
  to a web page, and a printer spooler is able to list printer
  features such as color support or two-sided printing.

  The service is registered on the network for as long as the instance
  of the Service class is valid.

o Bittorrent.Tracker

  Bittorrent tracker with support for scraping and UDP extension.

o Protocols.HTTP.Server.Proxy

  A simple HTTP proxy.

o Standards.TLD

  Country domains and other TLDs according to IANA. Useful when
  parsing log information or providing live information about clients
  connecting to your server.

o Tools.Shoot

  Several new tests have been added to benchmark and improve on
  various aspects of Pike. ReplaceParallel and ReplaceSerial measure
  the times it takes to replace multiple substrings of a string in a
  single call to replace and with subsequent calls.

  TagRemoveArraySscanf, TagRemoveDivide, TagRemoveLoop,
  TagRemoveParserHTML, TagRemovePCRE, TagRemoveSearch and
  TagRemoveSscanf measure different methods of completing the same
  task; to remove XML tags from a string.

o Web.CGI

  Provides a CGI interface on the callee side. Retrieves information
  from environment variables and populates the variables in the
  Request object.


Deprecations
------------

o The keyword nomask has been deprecated. It was functionally
  equivivalent with the keyword final.

o Stdio.File->set_peek_file_before_read_callback() is deprecated.


Incompatible changes
--------------------

  These incompatible changes can be solved by adding #pike 7.6 to your
  source file or starting Pike with -V7.6 unless otherwise noted.

o main() environment

  The main() method will no longer be given the environment as a
  mapping as third argument. Use an explicit call to getenv() instead.

o Array.transpose_old

  This function has been removed.

o Calendar

  Changes made to fix inconsistensies has created som unavoidable
  incompatibilities. See the entry for Calendar in the functional
  changes section for details.

o _Charset

  The parts of this internal module that were written in Pike
  have moved to Locale.Charset.

o Crypto

  The old crypto functions from Pike 7.4 have been removed. These
  functions produced a warning when used in Pike 7.6.

o Debug.describe_program

  The API for this debug function has changed.

o Image.Image

  The functions select_colors(), map_closest(), map_fast() and
  map_fs() has been removed. Use Image.Colortable operations instead.

o Parser.XML

  The XML parsers are now stricter in verifying the correctness
  of the XML. The function compat_allow_errors can be called in the
  create method of the Parser.XML.Simple and Parser.XML.Validating
  (with "7.6" as argument for 7.6 compatibility). Parser.XML.Tree can
  be created with PARSE_COMPAT_ALLOW_ERRORS_7_6 as flag.

o Protocols.LDAP.client

  The "dn" field wasn't properly utf-8 decoded in 7.6 and earlier. If
  your application does it yourself, you need to use the compatibility
  version of this class.

o Protocols.TELNET

  The interface functions create() and setup() are now protected symbols.

o spider.XML

  The spider module no longer contains the XML parser. The functions
  isbasechar(), iscombiningchar(), isdigit(), isextender(),
  isfirstnamechar(), ishexchar(), isidiographic(), isletter(),
  isnamechar() and isspace() have also been moved to the Parser module.

o Sql.Sql

  Pike will no longer create a .column entry in SQL query responses if
  there is no table name.

o Standards.UUID

  Functions new() and new_string() have been removed. Use
  make_version1(-1)->encode() and make_version1(-1)->str() instead.

o Stdio

  The functions read_file(), read_bytes(), write_file() and
  append_file() now always throw errors on error conditions,
  to allow easier use as errno doesn't have to be checked.
  read_file() and read_bytes() still return 0 if the file
  does not exist.

o The modules Mird, Perl and Ssleay have been removed.

  Note that these modules are not available via the backwards
  compatibility layer.


C level module API
------------------

o Improved support for embedding.

  Several declarations and definitions (most notably the debug and
  runtime flags) have moved from main.h to pike_embed.h, in an attempt
  to add support for embedding.

o Major compiler API cleanups.

  The Pike compiler is now executing in a pike function context
  (rather than in an efun context), and it is possible to customize
  some of its behaviour via inherit (rather than via handler objects).
  As a consequence the compiler is now much closer to being thread-safe.

o The global variable next_timeout is no more. It has been replaced by
  a backend-specific variable. Added backend_lower_timeout() for accessing
  the new variable. This fixes issues GTK, GTK2 and sendfile had with the
  new backend implementation.

  NOTE! C-API incompatibility!
  NOTE! Changed the argument for backend callbacks!

  The argument is now a struct Backend_struct * when called
  at entry (was NULL).
  The argument is now NULL when called at exit (was 1).

o Pike_fp->context

  Pike_fp->context is now a pointer to the current struct inherit
  rather than a copy of it. This allows for easier access to
  inherited symbols in C-code.

o Inherit level argument added to several object handling functions.

  In order to implement subtyping of objects, an extra argument
  "inherit_level" has been added to many object indexing related
  functions.

o .cmod:

  Voidable pointer types are no longer promoted to mixed.

o Support for class symbols with storage in parent scope.
  Also added support for aliased symbols.

o Machine code backend for PPC64

  - Machine code generation is now supported for PowerPC in
    64-bit ABI mode.

o Objectiv-C embedding framwork

  - Experimental support for interfacing with Objective-C code
    has been added.

    NOTE! This API is experimental and is subject to change
    NOTE! without notice.

o Added %c and %C to get_all_args to get char * without
  NUL characters (no 0 valued characters inside the string).

  %c: char *	      Only narrow (8 bit) strings without NUL.
      This is identical to %s.

  %C: char * or NULL  Only narrow (8 bit) strings without NUL, or 0


Building and installing
-----------------------

o Dynamic modules now become DLLs on Windows.

  This means the homegrown module loader is no longer used, but it
  also means some DLL limitations:

  - PMOD_EXPORT is now required to allow access to an identifier in
    the pike core.

  - DLLs have to be recompiled too if pike.exe is recompiled.

  The primary motivation for this change is to work with the new
  library tracking (so-called "side-by-side assemblies") in Visual C++
  2005 and later.

o Added ABI selection.

  It's now possible to select whether to compile in 32bit or 64bit
  mode at configure time by using the --with-abi option.

o MinGW builds.

  It's now possible to build Pike in MinGW on windows computers from
  source distributions.

o Cmod precompiler.

  - The cmod precompiler (pike -x precompile) now supports declaring
    all autogenerated identifiers as static.

    NOTE! C-API incompatibility!
    NOTE! As a side effect of this change, the DECLARATIONS statement
    NOTE! now required in cmod files.


New simplified method to write external C/C++ modules
-----------------------------------------------------

  It's now suggested that you do not use the fairly complex 'pike
  internal' style of external modules (configure.in with
  AC_MODULE_INIT etc).

  It's also no longer required that you have a configure script to use
  pike -x module.

  Instead simply locate the includefiles using 'pike -x cflags', and
  convert .cmod to .c files using 'pike -x precompile'.

  An example rather minimal 'pike -x module' compatible Makefile,
  without a configure script, using .cmod format files for a simple
  local module:

  | CC=gcc
  | CFLAGS := -O9 -fweb -shared -Wall $(CFLAGS) $(shell $(PIKE) -x cflags) -g
  | LD=$(CC) -shared -lGL
  |
  | all: Spike.so
  |
  | install: Spike.so
  | 	cp $< $(MODULE_INSTALL_DIR)
  |
  | Spike.so: Spike.o
  | 	$(LD) -o Spike.so Spike.o $(LDFLAGS)
  |
  | Spike.o: Spike.c
  |
  | Spike.c: Spike.cmod
  | 	$(PIKE) -x precompile $< > $@

  It's usually OK not to use pike -x module at all, but it will pass
  on a few extra variables to your make (and configure script):

   PIKE: How to start the pike interpreter used running pike -x module
   MODULE_INSTALL_DIR: Where modules goes
   LOCAL_MODULE_PATH: Alternative (user local) module location