CHANGES.185

This is the most current changes file for PennMUSH. Please look it
over; each version contains new things which might significantly
affect the function of your server.  Changes are reported in reverse
chronological order (most recent first)

[SW] is Shawn Wagner, a PennMUSH developer (aka Raevnos), also
 responsible for most unattributed changes.
[GM] is Greg Millam, a PennMUSH developer (aka Walker)
[MG] is Mike Griffiths, a PennMUSH developer (aka Talvo)
[3] refers to code by (or inspired by) TinyMUSH 3.0
[MUX] refers to code by (or inspired by) TinyMUX 2.x
[Rhost] refers to code by (or inspired by) RhostMUSH

==========================================================================

Version 1.8.5 patchlevel 7                                 Mar 16, 2015

Fixes:
 * Bytes received by connections weren't being properly recorded for
   things like SESSION and recv().
 * Fix some highly unlikely crash bugs/memory leaks in atr_tab.c [MG]
 * @attribute/delete now deletes all aliases for the deleted attr, too. [MG]
 * Help fixes. [MG]
 * SOCKSET didn't always send the corret lineendings. [MG]
 * unsetq() and @include/clearregs were unpredictable, and hopefully now
   function more reliably. Reported by spork, Mercutio and others. [MG]
 * Remove compile warning when info_slave isn't used. [MG]
 * %X now always returns the capitalized version of %x, even when the returned
   value contains ANSI. [MG]
 * Fixed a few niggly bugs in @include, registers() and similar things
   making too many or too few registers available at the wrong time. From a
   report by Paige. [MG]
 * lit(%) returned %) instead of %. Reported by Ashen-Shugar. [MG]
 * PCRE's jit engine causes crashes when passed certain malicious regular 
   expressions. Disable it for all public-facing uses of REs. Reported by 
   Ashen-Shugar. [SW]
 * align() would sometimes show extra column separators when columns were
   merged. [MG]
 * @lock/interact was wrongly checked when paging a player with an @pageformat
   set. Reported by Odin. [MG]
 * Channels you had both @chan/gagged and @chan/combined would still show
   connect messages. Reported by Paige. [MG]
 * Fix a memory leak in @dolist/inline when used with @break. From a report by
   spork. [MG]
 * @chan/decompile <hidden channel> by a mortal didn't give an error message.
   Reported by Paige. [MG]
 * It was impossible to remove a thing from a channel which had been 
   restricted so only players could join. Reported by Paige. [MG]
 * Players with an @pageformat which eval'd to an empty string wouldn't know
   they were being paged. Reported by Ashen-Shugar. [MG]
 * Fix crash in named q-registers. Reported by Qon@M*U*S*H. [DC]
 * Enabling the log_commands config option while a LOG`CMD event was set would
   cause endless queueing, until the event handler was halted as a runaway
   object. Reported by spork. [MG]
 * Fix a potential overflow problem in the code that computes queue load 
   averages for @ps. [SW]
 * @motd/down and @motd/full are actually shown now. [MG]
 * Properly negotiate telnet with clients who spit out telnet codes instantly
   upon connection like MUSHclient. Reported by Mercutio and AND. [MG]
 * Improved configure support for NetBSD. [SW]
 * Fix a compile error on non-Windows, non-BSD, non-Linux OSes. [SW]

Hardcode:
 * Improvements to networking code; robustness on errors and portability. [SW] 

Functions:
 * Added two new column modifiers to wrap(), x and X, to allow truncation of
   fields instead of wrapping. Suggested by Qon, spork and others. [MG]

Minor changes:
 * @attribute/delete now displays the info for an attribute, as per @attrib
   <attr>, prior to deleting it. [MG]
 * The *div*() and *mod*() math functions now take an unlimited number of
   args, instead of exactly 2 args. Suggested by Mercutio, patch by Luke.
 * A SUSPECT player changing their name to the same thing doesn't trigger
   an alert. Suggested by 0x0045. [SW]
 * valid(attrvalue, <value>, <name>) checks if &<name> obj=<value> is valid.
   For @attrib/enum and /limit. [MG]
 * Added @flag/decompile and @attribute/decompile, to decompile each table
   to copy to another game. Suggested by spork. [MG]  
 * Newly set player passwords are salted to be more resistant to cracking via
   rainbow tables by somebody with a copy of the database. You still shouldn't
   be re-using a password for anything important, of course. Existing 
   passwords aren't automatically updated. [SW]
 * Added /clearregs to @trigger. Suggested by Volund. [MG]
 * @motd with no args now acts like @motd/list, instead of clearing the MotD.
   Use @motd/clear to clear the current one. Suggested by Bodin. [MG]
 * @command can now be used by non-players. From a gripe of Kitai's. [MG]
 * Update some configure macros to the latest versions. [SW]

Version 1.8.5 patchlevel 6                                 Jul 29, 2014

Hardcode:
 * Removed an old debugging variable to reduce memory usage. Suggested by
   spork. [MG]
 * ssl_slave now gets passed its settings via a pipe, instead of on the 
   command line. [SW]
 * Rewritten the telnet handler, to make it more manageable/modifiable,
   instead of it being a huge bunch of nested switch()es. [MG]
 * Update the SFMT random number generator to 1.4.1. [SW]
 * Built in and persistant uses of regular expressions are JIT compiled if the
   host's PCRE version supports it, for faster matching.
 * Better support for BSD systems. 

Functions:
 * valid() can now test for valid channel names. Suggested by Volund. [MG]
 * valid() now takes an optional <target> argument for some categories,
   to see if <string> is a valid name for <target>. [MG]
 * tan(90,d) now reports an error instead of a really big number. Suggested by
   Peppel. [SW]
 * More math functions report errors. [SW]
 * Added reglmatch[all][i]() functions, regexp versions of match[all](). [MG]
 * isdaylight() now accepts time and timezone arguments. [SW]
 * convtime() takes timezones besides just UTC. [SW]
 * foreach() is now ansi-aware. Reported by Ashen-Shugar. [MG]

Minor changes:
 * @channel/gag no longer persists across a full @shutdown. Suggested by
   Trinsec. [MG]
 * 'give' now gives more informative error messages when you try and give
   pennies to someone who has the maximum, or take pennies from someone who
   is broke. [MG]
 * The message shown when you need to run 'make update' is now slightly 
   clearer about what to do and why. By Rince.
 * Removed some 20 year old Sun OS workaround code. Suggested by Fraibert.
 * Slightly less confusing error message for @chan/buffer suggested by Sketch.
 * 'HELP ENTRIES' is generated on the fly, and shows the total number of
   pages. Suggested by [MG] and Trinsec. [SW]
 * @success/@osuccess/@asuccess now get the old location of the moving object
   in %0. Suggested by Riorgan. [SW]
 * @mail/review with no args now lists all mail you've sent to everyone.
   Suggested by Paige. [MG]
 * @mail/review now shows the recipient in the message list, instead of the
   sender. This is slightly more useful, since the person using the command
   is always the sender. [MG]
 * @mail/status now lets you twiddle the 'urgent' status of messages. From a
   suggestion by Paige. [MG]
 * You can now always see a channel if you've been joined to it and can speak
   on it, even if you don't pass its @clock/see. Suggested by Rince. [MG]
 * 'page foo=' now produces an error, instead of paging nothing. Suggested by
   Paige. [MG]
 * 'make localized' now ignores charset when building translation files, which
   stops it throwing errors unnecessarily. [MG]
 * Use a more robust configure check for timezone data files. [SW]
 * Added a /spoof switch to @mapsql, to spoof the enactor a la @trigger/spoof.
   Suggested by Mercutio. [MG]
 * Improved error reporting when info_slave or ssl_slave exits unexpectedly. [SW]
 * hasflag(<player>, connected) now returns 1 if the caller can see that the
   player is online, instead of always returning 0 for mortal callers. Most
   recently suggested by Odin@M*U*S*H. [MG]

Fixes:
 * Windows compile fixes by ChaosMageX.
 * Fix a crash when expanding the flagset size. Reported by spork. [DC]
 * @attribute/access incorrectly cleared an @attribute/limit on the same
   attr. Reported by Ton Kyrion. [MG]
 * Help/translation fixes by spork, Riorgan and others. [MG]
 * The 'give', 'take' and 'empty' commands displayed the wrong object names
   when mogrifiers were used. Reported by Riorgan. [MG]
 * Fix crash in ssl_slave when a connection drops in the middle of hostname
   lookup.
 * Crashbug in SQL fixed. Reported by Mercutio. [MG]
 * @mapsql/colnames cleared q-registers in the first (colnames) action list
   queued. Reported by Mercutio. [MG]
 * @channel/join would give an 'ambiguous' error, even if you couldn't see the
   other potentially-matching channels. Reported by Mercutio. [MG]
 * Penn negotiated charset as iso-8859-1 (latin1) even if the MUSH was
   running using a different charset. Reported by Tyler. [MG]
 * Better support for mixing IPv4 and IPv6 connections. FreeBSD and NetBSD
   should not need the --disable-ipv6 option to configure any more. [SW]
 * Fixed an obscure buffer-overwrite bug in the channel system. Reported by
   Volund@M*U*S*H/Roll@Multiverse Crisis MUSH. [MG]

Version 1.8.5 patchlevel 5                                 Nov 20, 2013

Commands:
 * Added @moniker, to allow colored names. See 'help monikers' for more.
   Requested by Mikoto Misaka. Idea from [MUX]. [MG]
 * After a year of deprecation, @emit/room has been removed as an alias for
   @lemit. [MG]
 * @chan/list now takes /on and /off switches, to filter the channels shown,
   and a /quiet switch to give a less verbose output. From a suggestion by
   Fraibert. [MG]
 * @trigger now takes a /spoof switch, which preserves the enactor. [MG]
 * New @hook/extend, which allows adding new functionality to built-in
   commands via softcode, while preserving the original command. [MG] [Rhost]

Major changes:
 * Regexp $-commands can now used name captures. They can be accessed via
   r(<name>, args). Suggested by Mercutio. [MG]

Hardcode:
 * A number of changes have been made to accomodate monikers. Notably, a lot
   of calls to the Name() macro now use the AName() macro. See the 
   HACKING.185 file for more info. [MG]
 * wild_match_case_r(), regexp_match_case_r(), pe_regs_set_rx_context()
   and pe_regs_set_rx_context_ansi() now take an extra argument, a PE_REGS*
   flag to determine how matching subpatterns are stored in pe_regs. [MG]
 * add_flag_generic() now returns an enum, and takes an additional final
   pointer arg, where the new flag will be stored. add_flag() and add_power()
   pass this pointer as NULL. [MG]
 * Improve heuristics used for resizing the flagcache hash table. [DC]
 * Update hashing functions used by various hash table implementations. [DC]

Fixes:
 * Commands which take multiple right-hand-side arguments could only have
   8191-length arguments, instead of the usual 8192. Reported by Jules. [MG]
 * It was possible to return non-printable characters from an SQL database,
   which could cause undefined weirdness. Reported by spork. [MG]
 * Many things which listed attributes from a parent, like lattrp() or
   examine/parent, didn't respect no_inherit fully. [MG]
 * Minor crashbug in @boot fixed. Reported by thomaswdyoung. [MG]
 * @grep now ignores ansi when looking for a value. Suggested by Mouvar. [MG]
 * colors() now complains when the color is 'n', 'd' or 'D', since they
   have no actual color value. Reported by spork. [MG]
 * Buffer overflow bug in ansi() fixed. Reported by tsen. [MG]
 * Flag corruption issue due to incorrect flagset refcounts fixed. [DC]
 * When a player was nuked, any @mail they had filed into another folder
   would be retained, and visible to a new player created later with the
   same dbref. Reported by Valkyrie. [MG]
 * Crashbug in @message. Reported by Yuriko. [MG]
 * linsert(,1,foo) now returns "foo", like it always used to, instead of
   returning an empty list. [MG]
 * SSL connections which had not yet logged into a player weren't always
   fully cleaned up. Patch by Fraibert.
 * channels() could include an extra delimiter in some cases. Reported by 
   Mouvar. [MG]
 * Translations weren't working in the chat system. Reported by Riorgan. [MG]

Minor changes:
 * strreplace() now works harder to keep existing ANSI intact, when you
   replace an un-ansi'd string into an ansi'd one. [MG]
 * colors(...,xterm256) can now return colors 0 to 15, to avoid confusion
   when converting xterm -> hex -> xterm. Reported by Trinsec. [MG]
 * New substitution, %k, returns the enactor's monikered name. [MUX] [MG]
 * Help fixes and tweaks. [MG]
 * PennMUSH can now read MUX-style passwords, to make moving dbs from MUX
   to Penn less painful. Requested by Muse. [MG]
 * Commands are no longer queued for HALTed non-players. (Since they'd be
   dropped from the queue anyway, rather than being run, it saves us jumping
   through a series of hoops for no good reason.) From a report by Muse and
   Ashen-Shugar. [MG]
 * The connect-screen WHO now shows moniker'd (ansi'd) names, if so 
   configured. Requested by Trinsec. [MG]
 * @mapsql and mapsql()'s attributes now receives named arguments. That is,
   '@mapsql me/foo=SELECT bar FROM table' receives the value of the 'bar'
   field as both %1 and [r(bar,args)]. [MG]
 * Wizards can now give players names with spaces, even when the 
   player_name_spaces option is off. Suggested by Paige. [MG]
 * Removed the last traces of the 'events' help-style command. Noone ever
   uses it, and we removed the helpfiles for it ages ago when the Event 
   System was added, anyway. [MG]
 * Updated the README for adding help-style commands, with clearer examples.
   Requested by AnnePearn. [MG]
 * Per-player mail limits can be set with @mailquota, to override the
   mail_limit @config option selectively. Suggested by Ricochet. [MG]
 * Attempting to create a character (at the login screen or via @pcreate)
   is now specific about whether the name is already in use or invalid,
   rather than leaving you guessing. [MG]
 * Likewise, when failing to connect, the login screen tells you whether
   there is no such player, or whether the password is wrong. [MG]
 * beep() now returns out of range, instead of permission denied, when given
   an arg which is not between 1 and 5. [MG]
 * When an evaluation @search hits the CPU limit, it will now tell you which
   object it failed on, to make it easier for you to continue with a partial 
   search. [MG]
 * Players are now warned when setting an unanchored regexp $-command, as
   this is often an oversight which is problematic to track down later. [MG]
 * More effort now goes into checking whether an attribute contains a valid
   $-command when it's set, to avoid doing so unnecessarily during command
   matching; the latter happens a lot more often. [MG]
   
Functions:
 * delete(), replace() and insert() have been renamed to strdelete(), 
   lreplace() and linsert(), respectively, for clarity and conformity. The
   original names remain as aliases. [MG]
 * moniker(<object>) returns an object's monikered name. [MUX] [MG]
 * mapsql() only stored 10 columns, instead of the advertised 30. Reported
   by Mercutio. [MG]
 * fn() now takes an <obj>/ prefix to its first arg, to evaluate the built-in
   function as <obj>. Useful when overriding side-effect functions. [MG]
 * r() now takes two arguments, allowing it to return the stack (%0-%9),
   itext() and stext() values, as well as q-registers(). [MG]
 * Mortals can now use playermem() and objmem() on their own objects.
   Suggested by Paige. [MG]
 * ljust() and rjust() now take an optional <trunc?> arg which, if true,
   causes strings already longer than <len> to be truncated. Suggested by
   many people over the years, most recently Rince and Quasar. [MG]

Version 1.8.5 patchlevel 4                                 Mar 17, 2013

Major changes:
 * The player_name_len @config option is now the maximum length of a player
   name, instead of the maximum length+1, because doing it that way was
   rather silly. You may wish to lower yours by 1 now to compensate.
   Suggested by Mercutio. [MG]
 * Having unescaped commas in the final arg of a function call is now
   deprecated, and will eventually become illegal. Any existing code which
   does this will spam you with a warning until you change it. [MG]
 * The stack (%0-%9) can now hold up to 30 arguments, with the additional 20
   available through v(10) to v(29). $-commands, @functions and many built-in
   commands and functions (such as ufun(), @mapsql, etc) now make use of
   these args. [MG] [MUX] [Rhost]

Commands:
 * @dolist now accepts /inline and related switches, like @switch does.
   Suggested by Reaver. [MG]
 * @teleport/list teleports a list of objects to a single destination.
   Suggested by Nyssa, [MUX]. [MG]
 * @sockset and SOCKSET can now show/set PROMPT_NEWLINES. [MG]
 * @halt now takes a /noeval switch, and will always strip curly braces when
   run /noeval. [MG]
 * Added @mail/cstats, which shows the mail stats you see on connection.
   Patch by spork.
 * Added a new switch for @command/add, /rsnoparse, which stops the rightside
   arguments (after the '=') from being evaluated. Also clarified the help
   to show that /noparse only affects leftside args. Suggested by spork. [MG]
 * WHO, DOING and SESSION now take wildcard patterns for names, as well as
   a prefix. [MG]
 * New option for @sockset/SOCKSET, 'stripaccents', to strip accents for a
   single connection as per the NOACCENTS flag. [MG]

Functions:
 * lpids()' second argument has been altered slightly, to allow you to list
   the pids for one specific object only. Suggested by spork. [MG]
 * The "player" field in pidinfo() has been renamed to "executor", for
   clarity. "player" still works for backwards compatability. From a request
   by tsen. [MG]
 * stringsecs() can now handle the additional time periods that etime() can 
   return (years, weeks). [MG]
 * colors(..., name) is much more efficient. [SW]
 * alias(object,) clears object's @alias. Suggested by spork. [MG]
 * v(<integer>) is no longer the same as get(me/<integer>). Instead, it
   always the <integer>th stack entry. [MG] [Rhost] [MUX]
 * It's now documented that replace() can alter multiple words at once. 
   Reported by Jules. [MG]
 * Some terminfo() information can now be retrieved by mortals on other 
   players' connections. Suggested by spork. [MG]
 * Added a new option to colors() to include ansi underline, invert and 
   flash from the input in the output. [MG]
 * locate() takes a new flag, 's', which will only return objects
   controlled by the looker. Suggested by spork. [MG]
 * @@() now takes a huge number of arguments, instead of ignoring commas in
   the final arg, to stop it giving the 'unescaped comma' warning. Requested
   by everyone. [MG]
 * terminfo() now tells you whether a connection is having accents stripped,
   either via NOACCENTS or @sockset. [MG]

Minor changes:
 * ansi() now accepts a space-separated list of color codes, with later
   groups overriding earlier ones. Allows you to specify basic ANSI codes
   after XTERM color codes. Suggested by spork, [Rhost]. [MG]
 * Compiler warning fixed by ChaosMageX.
 * Help fixes by many, including spork, Wiggles and Bodin. [MG]
 * The tables mapping color names and RGB values are now generated by a
   script in utils/. [SW]
 * Converting from named colors to the nearest xterm or ansi equivalent is
   much more efficient. [SW]
 * Converting from extended colors to ansi colors is smarter. Based on
   work by Padraic. [SW]
 * Email registration no longer pretends to work with really!old!addresses.
   [SW].
 * Better error reporting in logs and event hooks on failed email
   registration. Suggested by Padriac. [SW]
 * @decompile <exit> now reports the exit's source room. Suggested by
   telandra2. [MG]
 * You now only need to control an object to change its zone, or to use it
   as a zone. Previously, ownership was required. Suggested by Cheetah. [MG]
 * The hardcode now always ANSIs a string (such as an object name or help 
   topic), even if the enactor isn't set ANSI. Fixes puppets, etc. [MG]
 * When clients negotiate charset via telnet, US-ASCII and ASCII are now
   presented as options. If the client picks one of them, that connection
   will have accents stripped (as per the NOACCENTS flag). [MG]
 * New attribute flag, quiet, suppresses the usual confirmation message
   when an attribute's value or flags are changed. Suggested by Trinsec,
   bribed by qa'toq. [MG]

Fixes:
 * %c contained the evaluated input (the same as %u), instead of the
   unevaluated input. Reported by Ingwaz. [MG]
 * repeat() could write partial ANSI tags, causing buffer overflows. From a
   report by Ashen-Shugar. [MG]
 * dbcks didn't reset corrupt exit's source room. Patch by Padriac.
 * lmath(and,) returned 1 instead of 0. Reported by Padriac. [SW].
 * Mixing old-style ansi colors and xterm colors was broken. [MG]
 * Players who couldn't see xterm colors would sometimes get a highlighted
   foreground color by mistake. [MG]
 * colors(<old-style-ansi>, hex) and colors(*, 16color) were broken. 
   Reported by Kimiko. [MG]
 * ANSI highlight would sometimes bleed when xterm colors were downgraded to
   the basic 16-color ANSI palette. Reported by Kimiko. [MG]
 * @entrances didn't correctly report the source of an exit. Reported by
   spork. [MG]
 * Fix a crashbug in @command/delete. [MG]
 * Fix database corruption on @flag/add [SW]
 * Attempting to @mail someone without permission failed silently. Reported 
   by Kimiko. [MG]
 * Players connecting dark with 'cd' were hidden even if they shouldn't have
   been. Patch by thomaswdyoung.
 * Fixed a bug in colors(*,16color) with no ansi. [MG]
 
Hardcode:
 * The Builder() macro, which is unused in Penn but exists for backwards
   compatability, used command_check_byname(), which spammed people who
   could not build with "Permission denied." errors. It is now spam-free.
   By spork.

Version 1.8.5 patchlevel 3                                 Aug 23, 2012

Fixes:
 * Fix a crash bug in new color code. Discovered by ChaosMageX, fixed by
   Intrevis and [SW].

Functions:
 * name() with no arguments is now an error. It should always have been, but
   previously returned nothing for backwards compatability. If you still have
   code that relies on that behaviour, you are extremely lazy and should be
   ashamed of yourself. [MG]

Version 1.8.5 patchlevel 2                                 Aug 18, 2012

Hardcode:
 * render_string has been modified to work with multiple output formatters,
   so it can display the appropriate color syntax for the client. [GM]
 * Tweak the networking code to try to reduce output latency/queue size. [SW]

Major changes:
 * 256 Color support. Using 24 bit internally for representation and
   downscaling to the client's possible output values. Use
   SOCKSET colorstyle=xterm256 to see 256 color support, with
   ansi(+colorname/+colorname,...) or ansi(#hexcode/#hexcode) for
   extended color support, or @set me=xterm256. [GM] [MG]
 * New SOCKSET socket command for changing socket options. Replaces the
   @sockset command added in p1 for mortals, and can be used at the
   login screen. Suggested by Sketch. [MG]
 * @sockset is now @sockset <desc>[=<opt>,<value>[,...]], to query current
   options or set options for any descriptor. [MG]
 * The old @doing command has now been removed, and replaced with the normal
   standard-attribute command (@doing <object>[=<message>]). [MG]

Commands:
 * look now takes an /opaque switch, which hides contents listings. Suggested
   by Wiggles. [MG]
 * @break and @assert now take a /queued switch, which causes the new <action
   list> to be queued instead of run immediately, and an /inline switch which
   gives default behaviour. For Rhost compatability. [MG]
 * Two completely separate, but equally stupid, bugs made @nsprompt show
   nospoof info just like @prompt. [MG]
 * @open, open(), @dig, dig() accept specific garbage dbrefs to use
   for the exits they make. [SW]

Functions:
 * colors() returns a list of all known color names known by PennMUSH, and
   info about a particular color. [GM]
 * valid() has two new categories, for validating color names and color code
   sequences for ansi(). [MG]
 * etime() formats elapsed times using the same rules as in WHO. [SW]

Minor changes:
 * The meaning of the COLOR flag has been restored. ANSI flag once again 
   means only highlight is supported, while COLOR means all ANSI 8 colors
   (8*2 including highlight), flash, invert, etc are supported. [GM]
 * Charset negotiation didn't work on MSVC, and now forces iso-8859-1 when
   compiled that way instead. Reported by Daniel Powell. [MG]
 * @config/set and @config/save now log to the wizard_log. Suggested by
   Israphel. [MG]
 * The ICLOC attribute, which was set on newly created players but never used
   in any way, is no longer set. Suggested by Paige. [MG]
 * @destroy'ing, @undestroy'ing, and the actual destruction of players is
   now logged. Suggested by Padraic. [MG]
 * Translation fixes by Stoko.
 * round(4.5,0) behaved differently on Penn than it did on MUX, and
   inconsistantly with rounding to more places on Penn. [SW]
 * The 'On for' and 'Idle' time fields in WHO are smarter and won't overflow
   their space. Suggested by Ari and Paige. [SW]   
 * Add a few new test cases. Requested by Sketch. [MG]
 * It's now possible to clear/reset a prompt with @prompt <player>. [MG]
 * New Pick_Dbref @power for specifying specific garbage dbrefs to use in
   @create and friends. [SW]

Fixes:
 * Reordered some checks in tz.c to compile on more distributions. Problem
   reported by Zen@M*U*SH. [MG]
 * Setting an exit Transparent didn't cause its destination's desc to get
   shown as it should have. [MG]
 * It became impossible in p1 to give a delimiter to wrap() without also
   specifying a first line width. wrap(%0,%1,,%2) will now use the
   normal <width> arg for first line width, the same as wrap(%0,%1) does.
   Reported by Kitai. [MG]
 * Free PCRE-allocated memory with pcre_free() instead of free(). Reported
   by ChaosMageX@M*U*S*H. [DC]            
 * Fix some potential memory leaks in timezone file reading code. [SW]
 * Off-by-one error in the helpfile index code. [MG]
 * Improve docmentation of ip_addr config option. [SW]
 * Fix an edge case in ansi_string_replace that could cause an overflow.
   Reported by Ton Kyrion@M*U*S*H. [DC]
 * Fix a memory leak in LOGOUT. [SW]

Version 1.8.5 patchlevel 1                                 May 15, 2012

Hardcode:
 * The local_data_clone() function now receives an additional argument,
   which specifies whether @clone/preserve was used. You'll need to either
   update the definition of local_data_clone in src/local.c from the one in 
   src/local.dst or, if you've made no changes to src/local.c, just remove
   src/local.c and it'll be rebuilt automatically. [MG]
 * The orator variable (defined in game.c), for tracking the last object
   which made sound, has been deprecated, and may soon be removed; don't use
   it. Just pass the correct dbrefs to the notify*() functions if something 
   talks. [MG]
 * Several of the notify_*() family of functions, and the do_*emit()
   functions now take an extra arg, for properly tracking both the executor
   and the speaker when spoofing (@pemit/spoof, etc). [MG]

Deprecation:
 * @emit/room and @doing/header have been deprecated. Use @lemit and @poll,
   respectively, instead. [MG]

Major changes:
 * Built-in attributes are now stored in the database. This means that
   @attribute no longer needs to be in an @startup, as it will persist
   across reboots and shutdowns. The attribute_alias option in .cnf
   files is now deprecated as a result. [MG]
 * The database now uses a new versioning scheme (alongside the old
   one, for backwards compatability), using incrementing integers
   instead of bitwise flags. Ideas by [GM] and Covenent. [MG]

Minor changes:
 * Logged messages now have events in the LOG tree. Suggested by
   Padraic. [SW]
 * Objects belonging to destroyed players now get chowned to a
   configurable player (probate_judge in mush.cnf) instead of
   God. [SW]
 * Commands involving movement now check the enactor is a
   player/thing, instead of just relying on the command's
   restriction. [MG]
 * @switch, switch(), @[in]filter and attribute @locks (foo:bar) now
   do lte/gte wildcard matches on <=foo and >=foo patterns. Inspired
   by Wiggles. [MG]
 * New channel mogrifier, MOGRIFY`NOBUFFER, can be used to exclude
   certain @chats from the channel's recall buffer. Suggested by
   Wiggles. [MG]
 * Nospoof players now see the nospoof header on @cemits, making
   @nscemit actually have some use. Report by rimnar@M*U*S*H. [MG]
 * The login screen WHO can now be overridden using the who_file
   config option, which can be set as the path to a textfile or an
   obj/attr, like connect_file et al. From suggestions by Paige,
   Trelane and Mercutio. [MG]
 * Talek's chunk attribute storage system, which pages attribute
   contents to and from disk as needed, is now optional by popular
   request. Controlled via options.h [SW]
 * New lock, @lock/dropin, controls who can drop objects into that
   location. [SW]
 * @ps now shows a running average of the number of queue entries
   executed per second. [SW]
 * Try to discriminate between ssl_slave and other sources of
   connections to the local unix socket. [SW]
 * Repeating the PUEBOCLIENT socket command now causes PennMUSH to
   re-send the 'start parsing HTML' codes, for when Pueblo-enabled
   clients hiccough and show raw HTML. [MG]
 * The DESTINATION attribute for variable exits now receives the exit
   name/alias typed by the player as %0 when moving, to allow different
   destinations based on the exit alias used to move. It also has access
   to %c, %u, and other %-subs available when moving, Suggested by 
   Javelin and [GM]. [MG]
 * Translation updates and messages fixes by Stoko.
 * Help fixes by Todd@M*U*S*H, Steve Varley, Wiggles and others. [MG]
 * @message now takes a /silent and /noisy switch, the same as @pemit, with
   the default controlled by the silent_pemit @config option. message()
   is now silent, like pemit(). Suggested by Padraic. [MG]
 * The MUSH can now negotiate its character set with clients. There's still
   only one charset available, but now the client can know exactly what it's
   receiving. Suggested by Sketch. [MG] Make sure your MUSH is running in the
   right charset; see http://tinyurl.com/penn-charset for more information.
 * Additional test suite tests for setunion/setdiff/setinter() by Sketch.

Fixes:
 * When regexp matching an ansi'd string, named captures weren't set
   at all.  For non-ansi'd strings, non-named (numeric) captures were
   sometimes overwritten with the values of named captures. Reported
   by Sketch. [MG]
 * The configure script incorrectly reported that ssl_slave was
   unavailable when told to disable it via
   --disable-ssl_slave. Reported by Cheetah. [MG]
 * Sound from an AUDIBLE wagon object used to propagate into other
   wagon objects in the same location with an @listen of *. It now
   does so again.  Reported by Xeero and Padraic. [MG]
 * Sound will now also propagate through two @listens, instead of just
   one, as it used to. Reported by Mac@M*U*S*H. [MG]
 * Loading a reboot db is more robust. Inspired by Padraic. [SW]
 * scan() couldn't be used by rooms. Reported by Padraic. [MG]
 * Possible crashbug in the queue triggered by events reported by
   Rince@Wing Commander: New Horizon. [MG]
 * @list attribute would show duplicates for aliases. [MG]
 * insert(,foo,-1) no longer appends an extra delimiter. [MG]
 * @function/clone didn't work correctly, and could potentially cause
   a crash. [MG]
 * @clone didn't respect Many_Attribs. [MG]
 * Making a connection to a ssl port on a game not using ssl_slave
   could hang the mush. Reported by Minion. [SW]
 * @cemit and friends would notify you twice if you cemited to a channel
   you weren't on. Reported by spork. [MG]
 * regrep() could try to read uninitialized memory. [SW]
 * Fix a possible DOS from a local attacker. [SW]
 * Fix a couple of possible memory leaks. [MG]
 * ordinal(12) now returns twelfth, not twelveth. Patch by Minion.
 * trunc() now ignores ANSI. Reported by Sketch. [MG]
 * Evaluation @locks didn't receive %0-%9 args, even when they should.
   Reported by Zen@M*U*S*H. [MG]
 * @message/spoof now tracks both the object using @message and the object
   being spoofed, allowing for more accurate priv checks when evaluating
   the format messages. Reported by Mercutio. [MG]
 * When spoofing (@pemit/spoof, etc), any errors for failed locks or bad
   object names incorrectly went to the object being spoofed, not the object
   doing the spoofing. [MG]
 * @lemit incorrectly ignored an = and anything after it in its arg. [MG]
 * wrap(%0,strlen(%0)) would incorrectly wrap the string onto two lines. 
   Reported by Padraic. [MG]

Commands:
 * @edit/regexp[/all][/nocase], command version of regedit[all][i]().
   Suggested by Minion. [MG]
 * The @sql command no longer uses UNIMPLEMENTED_COMMAND when SQL
   support is not compiled into the MUSH. Instead, it just reports
   that no SQL db is available, like @mapsql does. [MG]
 * @slave, for restarting the info_slave and ssl_slave daemons from
   within the game. [SW]
 * @sockset <option>=<value>, for setting options on your descriptor.

Functions:
 * version() now reports svn revision/date where available. Patch by
   Padraic.
 * ldelete(), replace(), elements() and extract() now accept negative
   position arguments, to count backwards from the end of the list,
   similar to insert(). Suggested by Minion. [MG]
 * replace() now takes an optional output separator. [MG]
 * regmatch() now returns an error for invalid regexps, whether you're
   storing subpattern matches or not. Reported by Sketch. [MG]
 * time(), timefmt() and convsecs() can use symbolic timezones
   (US/Pacific, EST, etc.) if the host server has a time zone database
   installed. [SW]
 * playermem() and objmem() are smarter about calculating space being
   used by atributes [SW]
 * rnum() has been deprecated in favor of locate(). [SW]
 * step() no longer pads the final call with empty arguments, allowing you
   to use %+ or registers() to differentiate empty and missing args. [MG]
 * e() and exp() are now the same function internally, with e(), exp() and
   exp(1) all being equivilent. Suggested by Ari. [MG]
 * When wrap() hyphenates a word, the hyphen will now share the ANSI of the
   first character of the newline. [MG]
   
Version 1.8.5 patchlevel 0                                 Feb 12, 2012

Major changes:
 * The PCRE library is now required to build Penn, instead of falling
   back to an old, buggy version formerly bundled with the MUSH. [SW]
   Daniko assisted greatly with Windows changes and documentation.
 * The OpenSSL library is now required to build Penn. Again, Daniko has
   provided instructions for installing on Windows. [SW]
 * A new command prefix, "}", enables debug for a single execution of the
   command  - }think foo, }+who, etc. See 'help }' for more info. Suggested
   by Wiggles and [GM]. [MG]

Minor changes:
 * Added a new event, OBJECT`FLAG, triggered when a flag/power with the
   "event" restriction is set or cleared. Suggested by [SW]. [MG]
 * Expanded help on lock types and @powers. [MG]
 * @startups are now processed in db order (#0, #1, etc) on reboot instead
   of giving priority to player objects. [MG]
 * @startups (and local_startup() in local.c) are now triggered slightly
   later, ensuring that the db is fully loaded, and that the connected player
   list is loaded on @shutdown/reboot. Suggested by Padraic. [MG]
 * Switch names that are defined but not used by any commands are
   warned about on startup, and several unused ones have been removed
   from the master list. [SW]
 * Tweak utils/mkcmds.pl for the above change and to remove references
   to generating the obsolete patches.h. [SW]
 * Tweak the parameters of the skip list used to store memcheck counts
   to save some memory. [SW]
 * Bring the scheme scripts in utils up to date with current versions
   of chicken scheme (The implementation of the language they're
   written for.) [SW]
 * The test harness code strips trailing whitespace before attempting to match
   against expected results, making it easier to use bounded regexps. [SW]
 * @destination (for variable exits) is now a standard attribute, and has
   a helpfile. [MG]
 * Sound is now propagated through audible variable exits, and exits @linked
   to 'home'. Suggested by Mercutio. [SW]
 * The internal QUEUE attribute is no longer used to track the number
   of action lists an object has queued; the objdata system is used
   instead.  [SW]
 * isobjid() now requires a full object id (with :ctime), and will return 0
   for simple dbrefs, or objids of destroyed/invalid objects. isdbref() still
   matches both dbrefs and objids. Suggested by Covenent. [MG]
 * @entrances and entrances() now use the same internal routines as @search,
   and have similar (and saner) privs. From a suggestion by Sketch. [MG]
 * Connection messages that are evaluated from an attribute instead of
   read from a file get the dbref of the connected player as %1, as
   long as they're ones shown post-connection (motd_file, etc.) [SW]
 * Passwords are, thanks to OpenSSL digest routines, now stored in a
   less error prone, more secure, flexible and extendable
   format. Existing player passwords will be upgraded automatically on
   login. [SW]
 * New script in utils/ for manipulating player passwords in an offline
   database. Useful if you forget God's password. [SW]
 * The my_vsnprintf function in the hardcode has been renamed to
   mush_vsnprintf to avoid a conflict with MySQL. Reported by Balerion @ 
   Blood of Dragons. [MG]
 * The enactor (%#) for the PLAYER`DISCONNECT event is now more useful,
   instead of always being set to the dbref of the player disconnecting (%0).
   Suggested by Padraic. [MG]
 * The SOCKET`LOGINFAIL event now gets passed the name that was entered as
   %5 when no matching player is found. Suggested by Daniel Powell. [MG]

Commands:
 * Added @mail/review for reading mail you've sent, and @mail/retract for
   deleting unread mail that you sent. Idea from MUX. Patch by Minion.
 * @hook/list now shows error messages if you don't have permission to use
   it, or if a command has no hooks, instead of failing silently. Suggested
   by Padraic. [MG]
 * Wizards can now @pcreate players with restricted names, as well as being
   able to rename existing players to forbidden names. Suggested by Nymeria @
   Blood of Dragons. [MG]

Functions:
 * Functions that take a sort type argument now accept 'mtime' for
   sorting based on object's modification time. Patch by Minion.
 * New registers() function, like listq() but returns all kinds of registers
   (q-registers, %0-%9 args, iter/switch context, regexp captures). Also
   takes an output separator arg. Suggested by Wiggles. [MG]
 * digest(list) returns all message digest algorithms the server supports. [SW]
 * sha0() has been deprecated in favor of digest(). [SW]

Fixes:
 * lit() didn't properly handle '\'s. Reported by Wiggles. [MG]
 * listq() incorrectly returned all registers (including %0-%9, %i0, etc),
   not just %q-registers. Reported by Wiggles. [MG]
 * The mud_url code didn't always work correctly in Firefox because http
   headers weren't sent. [MG]
 * %0 and %1 arguments weren't properly passed through verb attributes after
   a change in p8. Reported by Padraic. [MG]
 * Fix really stupid copy-and-paste error in magic sorts on strings with
   ANSI, which broke setinter() and friends. Reported by Minion. [MG]
 * Help fixes by Wiggles, Mouvar and others. [MG]
 * Memory leak in ldelete() fixed. [MG]
 * Fix a crashbug reported by Minion. [MG]
 * Sorting functions now ignore ANSI when sorting a string as a number
   or dbref. From report by Mouvar. [MG]
 * strreplace() was broken for overly long length arguments. Reported
   by Ashen-Shugar.  [SW]
 * Compile errors on Windows with OpenSSL. Reported by Daniko.
 * @uptime could report negative times for the first dump, purge, etc after
   a game was started. Reported by Cooee. [MG]
 * override/inplace hooks from .cnf files didn't work correctly. [MG]
 * The portmsg announcer daemon was broken. Reported by Minion. [SW]
 * A change in p8 broke the sending of the Pueblo hello string in some cases.
   Reported by Mercutio. [MG]
 * Fix crash bug in @sitelock/remove [SW]
 * ssl_slave would drop connections if it couldn't resolve a valid hostname
   for the connecting IP address. [SW]
 * Non-matching regexp captures ($0, $<foo>, etc) sometimes had the value
   of the previous matching capture. From a report by Sketch. [MG]