Changes

Changes in 2.5 PL8

- fixed a hang when hitting certain 8-bit chars in the From line

- fixed a file descriptor leak (from Andreas Luik)

- removed all Expires: processing


Changes in 2.5 PL7

- a few display issues fixed (from Kenneth Reek)

- buffer overflow in frm fixed

Changes in 2.5 PL6

- Fixed message-id in the In-Reply-To header.

- Actually show the -r option in the usage

- Whacked a few more places where we might follow a symlink in tmpdir

- Fastmail didn't grok commas correctly.

- Change lock.c to not use errno unless there's an errno

Changes in 2.5 PL5

- A few more buffer overruns fixed.

Changes in 2.5 PL4

- little bug in aliasdb.c (from Scot W. Hetzel <scot@genroco.com>)

- Message-Id: could overflow (from Darin Wayrynen <darin@deru.net>)

- Fixed the permissions setting problem with the edit mailbox function

- Added a ResizeScreen() call after show_msg() calls to hopefully deal wit
  SIGWINCH problems (inspired from Yuke Zhuge <zhuge@stone.pdx.informix.com>)
  
- Fixed problems under SunOS 4.1.3 (from Paul J. Menchini <mench@mench.com>)

- Fixed content encoding (from David Woolley  <david@djwhome.demon.co.uk>)

- Attempt to stop crashes with long recipient list (from Gilles Detillieux 
  <grdetil@scrc.umanitoba.ca>)
  

Changes in 2.5 PL3

- Fixed the declaration of outbound_from_name (from Richard Jackson 
  <rjackson@gmu.edu>)

- Fixed the leapyear macro (from Gerhard Niklasch)
  
- Fixed editing of a long CC line (from Derek Godfrey <djg@sjc.ascend.com>)

- Mail that lands in the sentmail with attachments now has the 
  Content-disposition header text added to the message (from Dave 
  Spensley)
  
- Made allto a bit longer (suggested by Dave Spensley 
  <davesp@typhoon.xnet.com>)

- Fixed Configure to find (again) fcntl on Linux

- can_access will now allow you to write to a char device (/dev/null)

Changes in 2.5 PL2

- Fixed (again, maybe?) savebyalias from Adam Benjamin <aebenjam@opentext.com>

- Fixed bug were an alias delete would crash elm if the mailbox was empty

Changes in 2.5 PL1

- Fixed savebyalias from Michael D Hughes <mdh@logcabin.woods.bridge.com> 
  and Robert Schwartz <rjs@elt.com>
  
- Changes for 8 bit clean from H.J. Lu <hjl@varesearch.com>

Changes in 2.5 PL0

- One minor change to make an easier compile on SparcLinux.

Changes in 2.5 PL0pre8

- Y2K stuff from Ed Symanzik <zik@msu.edu>

Changes in 2.5 PL0pre7

- minor change to deal with egcs

Changes in 2.5 PL0pre6

- newalias nows does a geteuid() instead of getuid().  This is to lets 
  SysAdmins do "su <user> -c 'newalias'"
  

Changes in 2.5 PL0pre5

- Fix for GECOS entries with comma-delimited sub-fields from Chip 
  Rosenthal <chip@unicom.com>.

- Fix? for keys getting screw up after a ^Z from Chip Rosenthal
  <chip@unicom.com>.


Changes in 2.5 PL0pre4

- A few changes for BSD from Chip Rosenthal <chip@unicom.com>.

- A few changes to Configure suggested by Chip Rosenthal <chip@unicom.com>.


Changes in 2.5 PL0pre3

- Configure didn't see fcntl locking on Linux.  Fix from Sean McDermott 
  <spm@intel.com>


Changes in 2.5 beta 3

- Fixed a problem where elm would try to close a file twice when
  sending an attachment (from reichler@ews.uiuc.edu)

- Added docs for environment variables to the man page (from Eric
  Haszlakiewicz <nimenees@uaccess.net>)

Changes in 2.5 beta 2

- IncomingFolders no longer set Magic Mode on for everything

- Configure should now deal with Linux.

- Make //pattern honor current limit (from Keith Neufeld<neufeld@pvi.org>)

- year 2000 fix.....?

- Building on Solaris requires /usr/ccs/bin to be in the PATH

Changes in 2.5 beta 1

- b)ounce honor DONT_ADD_FROM (from Keith Neufeld<neufeld@pvi.org>)

- removed [encode] display and reply (from Keith Neufeld<neufeld@pvi.org>)

- make right/left arrows work in the file browser the same as in the
  message index. (from Keith Neufeld<neufeld@pvi.org>)

- fixed a bug in the incomingfolders code (it wouldn't save) (from Bob Schwartz
  <rjs@elt.com>)
  
- fixed the meaning of noheader for forwards (from Keith Neufeld<neufeld@pvi.org>)



Changes in 2.5 alpha 11

- fixed a bug in the alias code (thanks to Keith Neufeld<neufeld@pvi.org>)

- fixed the generation of temp file names in newmbox.c

- allowed longer Cc: lines.  This was due to a crash with a group reply.

- added incomingfolders to the elmrc.  This is a list of folders that will
  automatically be opened with "magic mode" on. (inspired by a patch from
  Cristophe Kalt <kalt@ensta.fr>)


Changes in 2.5 alpha 10

- Press u on a message that is read but not deleted will mark the message as
  unread (from Larry Schwimmer <rosebud@cyclone.Stanford.EDU>).

- mmencode.1 man page is now installed with others (from Gregory Margo
  <gmargo@newton.vip.best.com>).

- Magic mode stuff is now in help and NLS (from Gregory Margo
  <gmargo@newton.vip.best.com>).

- allow elimination of aliases from larger aliases (To: bigalias -subalias).
  Based on incomplete patch by Waldi Ravens <waldi@moacs.indiv.nl.net>.  Also
  removes redundant declarations of get_alias_address(), which is declared in
  elm_globals.h (from Keith Neufeld<neufeld@pvi.org>)

- On an entry field that normally clears to be retyped from scratch as soon
  as the user presses a key (Subject: on header edit screen), allow right
  arrow and delete to begin editing the existing default answer. (from Keith
  Neufeld<neufeld@pvi.org>)

- Add a warning to Configure about disabling shell escapes (from Keith
  Neufeld<neufeld@pvi.org>)

- Eliminate complaints in doc/Makefile about duplicate makefile entries if
  mansrc and catmansrc are both "none" (from Keith Neufeld<neufeld@pvi.org>)

- Allow entry of uppercase commands on header display/edit screens, since
  that's how they're parenthesized on the display.  (from Bob Schwartz
  <rjs@elt.com> and Keith Neufeld <neufeld@pvi.org>

- Make 2+ -line error messages write over more of the screen instead of
  scrolling it. (from Bob Schwartz <rjs@elt.com> and Keith Neufeld
  <neufeld@pvi.org>

- Don't try to send a message if the user erased all the recipients on the
  header screen. (from Bob Schwartz <rjs@elt.com> and Keith Neufeld
  <neufeld@pvi.org>
  
- Extend save-by-name behavior to allow save-by-alias (if the user has an
  alias for this address). (from Bob Schwartz <rjs@elt.com> and Keith Neufeld
  <neufeld@pvi.org>

- Elm munged the permissions of a mailbox when not using system call locking
  and the mailbox was ordinary in all regards except for being on another
  filesystem than the temp mailbox.  Fixed. (from Bob Schwartz <rjs@elt.com>
  and Keith Neufeld <neufeld@pvi.org>.  Various other folks have also
  submitted patches that address this problem.
  
- Fix a bug in frm error message (from Bob Schwartz <rjs@elt.com>
  and Keith Neufeld <neufeld@pvi.org>


Changes in 2.5 alpha 9
======================

- Fixed a write call in lock.c (from Paul Close <pdc@lunch.engr.sgi.com>)

- When creating a new file, always make it so that symlinks aren't followed.
  This probably should only be done for temp files... 
  
- Add full name comments to the Cc: line of a group reply, and removes the
  To: recipient from the Cc: list (from Keith Neufeld<neufeld@pvi.org>)

- Make the keys on the status change screen change the status, instead of
  taking you up into the row to change the status with the space bar.  Also
  add he status change configuration to "elm -v" (from Keith
  Neufeld<neufeld@pvi.org>)

- In fastmail: fix the argument list generated to invoke the mailer to be
  space-separated rather than comma-separated; and change access() to
  can_access() (from Keith Neufeld<neufeld@pvi.org>)

- Fixed a problem with alt editor introduced in alpha 8 (from Robert
  Schwartz <rjs@elt.com>)

- Added vi-style `H'' and `L'' commands to go to the top |and bottom of the
  current indexed (Index, Alias) screen (from Keith Neufeld<neufeld@pvi.org>)

- More changes by Chip Rosenthal:

Totally new curses subsystem.

Encapsulated folder state into a (FOLDER) datatype.  Removes a
	gadzillion globals, and paves the way towards multiple
	folder support.

Maintain alias state separately, introducing stuff like num_aliases
	and curr_alias, to avoid the incestuous sharing of globals
	between folders and aliases.  This required some ugly
	changes to the menu stuff since they can no longer, for
	instance, just look at current -- they need to distinguish
	whether being run in folder or alias mode.  This likely is
	an interim step, Keith has mumbled stuff about replacing
	the wretched menu code.

Total rewrite of the Elm startup code.  Of particular note is that
	Elm entry into and exit from raw in a more controlled and
	consistent fashion.

Total rewrite of the Elm shutdown code.  Replaced the dozen or so separate
	procedures that all terminate Elm in slightly different ways,
	with a single leave() procedure.

New debug menu entered with '@' or '#' (the two are now equivalent).
	Added debug info on terminal (curses) settings.

The display_helpfile() program is now passed a just a topic (say "options")
	instead of a full pathname ("/usr/local/lib/elm/help-options").

File selection browser now has smart directory/pattern changing.  Instead
	of separate commands, just type in what you want (directory, pattern,
	or both) and it just figures it out.

When leaving the header menu for the ready-to-send screen, the display
	is updated properly.

Fixed the annoying screen-redraw-every-five-minutes.

MIME_SEND has been removed.  The MIME sending stuff is always on.
	The MIME_RECV stuff remains an option, controlled by the
	Configure answer to the MIME question.

Instead of just assuming base64 on binary data, see if quoted-printable
	will produce smaller encoded output.

New Makefiles.  There now is a Make.defs.SH file that includes all of
	the definitions (CC, BINDIR, etc.) plus the standard rules
	(all, clean, etc.).  The Makefile.SH files just define the
	list of targets, and the specific rules to build/install them.
	The Makefile.SH includes Make.defs plus depend.out to generate
	the resulting Makefile.  Thus Makefile.SH should be much smaller
	and simpler, and the subdirectory makefiles should now work
	consistently.

New "allow_setuid" setting.  This must be enabled in the global elmrc in
	order for Elm to operate setuid.  This prevents naive setuid
	installation without being too heavy handed.

Depreciated "keypad" setting.  New curses system knows whether or not
	to use HP keypad by termcap settings.

Reorganization of main header files.  The elmutil.h header is now
	obsolete.  All portions of the Elm package include elm_defs.h,
	and the main Elm program (but not utilities) include elm_globals.h.
	Moved many definitions specific to Elm (and not used by the
	utilities) from elm_defs.h to elm_globals.h.  Placed declaration
	of the dozen-or-so widely used globals in elm_defs.h.

Major overhaul of the global namespace.  Terminal state is now encapsulated
	into (struct term_info).  Folder state is now encapsulated
	into (FOLDER).  Most parameters incestuously shared between
	folder and alias now split apart (e.g. added num_aliases,
	curr_aliases).  New "opmode" parameter indicates Elm
	operating mode (OPMODE_READ, OPMODE_SEND, etc.).

New hdrs/port_termios.h encapsulates terminal setting commands in
	a POSIX wrapper.

sysdefs.h no longer defines individual help files, but rather a
	single "system_help_dir".

New lib/fast_getpw.c.  Does a "fast" getpwnam() by caching recent results.

Replaced gethostname() and gethostdomain() in lib/opt_utils.c with new
	get_hostname() and get_hostdomain() in new lib/gethostname.c.

Added new init_common() to perform initializations common to Elm and
	utilities in a consistent fashion.

get_word() not treats a (parenthesized comment) as a single word.

Added file selection browser and attachments menu to the message catalog.

New doc/help-selfolder with help on folder selection commands.

Reduced redundant macros, procedures, and definitions.  Stuff like
	use isspace() instead of whitespace(), streq() instead
	of equals(), TRUE instead of YES, strchr() instead of index(),
	etc.

All of the utilities now call initialize_common() rather than re-implementing
	the initialization code in an inconsistent fashion.

fastmail was broke.  Rewrote it.

Fixed readmsg so that when weeding is done, the default action is the
	opposite of the last action.  That is, with a weed list of
	"Foo: Bar:" the default is reject all other headers.  With
	a weed list of "Foo: !Bar:" the default is accept all other
	headers.




Changes in 2.5 alpha 8
======================


Lots of changes by Chip Rosenthal <chip@unicom.com>, including:


--- EXTERNAL CHANGES ---

New "make setup" target in base Makefile ensures .SH files up to date.

New "make tags" target in base Makefile creates tags for src and lib dirs.

New "mime-types" file added to $libdir.  (The mime-types file I've
    used basically is munged from my Gopher/HTTP server.  It might
    not be appropriate for the final release.)

Renamed "mlists" to "mlists.eg" in $libdir.

Renamed all help files from stuff like "elm-help.0" to "help-main".

Major overhaul of the elmrc file.  Options are now grouped together
    logically rather than appearing in alphabetical order.  (See next
    item for more info.)

Major change to $libdir/elmrc-info.  Previously, this was used simply
    as comments for annotating the elmrc.  Now, it directs the production
    of the elmrc.  The old elmrc-info format was:

	option-name
	# Comments to associate with the option.

    The new format is:

	# Comments to associate with the option.
	option-name

    Basically, just passes the elmrc-info file through, substituting
    in the "option-name = value" wherever it encounters an option
    name.  Appropriate checks are included so that settings missing
    from elmrc-info and extraneous settings in elmrc-info are noted
    with warnings.  If elmrc-info is missing, we fall back on a
    raw alphabetical listing.

Added "mmencode" to utils and "mmencode.1" to doc.  This is an
    encoder/decoder for base64 and quoted-printable.  It's based on
    the metamail version, which is freely redistributable.

"readmsg" now accepts "-w" to specify header weeding.  The default
    works like:  readmsg -w 'Subject: From: To: Cc: Apparently- Date:'
    The list entries are left-anchored "patmatch" patterns.

Added "printhdrs" elmrc option that allows header weeding to be
    specified when printing.  The value of this option becomes the
    "-w" arg to "readmsg".

Before, when printing a message, Elm discarded stdout/stderr if the
    "print" option had "%s" in it.  Now stdout/stderr is always left
    attached to the tty.

The "site hiding" stuff has been deleted.

Fixed potential problem where Elm was too aggressive stripping outside
    quotes from fullnames such as these:

	joe@acme.com ("Joe User :-)")
	joe@acme.com ("Joe" User "IV")

Pattern matching now allows \-quoting, such as "name\*with\*stars".

Fixed NLS utilities to compile on ANSIishy systems.

Improved SIGWINCH/SIGCONT support.  The places that already handled
    SIGWINCH (main menu and alias menu) now redraw automatically
    on SIGCONT as well.  The verify_transmission menu, file browser,
    and attachments menu do SIGWINCH/SIGCONT too.  Also, if you do a
    screen resize during a suspend, Elm now will notice.

Fixed possible problem with character reading.  (The problem was that
    ReadCh() was doing a read(2), while the rest of the package uses
    stdio.  Fixed ReadCh() to use getchar(3) instead.)

Temporarily ifdef'ed out the access() test on the ~e editor.  (I'd
    like to delete this entirely.)  This allows you to say
    "easyeditor = emacs" and not have to specify a full pathname.

Under the s)ave command, the cookie to mean "same as the last folder
    I used" has been changed from "." to "!!".

Under the s)ave command, if you specify a pattern you now get the
    new file selection browser instead of an "ls -C" listing.  The
    file selection browser also is started if you give a directory
    name.

Under the c)hange command, if you specify a pattern you now get the
    new file selection browser instead of an "ls -C" listing.  The
    file selection browser also is started if you give a directory
    name.

The "fwdattribution" setting now understands "%[begin|end]" formatting.
    Stuff to the left of the bar is printed in the opening attribution,
    stuff to the right in the close.  The %A escape is depreciated
    and is slated for removal.

Elm will try to use $PAGER or sel_pager to display help files, before
    falling back to its built-in help pager.

The chance of funky behavior during signals has been reduced.  (In
    other words, Elm was doing tons of things in its signal handlers
    that are dangerous.  Most, but not all, of them have been removed.)

Before, if you tried to f)orward a form Elm would allow it, but wouldn't
    let you edit or anything.  Now, it won't even do that.  It just
    gives a message prompting you to use b)ounce instead.

Elm is now MIME-sending capable right out of the box.  The answer that
    you give Configure with respect to the MIME question affects just
    the reading side of Elm.

Outbound mail messages (and MIME multipart/mixed parts) no longer
    have Content-Length: headers.  Content-Length: is generated only
    when messages are saved to a local folder.

The saved copy of a message will not have a Content-Type header
    unless the message is the source to a form.

If you omit the encoding from an [include ...] line, Elm will guess
    an appropriate encoding rather than assuming no encoding.

The content-type is now optional to [include ...].  That is, you
    can say just [include filename] and Elm will guess both the content
    type and encoding.

New [attach ...] command.  Use exactly the same way as [include ...].
    Difference is inclusions appear in-line with the message; attachments
    appear at the end.

New attachments menu.  To enter, after composing the message, strike
    "a" at the verify_transmission menu.

The MIME headers of outbound message have been changed.  Before
    the headers on a part would look something like:

	Content-Type: image/gif
	Content-Transfer-Encoding: base64
	Content-Name: /pub/archive/images/Drawing/bill.gif
	Content-Length: 5376

    Now they will appear something like:

	Content-Type: image/gif
	Content-Transfer-Encoding: base64
	Content-Description: GIF picture - version 87a 640 x 350, 16 colors
	Content-Disposition: attachment; filename="bill.gif"

The "textencoding" elmrc option has been dropped.  Elm now determines
    on its own whether a message is "8bit" or "7bit".  This setting
    will be ignored silently if it appears in elmrc.

Accordingly, Configure no longer asks about "defencoding".

The "charset" elmrc option is now used only if the message has 8-bit
    characters.  If the message contains solely 7-bit characters then
    this setting is ignored and US-ASCII is used.

Accordingly, Configure now offers ISO-8859-1 and not US-ASCII as the
    default "defcharset".


--- INTERNAL CHANGES - header files ---

Major reorganization of the header files.  Basically here is the mapping:

	old way			-->	new way

	defs.h			-->	elm_defs.h
	header.h + elm.h	-->	elm_globals.h

    Here is the general idea:  everything and anything in the Elm
    package should include "elm_defs.h".  It includes all of the
    definitions, inclusions, and portability hackups.  The *ONLY*
    files that should include "elm_globals.h" are those in the "src"
    subdirectory.  It defines the globals, and contains some of the
    definitions/declarations used with in Elm.  (Note that there are
    one or two routines in the library that require "elm_globals.h".
    I'd argue either that should be fixed, or they don't belong in
    the library.)

The "elm_defs.h" provides for ANSI prototyping.  It says:

	#ifndef P_
	# ifdef ANSI_C
	#   define P_(ARGS) ARGS
	# else
	#   define P_(ARGS) ()
	#   define const
	# endif
	#endif

    For example, the cvt_timestr_to_hhmmss() procedure (from
    lib/date_util.c) should be declared with something like:

	int cvt_timestr_to_hhmmss P_((const char *, int *, int *, int *));

A new header file, "elm_lib.h" has been introduced.  It contains
    declarations (and prototypes) for *every* entry in libutil.a.
    This header needs to be updated as changes are made to the library.
    This header is included automatically via "elm_defs.h", thus should
    not be included explicitly.  (Exception:  components of the library
    that have their own header files, such as ndbz.c, are not included
    in "elm_lib.h".)

Renamed "curses.h" to "elm_curses.h" to avoid name collision/confusion.

Added "port_wait.h" header.  Anything that needs to use wait(2) should
    include this.  Then, declare a "waitstatus_t" to hold the return
    status, and use the POSIX macros to test the status.

New headers "sndhdrs.h" and "sndparts.h" support the new mail sending
    subsystem.

A lot of stuff from "mime.h" has been dropped, retaining basically just
    that which is used by the receiving portion of Elm.

The MIME capability has been split into MIME_SEND and MIME_RECV.
    MIME_SEND is hardwired on.  MIME_RECV follows the Configure answer.
    (It is my hope that MIME_SEND will be deleted before release, and
    this stuff will just be standard.  As currently written, things will
    not compile if you try with MIME_SEND turned off.)

The "SHORT" macro has been dropped.  It was overloaded, used both as
    a string length as well as an argument to sort name types.  Instead,
    use TLEN (as in teeny length) to declare 10-byte strings.  As far
    as sort names (used to be SHORT/FULL/PAD), the padded sort names have
    been dropped (I fixed the menus to do the padding themselves), so
    the sort_name() procedures now accept a TRUE/FALSE boolean to
    return long versus short names.

I've dropped most of the happy macros for character names.  If you
    want to use ':' then say that, not COLON.

Renamed INCOMING/OUTGOING to LOCK_INCOMING/LOCK_OUTGOING.

Dropped JANUARY, FEBRUARY, ... definitions from elm_defs.h.

The MAIN_HELP, OPTIONS_HELP, etc. definitions have been changed from
    numbers to help file names and moved into sysdefs.h.

Rewrote the dprint() macro to fix subtle bug.  This means stuff like:

	if (something)
		dprint();
	else
		foobar();
    
    will work correctly now.  (Before, foobar() would be run if
    "something && !debug".)

Things in sysdefs.h/elm_defs.h/elm_globals.h were juggled about to their
    appropriate home.  For instance, the declaration of strstr() does
    not belong in sysdefs.h.  I put it in the new elm_lib.h header.

Deleted nearly all #includes from headers.h when converting over to
    "elm_globals.h".  This stuff belongs in "elm_defs.h".

Previously "headers.h" was used to declare the extern globals and
    "elm.h" was used to create/initialize them.  The new "elm_globals.h"
    does both.  Just declare INTERN before including the file (done
    in elm.c) to create them.

Drop explicit initialization of empty strings.  This means they now will
    be created in BSS rather than DATA, saving dozens of kilobytes of
    space in the executable and speeding startup.

All of the outbound message header globals are gone (to, subject, etc.)
    The new SEND_HEADERS datatype does this.

The "forwarding" and "text_encoding" globals are gone.

The "resize_screen" global has been renamed "caught_screen_change_sig" to
    reflect that it is used in SIGCONT handling as well as SIGWINCH.

One exception to the "all globals in elm_globals.h" rule: the "debug"
    and "debugfile" global declarations have been moved into "elm_defs.h"
    since so many things use them.


--- INTERNAL CHANGES - lib/* code changes ---

The entire library directory has been ANSIfied and checked with
    "gcc -Wall".  (Well...actually almost -Wall.  Due to the:

	/*#undef ... /* ... */

    stuff, I had to turn off the nested comment checks.

The can_access() procedure has been rewritten.  It now avoids an
    expensive fork when unnecessary.  Also, it now follows access(2)
    semantics (setting proper errno status rather than returning an
    errno value).  (All calls to can_access() have been updated to
    deal with this change.)

The error_description() routine has been dropped, and the entire
    package has been modified to use strerror().  lib/errno.c has
    been modified to provide a strerror() for those systems that
    lack it.

figure_out_addressee() routine has been modified so that you now
    have to pass the username as the second argument.  Previously
    it obtained this info from a global variable, which may or may
    not be initialized properly if figure_out_addressee() is called
    from a utility program.

The "next_field" value from parse_arpa_mailbox() now points to the
    character after scanning left off rather than where the next field
    begins.  What this essentially means is that whitespace is no
    longer skipped before setting the "next_field" value.

The patmatch() procedure now requires a third "opts" argument.

Added the following (possibly useful) routines to the library:
    basename(), trim_trailing_slashes(), trim_trailing_spaces(),
    and trim_quotes().


--- INTERNAL CHANGES - src/* code changes ---

The sort_name() and alias_sort_name() procedures used to take a
    FULL, SHORT, or PAD argument.  The menus were fixed to pad the
    name displays properly, so PAD is no longer necessary.  These
    routines now take a TRUE (display full name) or FALSE (display
    short name) argument.

Added ResizeScreen() to src/curses.c.  This may be called to effect
    a screen resize, and to update the associated globals.

Calling semantics to elm_date_str() [file src/date.c] changed.

New src/file_ops.c added that contains things such as file_open(),
    file_close(), file_seek(), file_rename(), etc.  They do exactly
    what you think they do -- plus they add error checking and
    print messages on failure.  In other words, rather than spending
    a half screenful of code to fopen() a file, just file_open() it.

The calling semantics to copy_message() have changed.  It no longer
    accepts a "prefix" argument.  New usage is:

	copy_message(FILE *dest_file, int cm_options);

    The following new options have been added:

	CM_PREFIX - Add "prefixchars" to each line.
	CM_FORWARDING - Replaces the (now removed) "forwarding" global.
	CM_ATTRIBUTION - Add an attribution line.

    Since attribution generation has been merged into this routine,
    the makeAttString() procedure has been moved into src/fileio.c

Calling semantics to display_helpfile() [file src/help.c] changed.
    Before, it used to take an integer to select a helpfile.  Now, it
    takes a helpfile pathname.  (These pathnames are now defined in
    sysdefs.h.)

Calling semantics to read_number() [file src/in_utils.c] changed.
    It used to have a hardwired screen line number and got its
    default response from a magic global variable.  Now you have
    to pass both of these as arguments.

Calling semantics to GetPrompt() [file src/in_utils.c] changed.
    Now you need to pass the timeout as a parameter, rather than using
    the global "timeout".  Say "0" for infinite timeout.  Also, if a
    SIGWINCH/SIGCONT occurs while in this procedure, the screen resize
    is done automatically.  Thus, the calling routine merely needs to
    do a redraw.  In other words, any menu that says:

	cmd = ReadCh();

    can get instant SIGWINCH/SIGCONT support by changing that line to:

	if ((cmd = ReadCh()) == NO_OP_COMMAND) {
	    do_redraw = TRUE;	/* or however you force a screen redraw */
	    continue;
	}

All of the signal initializations moved out of src/init.c and into
    the new initialize_signals() [file src/signals.c] routine.

The error message handling [file src/out_utils.c] is a bit
    smarter.  Now, for instance, the clear_error() routine performs
    output only if there is an error to clear.  It also returns TRUE
    when it does so, telling you that the cursor has moved.

The remail() procedure (implements "b)ounce") has been moved into
    src/reply.c.

The strip_commas() routine [file src/strings.c] has been modified to
    elide extraneous whitespace -- not create *more* of it!

The argv_zero() routine [file src/strings.c] has been dropped.  Use
    the new basename() library routine.

The system_call() procedure [file src/syscall.c] will set errno
    properly in the event of failure.  (By failure, I mean that something
    went wrong, and not that the subprocess had a non-zero status.)

If it has anything to do with sending mail, then chances are 87.43%
    that it has been rewritten, replaced, deleted, or moved to
    some other file.

All of the mail sending has been unified behind a new routine:

	int send_message(given_to, given_cc, given_subject, mssgtype)
	const char *given_to, *given_cc, *given_subject;
	int mssgtype;

    Supported values for the "mssgtype" argument are:

	#define SM_ORIGINAL     0 /* this is an original message            */
	#define SM_REPLY        1 /* this is a response to the current mssg */
	#define SM_FORWARD      2 /* the current mssg is being forwarded    */
	#define SM_FWDUNQUOTE   3 /* forward - without quoting or editing   */
	#define SM_FORMRESP     4 /* this is a response to a fill-out form  */


- Configure now looks for /var/spool/lock (from Bernd Ernesti
  <bernd@arresum.inka.de>)
  
- regexp.h was removed from the distribution (from Keith Neufeld<neufeld@pvi.org> )



Changes in 2.5 alpha 7
======================

- Minor tweaks to eliminate superfluous strlen() calls (from Chip
  Rosenthal <chip@unicom.com>)
  
- Significant rewrite of rfc822_toklen() (from Chip Rosenthal <chip@unicom.com>)

- Add %)F and %>F to attribution, for from name and from address (from Paul 
  Close <pdc@lunch.engr.sgi.com>)
  
- Add header fields to the send menu.  Only list to/cc/subject and any
  non-null fields (from Paul Close <pdc@lunch.engr.sgi.com>)
  
- More efficiency changes (from Paul Close <pdc@lunch.engr.sgi.com>)

- Don't use _down termcap if it is \n, since that doesn't work most
  places without mucking with stty (from Paul Close <pdc@lunch.engr.sgi.com>)

- elm won't allow the tmp mailbox to be a symlink (from Syd Weinstein
  <syd@Myxa.COM>)
  
- removed move_old_files_to_new() since it is no long needed (from Syd Weinstein
  <syd@Myxa.COM>)
  
- added the Status change screen (from Keith Neufeld<neufeld@pvi.org> )

- fixed a bug with the timezone stuff (from Chip Rosenthal <chip@unicom.com>)

- fixed a small bug in fwdattribution (from Chip Rosenthal <chip@unicom.com>)

- fixed readmsg to deal with long header lines (from Chip Rosenthal 
  <chip@unicom.com>)


Changes in 2.5 alpha 6
======================

- added handling of Apparently-To: (from Paul Close <pdc@lunch.engr.sgi.com>

- fixed a bug with MMDF & MIME (from Michael Elkins <elkins@aero.org>).

- fixed up the behaviour of ask & friends (from Keith Neufeld<neufeld@pvi.org> )

- added the ability to make entries of the form "!attribute" in a DT_MLT
  list or as a DT_SYN (which makes it an antonym, but it's still processed by
  DT_SYN).  It does _not_ allow "!attribute" in the elmrc, which would be
  silly.  It reimplements autocopy as a DT_MLT of !askreplycopy and replycopy,
  and it fixes alwaysleave to be a DT_MLT of alwayskeep and !alwaysstore,
  removing the `kludge that should be part of the "machine"'' from
  src/read_rc.c.  (from Keith Neufeld<neufeld@pvi.org> )
  
- prlong moved from $libdir to $bindir

- corrects inconsistent names used for header file inclusion.
  While most files include "foo.h", a couple used "../hdrs/foo.h".

- add a "fwdattribution" elmrc variable that enhances the handling of
  forwarded messages.  If undefined (the default), the current behaviour is
  preserved (standard attribution and prefix added on edited messages).  If
  this variable is defined, the prefix is never added, and instead of the
  standard attribution the inclusion is bracketed like:


Changes in 2.5 alpha 5
======================

- autocopy expanded into replycopy and askreplycopy (in elmrc).  replycopy 
  determines whether the default should be yes or no, and askreplycopy 
  determines whether it should happen automatically. (from Keith Neufeld 
  <neufeld@pvi.org> )
  
- askstore can be used to separately prompt to move read messages to 
  =received when you leave your mailbox.  I see its main use for people who 
  don't want to use =received but still like confirmation of deletion. (from 
  Keith Neufeld <neufeld@pvi.org> )
  
- Slight change to the behavior of resolv. (from Keith Neufeld < neufeld@pvi.org>)

- If you delete the last message in a folder that is a symlink, the folder will
  be emptied (this didn't work before).  (from Keith Neufeld < neufeld@pvi.org>)

- Added Chip Rosenthal's <chip@unicom.com> change to allow builds in a separate
  directory.

- Elm will now no longer stomp on folders if the user is over quota (from Larry
  Schwimmer <rosebud@cyclone.Stanford.EDU>).
  
- Changed the test for the Ultrix fflush bug.

Changes in 2.5 alpha 3
======================

- Added a change from Keith Neufeld <neufeld@pvi.org> to make the confirm for
  saving tagged messages a user configurable option (confirmtagsave in elmrc).
  
- Flush the edit buffer after an edit and now make it a user config thing 
 (editflush in elmrc).



Changes in 2.5 alpha 1
======================

- Added the changes that allow other folders to be treated as spool folders
  (the "Magic Mode" change).  Hitting '&' toggles this as well as the use of 
  the command line option -M.  The -M option also works on frm/nfrm.

- Added the "replied" code.  You must set showreply=ON in your elmrc for this
  to work.
  
- Added a key aliases to make emacs users happy:
  cntl-V is same as +
  cntl-N is down
  cntl-P is up

- Added the Ultrix fflush bug fix.

- Added some extra attribution code (inspired by a patch from 
  serini@ghost.sm.dsi.unimi.it (Piero Serini)).  The attribution string
  may now contain:
  
  %F      who the message is from (this is the same as %s)
  %D      date the message was created (the from message)
  %I      message ID
  %S      message subject

- Added patch to allow embedded date in folder names (from Mike Kenney 
  <mike@wavelet.apl.washington.edu>).  This allows:
  
          %h      month name ( 3 letter abbreviation )
          %y      last 2 digits of year
          %m      month number
          %d      day of the month
          %j      day of the year
				
  This patch also allows environment vars to be enclosed in {}.
  
  
- Added Wayne Davison's patch to confirm on saving tagged messages.

  
- If DONT_ADD_FROM is defined the Your name option doesn't do anything so it 
  now doesn't show up (from Larry Schwimmer <rosebud@cyclone.Stanford.EDU>).
  
- There were a couple places in mailmsg2 that strincmp was called with only 2
  parameters.

- The wildcard '*' is now allowed in the alternatives list.  For example, 
  wfp5p@*virginia.edu matches user wfp5p for all machines at virginia.edu.
  (from Michael Elkins <elkins@aero.org>).

- Added the mailing list stuff done by Paul Close <pdc@sgi.com>

- Allow the entry of aliases like the To: line 
  (from Keith Neufeld <neufeld@pvi.org>)
  
- Change to deal with parenthesized comments in when eliminating members from 
  an alias. (from Keith Neufeld <neufeld@pvi.org>)
  
- Allow a shell escape from the alias screen (just like from
  the index screen).  It does not put the shell escape onto the alias
  screen menu. (from Keith Neufeld <neufeld@pvi.org>)
  
- Allow the use of "T" from the builtin pager. (from Keith Neufeld 
  <neufeld@pvi.org>)

- Changes over to enhanced mode in for the print command.  From Chip Rosenthal
  <chip@unicom.com>

- Readmsg fixed to not truncate messages when called from within elm (it'll
  still have the the problem if run standalone).  From Chip Rosenthal 
  <chip@unicom.com>.

- Added fix for utime() from Chip Rosenthal <chip@unicom.com>

- Flush the edit buffer after an edit and now make it a user config thing 
  (editflush in elmrc).
  
- Made a few Configure changes