Support extended length (71 character) To: and From: fields in QWKE replies. Reported by “g00r00” et al.
For system names between 6 and 10 characters, use “MMail/name” instead of “MultiMail” for the Blue Wave tear line.
Fixes for various warnings when building with current compilers; less cluttered build output.
Use CURS_DIR to specify the default library location on the command line when building; removed references to CURS_LIB and CURS_INC. Partly due to Rob Swindell.
Use “SYS=DOS” with Makefile.bcc, as with others, instead of “DOS=Y”.
Makefiles renamed PDCurses-style, for a simpler build process.
Binaries built against PDCurses current git (post-3.8); Makefiles modified accordingly.
Cleaned up documentation; also, the man page and README no longer include the version number, which is now only specified in config.h (and here).
See the git log for more details.
It is risen. No fooling.
Enable blinking in PDCurses for the ANSI viewer, wherever possible. (Binaries of this version were built with PDCurses 3.6, which adds software blinking for most platforms.)
The Windows build now allows resizing under Windows 10, or with ConEmu in some earlier versions.
Workaround for corrupt messages.dat when using the fallback (non-.ndx) QWK scan – keep looking for next valid header. Based on bug report by “XBessa”.
Blue Wave issue: macros string fields are different lengths in INF_HEADER vs. PDQ_HEADER. Reported by Frederic Cambus.
A 64-bit version built with MSVC would crash on startup (other 64-bit builds not affected) – fixed.
NetBSD fixes by Frederic Cambus.
The Mac build now reports itself as “MultiMail/Mac” instead of “MultiMail/Darwin”; also, “/DOS” vs. “/MS-DOS”, “/OS2” vs. “/OS/2”, “/Win” vs. “/Win32”.
Makefiles reorganized and reduced – elminiated the per-directory build process; all objects are built in the top-level directory now. DJGPP and MinGW are merged into the main Makefile; DEBUG, SDL and X11 are command-line options. MSVC does an all-at-once build (similar to PDCurses). Borland for Windows is restored, and building with old Turbo C++ is combined into Makefile.bcc via a command-line option. And Watcom now allows building for any platform via an option (instead of editing the Makefile).
Removed separate “modules” file (now builds all) and “version” file (now defined in config.h).
Dropped the mmail.spec file – I think this is better left to distro maintainers.
Dropped support for EMX; cleaned up some lingering references to RSX/NT.
Various compile warnings and errors with more modern compilers, undone.
New source code formatting – four-space indentation, no tabs.
Documentation “converted” to Markdown format (mainly just by renaming files); removed trailing spaces.
Most references to “Win32” changed to “Windows”, per Microsoft recommendations.
Spelling and other doc fixes by Robert James Clay.
Updated web and email addresses.
The Artist Formerly Known as Peter Karlsson is now “Peter Krefting”.
There was never a formal 0.50 release, but over the years, people have distributed versions of “0.50”, based on the source code repository. So, in an attempt to reduce confusion, I’m skipping “0.50” for the next release.
Built against PDCurses 3.3 (and 3.1 is now the minimum version that will work); removed XCURSES references, for the sake of PDCurses 3.2+. This is faster, and allows me to finally chuck the resize_term() kludge introduced in 0.42 (for the problem where the window scrolled up on return from shelling out in Win32).
You can now use the “Transparency” keyword with PDCurses – with the SDL port, you can achieve the same effects as in, e.g., Gnome Terminal with ncurses.
I’m distributing the Win32 binary with pdcurses.dll dynamically linked, purely so that you can swap it out with the SDL version if you like (see http://pdcurses.sf.net/). It’s fun to play with. :-)
In “animate” mode, the ANSI viewer now pauses after each character. This feature was omitted for a long time because the animation was slow enough on older machines, and the shortest possible delay was still too long. But animated mode is useless nowadays without it.
If the config files can’t be written (updated), MultiMail no longer aborts with an error; instead, a message is displayed, with a pause, and it continues.
Updated to GPL version 3.
This version now builds with PDCurses 3.0, and if using PDCurses, will no longer build with an earlier version. This gives us mouse support for DOS and OS/2, and better idling in the XT version, among other improvements.
New keyword: “Mouse”. Setting it to “No” turns off support for mouse input. Mainly this is to allow disabling of the mouse cursor; it also means you can select without holding shift in X11.
Use setlocale(), to give a correct display in environments such as the console under Ubuntu.
The FAQ and TODO files were outdated, and not very useful.
Removed the makefiles for EMX and Borland C++ for Windows. These compilers are basically dead. I also removed Makefile.wid – full- fledged wide-character support will be coming in the future, but for now, this file was just more clutter.
Changed “install” to a double-colon rule in Makefile to avoid conflict with INSTALL on non-case-sensitive filesystems, like Mac OS X’s HFS+.
Added support for Microsoft Visual C++, MinGW, and Open Watcom. I highly recommend Watcom, and I’m now using it for the OS/2 port. I’d use it for the Win32 port, but sadly the directory scan (used in the packet list) is using the slow POSIX-emulation method rather the fast native method, and I haven’t got that sorted out yet. MinGW is also very nice, and I’d use that, except that it needs MSVCRT.DLL, which doesn’t come with old Windows 95. So – rather than continuing with the 5-year-old Borland 5.5 (although it still works) – I’m delivering an MSVC-built version. This version doesn’t support set_new_handler(), but no one is likely to notice (famous last words). Also, I’ve finally dropped the pretense of supporting RSX/NT.
The starting directory and home directory are now added to the PATH searched for archivers and editors, so you can include them in the same directory with mm.exe. I’m taking advantage of this to bundle InfoZip’s zip and unzip, hugely bloating the Windows zip, but hopefully stopping the complaints I get from people who can’t figure out how to set up archiving themselves.
Support for the “GreekQWK” variant, which uses 0x0C instead of 0xE3 as the line delimiter. Reported by Dimitris Mandalidis. Note that there’s still no particular support for the Greek character set, but you can probably make it work by disabling character set conversion.
A mouse click in the lower half of the screen, in the ANSI viewer, now counts as if the spacebar had been pressed, instead of PgDn. Also, when the screen is at the start, even clicking in the top half counts as a space. This makes for easier mouse-only navigation with the new welcome screen display.
Included Makefile.wid for wide-character ncurses. This doesn’t really do anything yet, except give you the proper background character instead of the octothorpe, and work semi-decently in a UTF-8 Linux VT. For now, I don’t recommend using it, nor distributing binaries based on it.
Various documentation updates.
Revisions for newest version of PDCurses – fewer ifdefs needed. Note that MultiMail now must be compiled with PDCurses 2.7 or later.
Changed some compiler options to work better with recent gcc. Unfortunately it bloats the executable.
Removed the distinction between Bulletins and New file lists, so there’s only one prompt to bother you with now.
Minor code cleanup.
First, an apology: This version probably should’ve come out in July, when most of these changes were done… though there was a bug in one of the new features that didn’t show up until Halloween.
MultiMail now displays the appropriate screens (when it can determine them) on opening and closing a packet, instead of just adding them to the bulletin list. This is more like the way most other OLRs work (though not, e.g., Blue Wave). I hope you like it. :-) I had to adjust to it myself, but now I wouldn’t go back.
The ANSI viewer now includes partial support for AVATAR (level 0) and BSAVE (.BSV; text-only). As with the parsing of AT codes, these can be toggled, via ^V and ^B respectively.
The lower portion of the area list display has been redesigned. It now includes the name of the Door and BBS that created the packet, if available, and omits the duplicate description. The field formerly labelled “BBS:” is now “Name:”; “BBS:” is used for the software type instead. In the upper portion, there’s a little more room for the descriptions. Door type in OPX mode is not entirely satisfactory; some SX doors seem to use the field for BBS version instead. Neither of the new fields is available in Blue Wave mode.
With this version, I’ve added MIPS and AMD64 to my list of tested systems (no mods needed). However, I’m not including those binaries, because a) the MIPS binary I built is only for a Tivo, and doesn’t display onscreen anyway, so you’d have to telnet in or use a serial console; and b) there’s a small but annoying bug in the (beta) version of ncurses used on SF.net’s AMD64 compile farm machine, such that the window borders all show up with the wrong color. (The problem is not actually 64-bit related. Also, please note that MultiMail itself has been 64-bit ready for a long time, since being ported to the Alpha processor.)
Fix for some incorrect CONTROL.DAT files – a failed read (because a file was too short) was returning a dupe of the last line; now returns a blank.
Make reply endings in OPX conform better to what SX does – CR/LF, and a trailing zero. (I noticed that the tearline was being eaten by the Wildcat SX door on Doc’s Place, and there was no final CR/LF… I don’t think it used to do that, but anyway, this fixes it.)
ANSI viewer: Allowed right arrow to exit even in lynxNav mode – makes more sense with new Hello/Goodbye behavior; fewer false positives when interpreting AT codes; better handling of chars 0-31 and some others, under ncurses; no more overlapping of screens when a screen clear command is rendered.
Some code simplification.
New temporary directory and file system: Instead of using tmpnam() – which frankly never worked properly outside of Unix, and is sometimes regarded as a security risk there – MultiMail now generates temp directories of the form “workNNNN”, where NNNN is a random number, under the directory specified by the new .mmailrc keyword “TempDir”. By default, TempDir is set to the same value as mmHomeDir. NOTE THAT IF YOU’VE BEEN USING ENVIRONMENT VARAIBLES TO SPECIFY A TEMP DIR, IT WILL NO LONGER WORK. They aren’t even used to set the default (as with EDITOR). In the XT port, TempDir also now specifies the location of the swap file, if that feature is enabled. And temporary files are now created as “tmpNNNNN.txt” under the workNNNN directory, where NNNNN is a serial number; so they no longer clutter the base temp directory, they have the “.txt” extension some editors depend on, and there’s no longer a problem with editor-created backup files being cleared out.
Big improvements in the ANSI viewer under Unix: For those terminals (like the Linux console) that support the “smpch” terminfo capability, the IBM PC characters are passed through directly. For other terminals (like xterm), the use of the widely unimplemented ACS_BLOCK and ACS_BOARD has been replaced with an inverse space and ACS_CKBOARD, respectively. (In ncurses’ curses.h, ACS_BLOCK and ACS_BOARD are identified as “Teletype 5410v1 symbols”, while ACS_CKBOARD is in the “VT100 symbols” group. Everything supports VT100, but few terminals support the other class.) Note that the use of the inverse space does present a problem when characters have the A_BOLD attribute, as it’s interpreted differently by different terminals, so that some will make the cell “bright”, and some won’t; but this seems to look best on the greatest number of terminals. Also note that “xterm -fn vga -tn linux” is still the best way to view ANSI under X. ;-)
A new method for setting mmHomeDir: If it’s based on the environment variable “HOME”, add “mmail” to the end, as before; but if it’s based on “MMAIL” or the start directory, don’t. Practically speaking, this should make it apply “mmail” under Unix and not under DOS-ish systems, in most cases. That means that if you create a directory like “C:\MMAIL” and run mm.exe from there, it will no longer create “C:\MMAIL\MMAIL\DOWN”, etc., but simply “C:\MMAIL\DOWN”. Note that this (unlike the new temp system) applies only to new installations; upgrades will not alter existing MMAIL.RC pathnames.
The name shown at the top of the screen is now “MultiMail/Sysname vX.X” – i.e., the same as would appear in a QWK reply tearline – instead of “MultiMail offline reader vX.X”. Shorter, yet more informative. :-)
Doing a ^X during a bulletin would cause a segfault.
A malformed ‘H’ command (e.g., using “0” as a parameter) in ANSI could cause a segfault.
The window title is explicitly set to “MultiMail” in the Win32 and XCurses ports. This eliminates the problem where “UNZIP” would remain in the title after unarchiving a packet under Win 9x.
Revived cursor mode save/restore routines for PDCurses (except for Win32, where it doesn’t work); moved cursor restoration after endwin() call in mysystem() (this is why it didn’t do the job before).
Updated Makefiles for PDCurses 2.6, and GCC 3.2.1 (in the case of EMX). I strongly recommend upgrading to PDCurses 2.6, in most cases, as it already incorporates my patch for halfdelay(); however, there’s a fatal bug in the OS/2 version. (Write me for a patch.) I also recommend ncurses 5.3 over 5.2 – for one thing, it fixes the problem with the bool declaration that broke searching and animation in some installations – but I’ve left it as 5.2 in the Makefile, for now, because I have too many systems to recompile it for. :-)
Remaining compilation warnings eliminated for most platforms: Got rid of tmpnam() (see above), moved “%y” strftime() strings outside the calls (copied this from Ingo Brueckl), and added “-Wno-deprecated” to the Makefile options for the latest versions of gcc.
Useless .mmailrc keyword “homeDir” removed. (It specified the base value for mmHomeDir, but in most cases that was already specified explicitly on the very next line.)
Stronger attempts to use backslash characters instead of slash where appropriate in MMAIL.RC (depending on the platform). But again, this is cosmetic.
Optimizations, comments, etc.
Changed the way that the length of QWK replies is calculated. This is a speculative fix for a reported problem with corrupt replies that I couldn’t reproduce; but from what was described, this should do it. At any rate, the new code is simpler and cleaner. Reported by Jimmy Day et al.
Pressing the “END” key in string input fields now works correctly, taking the cursor to the end of the field instead of ending the edit. Reported by Touko Rajala.
Bogus/unintentional ‘@’ color code sequences could cause segfaults in the ANSI viewer. Reported by Neall Mercado.
In the Win32 version, under NT (though not 9x), the temporary directories were not being deleted.
The filter was not working correctly in the LittleAreaListWindow.
Junk could appear in the LetterList on a resort if some messages were read.
Narrowed the width of some windows as a workaround for display bugs in xterm.
Added “-inul” to the default command lines for RAR to suppress the output.
Minor internal changes to allow compilation and remove warnings under Sun’s SUNWspro C++ compiler. This also entailed the elimination of the sillier of the “set_Letter_Params()” functions, which was long overdue. :-)
Miscellaneous small changes for efficiency.
A big one: Filter feature in all list windows, invoked by pressing ‘^’ (shift-6). See the “FILTERING” section of the man page for details. This was inspired by similar features in the Pan and BNR newsreaders.
Clock in the upper right corner of the letter window. (I would’ve added it to the others, but they lack an obvious place to put it. I figured that one spends most of one’s time in the letter window anyway.) It can show time of day, time since startup, or be disabled, as controlled by the new keyword “ClockMode”. Note that this required changes to the keyboard routines (now using halfdelay() instead of a blocking call), which in turn affected the ability of MultiMail to give up CPU time when idling. Most ports still use little CPU, but the XT version pegs it high, except under DOSEmu. (Old versions of DOSEmu, ironically, instead show high CPU usage with the new DOS (DPMI) version, which doesn’t have that problem under Windows or OS/2.) Also, I had to fix PDCurses to make this work – write to me if you want the patch.
“AT code” parser for the ANSI viewer, for PCBoard and Wildcat ‘@’ color codes. (I’ve only seen these used in the sample packets that came with a few readers – do they exist in the wild?) Also, “ANSI music”, though still not played, is now filtered out, so at least it doesn’t clutter the image.
The “UseColors” keyword is now available in the PDCurses (DOSish) ports. That means a monochrome mode, which should help on some displays.
The packet name is now displayed when in the letter list, letter window or ANSI viewer. No more having to back out to check. :-) The placement of the packet name in the area list window has been changed for consistency.
The tagline window can now be sorted, optionally.
An “Opening…” notice is now displayed when entering an area, as when opening a packet. (This operation can be slow, so it’s good to have a hint that MultiMail got the message.)
The type ‘m’ (Unix mailbox) parser in the SOUP module was missing some valid “From “ separators, so I incorporated code based on the VALID macro from the C-Client library by Mark Crispin. I’ve read conflicting claims about whether its “free-fork” license is GPL-compatible, but since I don’t see a better way to do this, I’ve chosen to assume that it is, at least for now. Anyway, this may in fact be the only valid (no pun intended) way to parse a Unix mailbox without false positives or negatives.
The ANSI viewer was not displaying the headers of SOUP messages when animating them, or thereafter. This was really a problem with multiblock messages (otherwise not noticeable except in the XT version, and then only with a message over 64K). Also, the ANSI viewer now acknowledges “expert” mode (i.e., it omits the “F1 or ? - Help” message from the lower right).
The address book and tagline windows now show the number of items, like the other lists.
In an xterm, when the tagline window was active, bits of the scroll bar and the screen border were being replaced with blanks. I’ve fixed this in an ad hoc way, but I’m not quite sure what caused it.
An empty file list in the packet window caused problems. Normally you wouldn’t see this, because there’s always at least a “..” entry for the parent directory. But the filesystem on a Windows CE device apparently lacks them, as I found when running MultiMail/XT under PocketDOS.
Setting the “UseColors” keyword to “No” now forces the use of the default colors (or rather, the monochrome version of them), as the .mmailrc comment had already claimed it did, instead of the monochrome version of the colors specified by “ColorFile”.
The “Reverse” attribute will no longer be stripped from a color file that’s rewritten by a PDCurses (DOSish) port.
If the tagline file existed, but was completely empty, a segfault would occur on selecting the (nonexistent) tagline when composing a message. Reported by Touko Rajala.
Some really basic stuff added to the man page – it always should’ve been there. Further revision is needed.
Minor speedups.
In the Win32 version, under Windows XP and perhaps 2000 (?), the window was scrolled up on returning from a shell (unarchiving, etc.), unless the scrollback buffer was disabled. Although cosmetic, this was enough of a nuisance that I’m releasing 0.42 now.
Added strings.h to the includes in mmail.h to get it to compile under QNX. (If you’re trying this, edit the Makefile to say “-lncurses” rather than “-lcurses”.) Note that the background characters (ACS_BOARD and ACS_CKBOARD) aren’t what they should be in QNX; otherwise, it works well.
New error message for the case when no files are found in the work directory after uncompression, but no error code was returned by the unarchiver. (Previously this resulted in the misleading message “Packet type not recognized”; that message is now reserved for times when there are actually some files in the work directory.)
Since version 0.38, when I moved development to a new computer, the documentation file “mm.txt”, included with the DOSish ports, had not been converted correctly by the script that builds those archives – it was left as raw troff code instead of a rendered man page, and the line endings were left as LF (Unix style). In other words, it was basically illegible. The discouraging part of this is that no one reported it, which makes me think that no one’s reading the documentation. :-(
Other, minor updates to the documentation.
Options can now be specified on the command line. Any .mmailrc option may be used, though not all will actually work (some of them are used in initialization before this point). Options are preceded by ‘-‘ or “–”, and must finish before a packet is specified. See the man page for more details.
New keyword: “IgnoreNDX”. This makes MultiMail ignore the *.NDX files in QWK packets, in favor of the new (slower, but safer) indexing method that uses only MESSAGES.DAT. Turn this on if you get bogus *.NDX files from your BBS (as seems all too common).
Big bug: Reply splitting in QWK on DOSish systems (DOS, Win32, OS/2) generated a corrupt reply packet (always). It’s been this way a long time; possibly since the message split feature was introduced. I never noticed it because a) I rarely split messages, b) I seldom use QWK anymore – all SOUP and BW now, c) I never use the DOSish ports, except for testing, and d) the split replies look OK, until you close and reopen the packet. I had a lot of bug reports which, in retrospect, seem to be about this, but I never got enough detail to make sense of the problem, until Ken Hrynchuk reported on it. Many thanks to him.
Bogus values in the *.NDX files of a QWK packet are now checked for and often handled without having to abort. In some cases, MultiMail is able to revert to the alternate indexing method automatically; in others, the message “ERROR READING MESSAGES.DAT” appears in the letter list, instead of the gibberish that resulted when following a bogus pointer. For those cases which still cause problems, see the new “IgnoreNDX” keyword.
Problems with the ncurses configuration on Mandrake 8 (well, on my system, anyway) caused MultiMail to hang on searches or ANSI animation. There’s now a partial workaround for this (searching works OK, ANSI anim still hangs) in MultiMail. But the proper fix is this: Download the ncurses source; configure; manually edit the line in curses.h that defines bool to say “typedef unsigned char bool” instead of “typedef unsigned bool”; compile; and install. I still have to figure out why the typedef comes out that way; the same version of ncurses (with the same configure script) works correctly on my old Slackware system. You might see this problem on other systems as well.
Blue Wave messages with Internet kludge lines (as produced by version 4 doors) were causing segfaults. This is what happens when a feature doesn’t get exercised for a while. :-(
Segfault on a message with a MSGID kludge line, but no other body text.
In the Win32 port, packet time stamping now works under Win 9x, though it still doesn’t work on network shares under NT.
Some unprintable characters weren’t being filtered out when running in an xterm, resulting in a garbled display.
In the 16-bit port, in QWK packets with no *.NDX files, MultiMail was failing if the MESSAGES.DAT file was larger than 32K.
The END key now works under screen (for example) with ncurses.
The message-splitting function now uses parentheses instead of brackets, and the part number is padded with leading zeroes. I did this to match standard practice in the alt.binaries.* newsgroups.
Added the undocumented option “-#” to the default command lines for PKZip and PKUnzip. This suppresses output, equivalent to the “-q” option used in the default InfoZip command lines. Suggested by K.H. (Note that if you’re upgrading, you’ll have to put it in manually, if you want it; as always, MultiMail won’t alter values that are already specified in its config file.)
Shelled commands (zip, unzip, edit) can now handle paths with spaces in them.
On MultiMail’s first run, on DOSish systems, the MMAIL.RC ended up with paths that had a mix of slash and backslash characters. Although this didn’t interfere with operation, it was a cosmetic defect.
The defaults in the top-level Makefile have changed to “
I figured out the problem that led to the hasty release of version 0.40 –
it was a missing pair of parentheses.
Undid one part of the BSOC, which (perversely enough, since this is exactly what it was intended to cut down on) was causing crashes in the Win32, XT, and OS/2 ports. I still haven’t nailed down WHY it was doing this, but I isolated the source of the trouble by brute-force analysis. :-( I promise better testing next time.
Useless mouse code now omitted in DOS, XT and OS/2 ports. (Yes, I hope to actually support the mouse on these platforms eventually. First, mouse support has to be added to their respective ports of PDCurses.)
Mouse functionality. Currently, this is only supported under X (with either ncurses or PDCurses), in the Linux console with gpm, and in the Win32 port. See the man page for a description of how to use this – not well-described there, but it should be fairly intuitive.
A new flag to join Marked, Replied, Read and Private: Saved. It’s set automatically when a message is saved (natch). It appears in the letter list as a lowercase ‘s’, and as a new flag in the group displayed in the letter window. (Note that if the message is also Marked, only the ‘M’ will appear in the letter list – I was short of columns.) Because of this, the names displayed in the letter window have been shortened. Saving a message no longer marks it as Read, since that seems redundant now (though it still does clear the Marked flag). The new flag is mapped appropriately in .XTI files, but not in MAIL.FDX, so it’s not preserved in OPX packets.
New transparency method: The transparent background color is the one specified by “Main_Back” in the colors file, rather than always black. This allows the use of transparency with light backgrounds, etc. (See the screenshots web page.)
A longstanding bug: It was impossible to rename packets after opening them. Finally fixed, as part of a big string-operation cleanup.
Another longstanding one, also part of the BSOC: On some platforms, when
specifying a packet name on the command line, names of certain lengths
wouldn’t work. After I finally sorted this out, the surprising thing was
that it had ever been working for names of any length.
Many potential buffer overflows and bogus limits fixed in the BSOC. Others remain. :-(
Since time immemorial: When pulled from the address book into a message, the “To:” field was being restricted to 29 characters, regardless of packet type. Also, the length of names in the address book had been limited to 44 characters.
A fresh one: Reply packets were being stamped with a bogus date.
The packet list, if sorted by time, is no longer rearranged when a packet is Touched.
Replies are now preserved in the event of a crash during the write-out process. Inspired by Matt Munson.
The highlight bar wasn’t showing up in XCurses – probably since about version 0.18, which shows how often I try that port. (I also had some other problems with XCurses, as yet unsolved; it sometimes locked up on starting.)
Stupid bug in SOUP: segfault on message areas with 8-character filenames.
The line showing “BBS:” and “Sysop name:” in the area list is now omitted if neither is defined (as in SOUP packets), leaving an extra line for the areas themselves.
The minimum screen size is now 60x17. I’m shooting for 40 columns. ;-)
First new version in over seven months! See, it’s not dead after all.
Transparency: There are three new keywords in the .mmailrc which are mainly for making the background transparent, for use in programs like Eterm which can put an image in the background: “UseColors” (default: yes), “Transparency” (default: no), and “BackFill”. “Transparency” only works in ncurses; “BackFill” is the only one available on all platforms. See the man page for more on these.
New date/timestamp handling: When updating packets with the .red markers, MultiMail now preserves the original timestamp of the file. This means that date sorting will reflect the order in which packets arrived, rather than the order in which they were last read. (Note, however, that this feature doesn’t quite work correctly in the Win32 port (see README.win). One way around this would be to use PKZIP with the “-k” option; or just use the DOS port.) Inspired by Ingo Brueckl. At first I had this toggled by a keyword, but then…
The complement to the new timestamp handling is the ‘T’ function in the packet window, which “Touches” a file, setting its timestamp to the current time. I needed this because I actually depend on the timestamps being updated sometimes.
New “ExpertMode” option suppresses the help text, for those who are familiar with the commands and would rather use more of the screen for text. Four lines are gained in the packet, area and letter lists, among other gains.
In a related note, you now can gain seven lines in the packet list by undefining VANITY_PLATE in config.h, but this is strictly a compile-time option. :-)
OPX secretly restored. ;-)
Removed several older keywords to simplify the .mmailrc: BuildPersArea (even on my XT, it didn’t really take much longer), MakeOldFlags, UseScrollBars, and AutoSaveRead (all these are now always on).
The default editor for the Win32 port is now “start /w notepad”. I’m not crazy about this, but under NT (in VMware), EDIT was completely locking up. Unfortunately MS provides no text-mode, native Win32 editor. (BTW, an interesting aside: Under NT, “notepad” is actually sufficient (the calling program waits for it to return); the “start /w” bit is for the benefit of 9x, which otherwise launches notepad and returns immediately.)
The Win32 port had trouble with the most recent Win32 version of PDCurses, because it now returns shift, alt and control immediately as keystrokes. (XCurses presumably had the same problem, but I didn’t test this.) This is worked around now.
SOUP (and BW v4): Newsgroup lines which exceed the width of the display now end in “…” to indicate a continuation. (You can see the entire “Newsgroups:” line by pressing ‘X’ for full headers.)
SOUP (and BW v4?): Broken “References:” lines which contained no proper message ID’s were causing segfaults. Reported by Peter Krefting. BTW, would the perpetrators of these broken messages please fix them?
SOUP: In ‘m’ mode, checking for “From “-line dividers is a little more rigorous; the line must now also contain an ‘@’. This eliminated all the false positives I was seeing on Pine-generated mail folders.
In packets where the list of subscribed areas is not known (e.g., all plain QWK and SOUP packets), the “Subscribed” mode in the area list is now skipped automatically.
The default color scheme is now a little less day-glo, :-) and shows up better on some terminals. The old one is still available in tradit.col. Let me know what you think.
Ncurses’ built-in SIGWINCH handler is on by default in 5.0+; this was conflicting with MultiMail’s handler. It should now work OK, but see NCURSES_SIGWINCH in interfac.h if you’re not using ncurses’ defaults.
When going backwards through messages via ‘-‘ and backing out of an area, the display was not updated correctly. I’ve also changed the behavior of ‘-‘ in this instance to go back to the last non-empty group instead of the last group (more consistent with ‘+’).
Moved web site and email from ClarkNet to SourceForge. Among other benefits, this means that old versions of the MultiMail source code are now archived online, going back to 0.1. Binaries (for certain platforms) from 0.36 on will also be archived.
New compression type: tar/gz. (New .mmailrc keywords: “tarCompressCommand” and “tarUncompressCommand”.) Unlike the other types, this recompresses the entire packet when updating the .red flags, so it can be a bit slow. Also, the supplied command lines assume GNU tar, which has gzip built-in. Separated gunzip/tar and tar/gzip command lines would require a (simple) external script. MultiMail only checks for the gzip signature. Requested by Peter Krefting for compatibility with GNUS.
The packet list could crash on directory names longer than 32 characters. Reported by Tuukka Lehtinen.
Zero-length messages caused a segfault. Reported by T.L.
In SOUP packets with index files, the .idx file would be opened instead of the corresponding .msg file. Reported (indirectly) by Peter Krefting.
GNUS apparently uses a bogus identifier ‘n’ in the first position of the attribute fields of the AREA files in the SOUP packets it generates, equivalent to the standard type ‘u’. MultiMail now recognizes this flag. Reported by P.K.
In MakeChain(), rejoin is now always applied to non-quoted lines. This looks better in most cases; in particular, where space padding on the end of a line resulted in a line of exactly 80 columns. (Previous versions of MultiMail would print an extra blank line then.)
relist() is now always called from MakeActive() in the area list and little area list. This means that the list is updated properly in Active mode, among other things.
In the packet list search function, the wrong packet was highlighted.
On some platforms (including the NetBSD m68k binary distributed on the web site), forced word-alignment of structs broke the “old-style” (.PDQ-based) offline config in Blue Wave mode. (Yes, I should’ve anticipated this after the similar bug with OPX mode in 0.32.)
The signature-adding routine will now skip CR characters, so (in principle) a .sig could be shared between Unix and DOSish readers. Other text files should still be kept separate. Inspired by Martin Prieto.
The backslashes that had briefly appeared in the DOSish versions’ packet windows had reverted to forward slashes again in 0.36. :-)
The area list has a new mode, “Active”, to accompany the existing “All” and “Subscribed” views. The Active mode lists only those areas that contain messages. The L key now cycles among all three modes. (Note that in some cases, two modes may be equivalent; e.g., in plain QWK packets, there’s no way to distinguish Active from Subscribed.)
There’s also a new “Marked” view in the letter list. If any messages in the area are marked, the L key cycles between all, unread, and marked views. (Otherwise, it toggles all/unread, as before.) Marked messages are no longer automatically included in the unread view.
In the packet and letter lists, the current sort modes are now displayed. And in the area and letter lists, the current mode is displayed. The number of files is shown in the packet list, but only if the total number of items (files + directories) exceeds the window length. The area and letter lists now work similarly, showing the number of items only when they exceed the window length.
By popular demand, the default modes for the area list (all, subscribed, or active) and the letter list (all or unread) can now be set in .mmailrc. See the “AreaMode” and “LetterMode” keywords.
Old-style Blue Wave offline config, using .PDQ files, is now supported. (.UPI/.NET replies still aren’t, yet.) Requested by Russell Tiedt.
Specifying a directory on the command line now opens the packet window on that directory. No more items are read from the command line afterwards.
In Unix, the “~” character is now recognized as a synonym for the contents of the “HOME” environment variable in input, and is also used that way in the packet list display. (However, “~username” is not supported.)
The Tab key works as an alternative to Enter in input windows.
The BCC port is now the “standard” Win32 port. (The RSX/NT port is still available, renamed mmrsxNNN.zip.)
New BeOS binary.
Makefile.tcc and Makefile.bcc are now sensible – no more renaming or moving kludges. :-) Also, the “UNUSED_PARAMS” nonsense is gone, and MultiMail compiles without warnings under BCC.
The message and ansi viewers can now handle > 64k in the 16-bit version, depending on available memory.
In myreaddir(), the DJGPP, Turbo and Borland ports (mmdos, mmxt, and mmwin) now use findfirst()/findnext() instead of readdir()/stat(). This means that the packet window opens significantly faster on large directories. (stat() is fine in Unix, but slow in MS-DOS, where it’s implemented via findfirst().) Also, the list is now built in a single pass, instead of one pass for files and another for directories, resulting in a speedup on all platforms.
Test of read-only status now works in Borland/Turbo ports.
Grabbing of Fidonet addresses was broken in some situations, causing segfaults. Reported by Greg Paski.
When lines ended with spaces, they could be quoted incorrectly. Reported by Jim Hanoian.
Tab characters in messages could cause segfaults in some cases.
Soft CR stripping wasn’t working.
Looking through my (very) old packets, I found some in which another reader (?) had added its own files to a packet with the names “a0000.ndx” and “a0001.ndx”, which made MultiMail generate a bogus area 0 for those packets. Fixed.
The keypad mostly didn’t work in the BCC (Win32) port of MultiMail. Reported by Jeff Foy.
Using the Goto function in the packet list now puts the highlight bar on the first file, as when a directory is selected by pressing Enter. In Lynx mode, going left (back) in the packet list now actually goes to the previous directory, instead of the parent directory, and the previous highlight position is restored.
When cycling through sort modes in the letter list, sort by “To:” is now skipped for areas which have no such field (i.e., Usenet areas).
Wrapped header lines in SOUP are now displayed in the correct color.
When writing out offline config (add/drop), the area list was scanned, but the current area number was not preserved. This sometimes led to minor cosmetic problems in the reply area. Reported by J.H.
Finally replaced the ugly #define list in mmcolor.h with an enum.
Some malformed Internet “From:” lines appear without a space between the name and address. In such cases, MultiMail was stripping the last letter of the real name.
Several new error messages.
The color files are now archived separately within the binary archives, to reduce clutter. Now if I can just figure out how to combine some of the documentation files in a sensible way…
Two new ports officially supported: An alternate Win32 port, via Borland C++ 5.5 (available free at the Borland web site); and, more interestingly, a 16-bit MS-DOS port, via Turbo C++ 3.0 (an ancestor of Borland C++ 5.5). I released an experimental version of the latter a couple weeks ago, but the official 0.34 version is much better, mainly in reducing memory usage and in not (always) giving up when memory runs out. :-)
Areas that have replies are now always included in the short list, which prevents the replies being misdirected on re-edit. Such areas are also now flagged with an ‘R’ in the main area list.
The “References:” line disappeared from SOUP replies if they were re-edited before posting. Reported by J.B. Moreno.
Massive reductions in memory usage for the LetterWindow and AnsiWindow classes. The plain message text takes up about half as much space as before; the ANSI viewer savings vary, but can be even greater.
The environmental variable “TEMP” now works as an alternative to “TMP” with the DJGPP port. If neither of those is defined, nor “DJGPP”, it will now use the startup directory instead of “C:".
TEMP or TMP values ending in a slash or backslash now work correctly in the Win32, OS/2 and 16-bit MS/DOS ports. Also, all these ports now use the startup directory rather than the MMAIL directory as the default location for the temporary directories and files, in the absence of a TEMP or TMP environment variable.
SOUP replies sometimes had the wrong length in the length field (one byte too long), resulting in spurious characters at the ends of messages.
Skip trying to add to read-only files. (But this check doesn’t work yet in the Borland/Turbo ports.)
References to “interface” in the code changed to “ui”. (Apparently, “interface” is a reserved word in Borland C++.)
Soft CR stripping moved to the Blue Wave class. Let me know if this is a problem. (I haven’t seen them in QWK or OMEN packets, and they’ll never appear in SOUP.)
Various code cleanup. Now 16-bit clean, among other things. :-)
Support for encoding and decoding Quoted-Printable headers and bodies, primarily in SOUP packets. Decoding is always performed; encoding is performed subject to the new .mmailrc keywords “outCharSet”, “UseQPHead”, “UseQPMailHead”, “UseQPNewsHead”, “UseQPMail” and “UseQPNews”. Thanks to Peter Krefting for some of the code.
Support for the OPX format has been removed. I hope this will be only temporary. The reason is not technical. If you’re affected by this, and you’d like to continue using OPX, please email me privately.
Support for individual packet types may now be disabled when compiling, in order to save a little space in the executable. The packet modules to include are selected in the “modules” file.
The Rename function in the packet list now brings up a Blue Wave-style default filename, with a numeric extension based on the last one found plus one, unless the packet already has a numeric extension. Also, rename failures are reported, and you can no longer rename to the name of an existing file.
Some extra pop-up notices like “Opening…”, mainly for the benefit of slower systems. :-) There are also additional non-fatal error messages in several places, as when a Save fails.
.XTI support is now fully segregated into the Blue Wave module. This is cleaner, but it means that .XTI files are only recognized in Blue Wave packets, and not in other types. The default for the “MakeOldFlags” option in .mmailrc has been changed to “Yes”.
The automatic prefixing of “Re: “ on the subjects of replies can now be disabled, for non-Internet areas only, via the “ReOnReplies:” .mmailrc keyword. By popular demand. :-)
New color scheme contributed by Tuukka Lehtinen.
Before the removal of the OPX module, these features were added:
In OPX mode, MAIL.FDX is now handled in the same way as .XTI in Blue Wave mode. It’s written instead of .red when “MakeOldFlags” is set.
Offline config in OPX mode. Thanks to Armando Ramos for the format.
“To:” and “From:” kludge lines in text are now checked only in QWKE packets, instead of all QWK packets. This should help with some systems that require “To:” lines for Internet addressing.
Blue Wave anonymous areas are now recognized as alias areas, as in QWKE.
Areas flagged as read-only will no longer accept replies, and will pop up an error box instead. This works in Blue Wave, QWKE, and OMEN. Requested by Jaakko Lintula.
The MS-DOS version sometimes segfaulted due to bugs in DJGPP’s qsort(). Fixed by upgrading to a newer DJGPP library. Reported by Laird Kelly.
When reading the .mmailrc file, unrecognized keywords are now reported. May help to catch typos. :-)
In OMEN mode, when the UserName was not set in .mmailrc, all messages were being flagged as personal.
Defaults for “AutoSaveReplies” and “AutoSaveRead” changed to “Yes”.
Compiling on an Alpha revealed several bugs that didn’t show up on systems where pointers and ints were the same size (i.e., all the 32-bit systems); mainly inappropriate usage of resource::get() vs. ::getInt(). MultiMail is now 64-bit ready. :-) Thanks to Compaq for the free account.
In Blue Wave mode, the echotag field is now used as the area description if the description field was left blank. Problem reported by Greg Mayman.
Internet address parsing modified to deal with badly-formed addresses like “(Name) <address>”. Problem reported by Dane Beko.
There were intermittent segfaults in the letter window with screens wider than 80 columns, due to an off-by-one error in memory allocation.
The following fixes were made to the OPX module prior to its removal:
Some versions of the Silver Xpress door eat the last character of the tear line. MultiMail now pads it out with a sacrificial space in OPX mode. Reported by Jim Hanoian.
Some OPX packets split their areas into BRDINFO.DAT and EXTAREAS.DAT; MultiMail should now be able to handle this (untested). Reported by Armando Ramos.
On some platforms (e.g., the NetBSD version distributed on my site), OPX support was not working due to structure alignment padding. Fixed.
New line ending recognized: plain CR. (This is in addition to the LF, CRLF, and 0x8D endings recognized in 0.32.) If you got messages with blank bodies, or immediate segfaults when opening messages, this will probably fix it. Reported by Laird Kelly.
Certain packets (with messages over 64k?) reportedly had incorrect values in the length fields of some messages, causing everything after that point to become unreadable. This is fixable by using MAIL.FDX to index the messages, which MultiMail now does. Reported by Armando Ramos.
Support for the OPX packet format. This was reverse-engineered, and has really only been tested with the WINS door. See mmail/opxstrct.h in the source code archive for the specs.
A “Personal” area (messages addressed to you) is now created automatically for all packet types except SOUP. This can be turned on or off (default is on) via the “BuildPersArea” option in the .mmailrc – except in QWK packets with .NDX files, where the presence or absence of a “personal.ndx” file determines whether the area is generated (the same as in older versions of MultiMail).
With Blue Wave packets (and only with Blue Wave packets), MultiMail now has the option (off by default) of storing its last-read pointers in an “.xti” file, as used by the Blue Wave reader and some compatible programs (like BWSave), instead of in MultiMail’s native “.red” format. This is toggled by the “MakeOldFlags” option in the .mmailrc. (I made the name generic in anticipation of support for other such formats, like OPX’s “mail.fdx”.) In reading, .red no longer takes precedence over .xti; if both files are present, the one with the later date is used.
Character set translation now works correctly on Linux text consoles that use Unicode-mapped fonts. Problem reported by Peter Krefting. Also, OMEN replies to Latin-1 packets were not recognized as being Latin-1 on rereading; initial printing of translated fields in the header editor was incorrect; and the ANSI viewer always assumed CP437.
Forwarded messages now add only the changed header lines to the body of the text, instead of using the same format as saved messages.
Date in SOUP replies was being localized, which is undesirable. Reported by P.K.
The Personal area now always shows the correct number of unread messages in the area list.
Blank lines are now added automatically before signatures, and the sigdash (“– “) in Internet and Usenet areas is correctly followed by a newline before a sig. (Previously, it worked right with taglines, but not sigs.)
Areas with a very long description could cause a segfault when opening the letter list. (They were truncated, but not quite enough since the addition of the message count.) Also, the header of the letter list now takes advantage of the full width of larger screens.
Can now read net-status QWK packets correctly, even without .NDX files. (End users shouldn’t really be reading net-status QWK packets; but judging by a packet I was sent, some are.)
Empty subject lines are now forced to the end of the letter list when sorting by subject, instead of messing up the sorting as they were doing.
Precautionary fix for hypothetical corrupted Y2K date fields in Blue Wave packets. Hopefully not needed. :-)
The User-Agent line in SOUP mode now includes “SOUP” in the comment. :-)
Text files (e.g., MMAIL.RC, or the DOOR.ID or CONTROL.DAT from a QWK packet) which did not end in a line feed character would cause MultiMail to lock up. This was due to a new line-reading function introduced in version 0.30, which fixed some other problems, but stupidly added this one. Due to the seriousness of this bug, I’m releasing 0.31 early. Reported by David Pratt et al.
Searching did not work in the little area list. (This dates to 0.29, though I just noticed it.)
In SOUP packets, extra spaces between header keywords and values are now ignored.
Support for SOUP packets. Note that type ‘M’ areas and index files are not supported yet; only ‘m’, ‘b’ and ‘B’.
Support for OMEN packets.
The new keywords “UserName” and “InetAddr” are used to build a default
“From:” line for SOUP replies, of the form “UserName
Replies in Usenet areas may now be cross-posted; the newsgroups list can be edited.
Quote headers can now be set in the config file (separately for Internet/ Usenet areas, and others), via the “QuoteHead” and “InetQuote” keywords. The maximum width for quoted material can also be set, via “QuoteWrapCols” (suggested by Holger Granholm). See the man page for details.
The “Lines:” indicator now includes a percentage.
Long lines in text files would sometimes cause the remainder of the file to be read incorrectly. Mainly, this could show up with the tagline file. Reported by Dane Beko. Also, blank lines in the tagline file are now ignored.
Area 0, if present, was not being sorted in the letter list.
In the MS-DOS version, the first shell-out would trash the old prompt, so second and subsequent shells had a bogus prompt. Reported by Jerel Arbaugh.
When re-editing a reply and shortening it, the remaining text could end up off the screen when returning from the editor. Reported by Jim Hanoian. Also, a similar effect could be briefly seen when editing a reply from the original, if the reply was shorter.
Text entry fields did not work correctly when backspacing if the text filled the field. Reported by Tom Rutherford.
Changed default zip commands – removed “-o” (for both PKZIP and InfoZip) and added “-k” for InfoZip. The latter option forces stored pathnames to an MS-DOS compatible 8.3 unicase format, which ensures maximal compatibility with doors and such. You may want to add this yourself if you’re upgrading; but note that it could result in two “.red” files being stored into packets that had been read without this option.
Saved messages can now have longer header lines, and the headers are more appropriate (closer to what you see in the letter window) for Internet/ Usenet areas.
Signatures and/or tearlines in Internet/Usenet replies are now prefaced with “– “, and “… “ is omitted. Signatures are recognized in text in these areas, are colored with the tagline color (I’ll make this a separate color later), and are not quoted when replying.
Long subject lines and other long lines can now be fully edited, regardless of screen width, and will be fully displayed if the screen is wide enough.
In Internet/Usenet areas in Blue Wave packets, the subject, date and newsgroups are now extracted from the kludge lines in the text, as with references and message-id. (Most of the changes in this version of MultiMail are for enhanced Internet/Usenet support; although this was primarily for the benefit of SOUP, it carries over into Blue Wave.)
TAB characters are now decoded into spaces in MakeChain(). This prevents them from pushing lines off the screen. But to save a message with its TAB characters intact, you have to save it from the ANSI viewer.
Strip any number of leading “Re:”’s, not just the first.
Filenames within reply packets are now created in upper case, even if you don’t use the “-k” option (see above). Mainly this was for compatibility with the MBSE BBS, though there may be others with the same problem. (The format specs generally show the filenames in upper case, so this is arguably the correct behavior for a case-sensitive filesystem; but most were designed with the FAT filesystem in mind, and only SOUP explicitly addresses the issue of case.)
“Re: “ will not be prepended if it would make the subject exceed the maximum width for the format. Also, non-QWKE QWK packets now have their subjects limited to 25 characters (again); the “Subject:” kludge will not be added except in QWKE packets. (I’ll add it back for some later, when I can identify the particular door’s capabilities better.)
’:’ and ‘|’ are now recognized as quote indicators, but only when appearing in the first column. (But if ‘:’ is the start of a smiley, it’s ignored.)
When quoting in Internet/Usenet areas, blank lines now have a quote marker prepended, as this seems to be the standard.
Intitial check for 80x20 instead of 60x20, error if screen too small.
I made a (temporary) patch to PDCurses for the MS-DOS version, so that MultiMail will work correctly in non-standard video modes. Problem reported by J.A.
Overdue documentation of the .mmailrc file in the man page.
List windows now have a “scroll bar” when the number of items in the list exceeds the number that will fit onscreen. This feature is selectable via the “UseScrollBars” keyword in .mmailrc (on by default). Note that it’s not (yet) mouseable.
The new “UseLynxNav” option (off by default) changes the function of the left and right cursor keys: left backs out, and right selects. (It’s named after the Lynx web browser. The idea comes from Pine, which has a similar option.) It also makes the left arrow equivalent to selecting “..” in the packet list, except on the top level (where it’s equivalent to Quit).
Fido “SEEN-BY:” lines are now marked as hidden if they weren’t already. (This allows you to turn on the “extended info” option in Blue Wave doors below ver. 4 with impunity.)
The Delete key will work as an alias for ‘K’ in all cases.
Major bug: In certain cases, the little area list would not set the correct area, and the reply would be misdirected. Reported by Jerel D. Arbaugh.
Outside of the packet/area/letter list, the search routine could abort in some cases (i.e., ‘/’ would simply not work). Reported by Jim Hanoian.
If a message was saved (from the letter list) before any messages had been viewed, it could be saved rot13’d!
The “soft CR” filter was not working if the console character set was Latin-1.
Re-editing of replies from the original area could move the letter list selection inappropriately (to the corresponding reply number).
If the last-opened packet was deleted and replaced by another packet of the same name, the old packet would be “reopened” even though it was gone. Reported by Raymond Cool.
Lines which contained a ‘^A’ character, but not at the beginning of the line, were erroneously being marked as hidden.
“Re:” is now left visible in the letter window, though still stripped from the letter list. (In this, I emulate Tin.)
During a search, the beep-on-personal would beep on each personal message scanned, even when the corresponding message was not actually displayed.
When shelling out, the prompt string could be truncated if it was excessively long. Not really a bug – I just hadn’t believed that anyone would actually have a prompt string that long. :-) Reported by J.D.A.
Due to its placement at the start of a line, an instance of the word “.mmailrc” was interpreted as a control code and disappeared from the formatted man page. Reported by Geoffrey Wilson.
Offline config! (Add/Drop only, at the moment.) For Blue Wave, QWKE, and QWK (with DOOR.ID indicating Add/Drop support). See the man page for details.
The area list window now has additional status flags: ‘+’ and ‘-‘ for added/dropped, and ‘’ in the full list for subscribed areas. The little area list has ‘’ also, and is now wider.
A more advanced search routine. It will now (optionally) descend levels, allowing even a full text search over multiple packets.
The address book now allows direct entry, as well as editing of existing addresses. The tagline window allows editing of existing taglines. (The old function of the ‘E’ key in the tagline window has been moved to the ‘A’ key.)
Reply splitting, both manual and (optionally) automatic.
Quick exit from any screen via ^X.
Command shell (for DOSish ports) via ^Z.
Packets can now be specified directly on the command line, bypassing the packet menu. (If multiple packets are listed, they will be opened in sequence.)
Better quote wrapping – paragraphs are rewrapped, and second-level quotes are not requoted (except in Internet/Usenet areas, which use a reduced quote for second-level quotes).
Optional stripping of “soft carriage returns” (those annoying accented i’s that sometimes appear).
Optional beep in letter window when opening personal messages.
The “Yes/No” prompt to keep old reply packets has been replaced by “Save/Kill”, to reduce the chances of deleting a packet by mistake (as can happen when one is in the habit of hitting ‘N’ in response to the new files or bulletins prompts). Suggested by several users.
In all string input windows (those that use ShadowedWin::getstring()), a backspace as the first character will now preserve the default text (if any). I’m not sure if it’s better this way or not, though.
The tagline window was displayed improperly on a screen wider than 80 columns when scrolled up or down a line at a time.
I’ve changed the message MultiMail shows the first time a new version is run, to something I hope is clearer. Please note that MultiMail will never wipe out the values from your old config file, only update the file with new keywords (if any).
When changing to a new directory (or on startup), the first file rather than the first directory is highlighted.
When composing a new message and not quoting, MultiMail no longer creates a 0-byte temp file before launching the editor. (You probably won’t notice the difference, but I think it may address one bug report I got a while back. For me, it makes pico say “New file” instead of “Read 0 line”.)
Some more egcs warnings fixed. (Odd; they should have appeared earlier.)
Internal reorganization: Most non-ANSI, non-curses code is now in mysystem.cc. I hope this may simplify porting to non-POSIX systems. Also, part of interfac.cc was split off into basic.cc; and some stupid #define and const int lists were replaced with enums.
In QWK reply packets, the binary conference number was not being included. This made them incompatible with some doors (at least some versions of Galacticomm’s software, and probably EzyBBS). The problem goes back to MultiMail’s beginnings. I didn’t notice it before now because most QWK doors use the ASCII conference number instead. (There are redundant fields in a REP packet.) Thanks to SparkAm@peyam.net for tracing the problem.
The precompiled MS-DOS binary of 0.26 wouldn’t work on machines without FPUs, due to the FPU code linked in with the difftime() function in the packet list. I’ve replaced this with an integral subtraction, which is adequate for my purposes. Reported by Greg Mayman et al.
curs_set() is now called from Win::cursor_on() and Win::cursor_off(), regardless of the curses implementation. (Formerly, these lines were ifdef’ed.) This is for compatibility with ncurses 5.0-beta1, which no longer calls it from leaveok(). I’m told that Solaris curses doesn’t, either.
A netmail address on a message would be retained if the message were forwarded to a non-netmail area, and it would not be editable. Reported by Jim Hanoian.
The packet list window length is now variable for lists shorter than the max, as with the letter list. The “info” window in the area list has been merged into the main window. (The visible effect of this is that the lower window no longer has a title.)
Some internal reorganization.
Automatic saving of replies is now available via the “AutoSaveReplies” keyword in .mmailrc. When this is enabled, MultiMail automatically rewrites the reply packet after a reply is created, edited or killed. It’s similar to pressing “F2” or “!” after each event, but without the warning window. I’m considering making this the default behavior for future versions – any comments?
The “AutoSaveRead” keyword turns on automatic saving of the last read markers when exiting a packet, bypassing the “Save lastread” prompt. (This is another thing I might make the default.)
Packets may now be renamed from the packet list window by pressing ‘R’. (This function can also be used to move them to different directories, though you must specify the filename as well as the pathname of the destination.)
Directories are now shown in the packet list, and are navigable; or you can specify a directory to jump to via the ‘G’ command. (On the DOSish ports, the ‘G’ command is the only way to change drive letters.) The current directory appears as the title of the packet list window. Note that this change means that the old behavior of exiting when no packets were found will effectively never occur.
By popular request, multiple replies to the same message (in the same reply packet) are now allowed. (This is a partial reversal of the change in 0.24.) When re-replying, you’re prompted to re-edit; if you say no, a new reply is generated. Only the first of the existing replies will be recognized for re-editing.
When opening a packet, you’re now prompted whether or not to keep any existing replies.
The tagline window can now be completely disabled via the “UseTaglines” keyword in .mmailrc. Requested by Jack Pfisterer.
Re-editing replies from the reply area no longer forces a return to the letter list. Instead, the re-edited letter is displayed.
Cosmetic problems with the wrong area being displayed on the status line (and the header window) during the reply process have been cleared up.
The “This will overwrite any existing reply packet” warning from the F2 function is now suppressed if there is no packet.
MultiMail will now display a warning if the reply packet is not created, or if the last read markers could not be saved. Also, it will pause for two seconds after getting an error code from a program it shelled to (archiver or editor) or attempted to shell to, so that any console error messages from the system may (hopefully) be read.
Speculative fix for reported problems with replies disappearing (which I couldn’t reproduce): calls to clearDirectory() now pass the name of the working directory explicitly, rather than “.”.
Added “-e” to the default add command for ARJ to suppress pathname inclusion. Suggested by Ken Whiton.
The separate colors for the Search window have been eliminated, as the Save dialog is now used for both (as well as for the new prompts in the packet list).
The “F2, !: Make reply packet” function is now available in the letter list and letter window, as well as the area window.
Shorter date format in the packet list; more space for long filenames.
On Blue Wave messages with no taglines, a blank line is now inserted before the origin line.
Fixed errors (and some warnings besides) that made MultiMail 0.25 incompatible with egcs. Thanks to Tony Summerfelt for reporting the problems and testing (and re-testing) the solutions.
If there was a tagline in the message on screen, and you chose not to use a tagline in the reply, the tagline from the original would be used instead. Reported by David Toutant and J.F. (With the reorganized tagline code, I’ve also reverted to having the tagline grabber pick the last tagline in a message, instead of the first.)
In Blue Wave mode, when the packet filename differed from the “base name” (BBS id), the former was used for replies, when it should be the latter. Reported by Jim Hanoian.
After animating ANSI, MultiMail would loop on a non-blocking keyboard check, sucking up CPU.
As usual, lots of internal reorganization.
Support for Internet and Usenet messages in Blue Wave mode. (Note: AFAIK, this is only supported by the Blue Wave version 4 door for PCBoard.) Also, Internet addresses can now be taken into the address book.
Character sets (either code page 437 or Latin-1) are now selected for each area or message. The default character set is CP437 (as before), except for Internet or Usenet areas (as marked by Blue Wave or QWKE flags), which are Latin-1. The character set in individual Fidonet messages can be overridden by the “CHRS:” or “CHARSET:” kludges. Replies are set according to the default for the destination area.
The console character set and the default sort types for the packet and letter lists can now be set in the config file. The default packet sort is now by time (latest first) instead of by name.
The “To” field is now omitted from the letter list in Usenet areas and the QWK Personal area. The “From” field is omitted in the REPLY area. This leaves more room for the “Subject” lines by eliminating redundant info.
Direct support for RAR archives. Note: there’s a problem with this if you read the same packets in, e.g., both Linux and DOS: RAR will save the last read marker file twice (once in upper case, once in lower).
More ANSI navigation: The ENTER key now works the same in the ANSI viewer as in the letter window (advance to next message).
The packet list can now be rescanned (to pick up new packets) by pressing ‘U’.
The message count now appears in the title bar of the letter list (like the area count in the area list).
New color scheme by Gary Gilmore.
The alternate QWK private message flag, ‘+’, was not being recognized. Reported by Tom Rutherford.
Changed the message that prompts to save replies to something I hope is clearer. Suggested by Max Chamberlain.
If the reply packet had a different archive type than the main packet, MultiMail would attempt to use that archiver (the reply packet’s) to add the last-read markers to the main packet. The correct behavior is to use the main packet’s archive type when remaking the reply packet!
When quoted material was wrapped, the wrapped lines would have trailing spaces. These are now stripped.
In the area and tagline lists, a completely blank line would sometimes appear at the top of the list. (A long-standing bug, finally fixed.)
Strip blank lines after hidden lines. (Blue Wave puts blanks after the hidden headers in Usenet messages.)
Fewer keys to hit in the header editor. :-) The cursor is now placed on either the Subject or To line, as appropriate (you can still arrow up to edit the other lines); and the “Make message private?” prompt is not presented when sending to areas that are either all-public or all-private (as indicated by Blue Wave or QWKE flags).
Bizarre alignment bug in the EMX-based ports (OS/2 and Win32) caused crashes on some QWK packets. Reported by Jim Hanoian.
Source code reorganized somewhat… no more .a files.
Blue Wave packets could have identical areanums in multiple areas. Although this is an error, the Blue Wave reader handled it; now MultiMail does as well. This is also a little faster. However, it assumes that the mixRecords and INF area headers are sorted the same way, which I would’ve preferred not to assume. Problem reported by Scott Jones.
In 0.24, the new packet kill routine failed to exit after killing the last packet, leaving garbage on screen. Reported by Gary Gilmore.
A filename without an extension would mess up the display in the packet list. A long-standing bug that I never really cared about, but which J.H. mentioned.
All directories are now omitted from the packet list.
One more attempted cursor fix for the Win32 version.
Minor display bug with messages 100,000 lines or longer. (Now OK up to 1,000,000.)
In the EMX ports, in the letter sort by subject, the secondary sort by message number was still broken. Reported by J.H.
“Save” function in ANSI mode. This differs from the letter window save in that no wrapping is performed, and no header is prepended. This is also the only way to save bulletins and new file lists.
New navigation in ANSI mode: the space bar and left and right arrow keys now work in the same way as in the letter window. This also means the interface for the bulletin viewer is slightly different: ‘Q’ exits from the entire list, instead of just the current bulletin, and you can move back and forth between bulletins as in the letter window.
Multiple sort modes in the packet list: by date (latest first), and by name. By name is still the default. “$” toggles between them.
Multiple sort modes in the letter list: by subject (still the default), message number (i.e., no sort), from, or to. “$” cycles through them.
The help window at the bottom of the packet, area and letter lists can now show multiple pages of options (similar to Pine’s “Other commands” menu function).
When character set translation is on in the ANSI viewer, some IBM characters are now mapped to curses equivalents, giving a better result on non-PC terminals (e.g., xterm, vt100). This works best with the box- drawing characters. The letter window still uses a plain ASCII mapping for these characters.
Pop-up help is now available in the ANSI viewer, as in the letter window.
Messages can be forwarded from any area, with the original headers preserved in the text.
Support for the Win32 console via RSX/NT. See README.rsx if you want to compile for this platform.
Fidonet MSGID/REPLY kludge lines are now supported, in Blue Wave mode.
Additional QWKE support – in QWKE packets, the short list now shows “subscribed-to” conferences, as in Blue Wave mode, instead of just the non-empty ones; and alias areas are supported. The “QWKE” type appears in the Info window of the area list.
When Replying to a message that’s already marked “Replied”, the previous reply (if available) will be re-edited instead of a new reply being created.
Some new color schemes.
The help window now survives a screen resize, instead of being closed.
Killing a packet no longer forces a reread of the directory.
The addressbook will no longer grab a new entry when no Fido address is available. Also, the addresses are now sorted alphabetically on startup. (Anything added during the session will not be alphabetized; however, the list will be sorted again on the next startup.)
The full length of Blue Wave “To” and “From” fields is now available in the header editor. (Previously, they were restricted to 25 characters.)
The subject sort in the letter list was not correctly performing the secondary sort, by message number, in runs of the same subject.
Time/date printout code changed in packet list and in Blue Wave reply headers. I think this looks a bit better.
The experimental Win32 console version had a problem with slow screen output. This seems to be a problem with the console API (?), but it’s worked around in this version. (Normally, PDCurses checks for a keypress after printing each line. In Win32, this call takes a noticeable amount of time to return, even though it’s supposed to be non-blocking. So, in MultiMail, the check is disabled for the Win32 version. This also disables typeahead, but I doubt anyone will even notice that.)
A last attempt at fixing the remaining cursor problems in the PDCurses versions. (The cursor size is now checked at startup, and that value is explictly restored instead of using curs_set(1).) Scott Jones reported the persiting problem in the OS/2 version.
The index of the Personal area is now updated just by entering the letter list, or anytime a message is read in that area. Better than previous versions, though still not ideal.
Support for the keypad plus, minus, and slash keys under PDCurses. (This completes the keypad “Enter” fix in the previous version.)
The experimental Win32 version supported only “" in the pathnames specified in MMAIL.RC, where the DOS and OS/2 versions allowed either “" or “/”. Now both work in Win32, and both will be converted to “" when MMAIL.RC is updated. Reported by Rafael Cresci.
One more character allowed for unames in Blue Wave reply packets (six instead of five). Specifically, that means “MultiMail/MS-DOS” and “MultiMail/NetBSD” will now appear in Blue Wave tearlines, instead of the abbreviated form “MultiMail”.
Hidden lines at the end of a message would cause the line count to be over by one. Reported by Jim Hanoian.
In the Save dialog, user-entered names were getting spaces converted to underscores, though only the second time through (when the name was pulled up as the default). Reported by J.H.
QWKE and PCBoard-style “Subject:”, “To:”, and “From:” line kludges are now supported in QWK mode, along with “Title:”, which is used on WWIV. The QWKE “Subject:” line is generated automatically for replies, if needed. Note that in regular messages (though not in replies), these lines aren’t parsed until the message body is accessed by entering the letter window, so the fields as shown in the letter list will at first be the short forms.
Reply forwarding – press <CTRL>-‘F’ while in the reply area (either the letter list or letter window) to forward a reply. (Forwarding of regular messages is not available yet.)
In the REPLY area, ‘R’ now works as an alias for ‘E’. Added at the insistence of Jim Hanoian. :-)
When (re-)editing or forwarding a reply, the area can now be changed.
REPLY and PERSONAL areas now show the original area (as well as the collection area) in the letter window. Suggested by J.H.
New behavior in the ANSI viewer: Instead of resetting everything when a
screen-clearing code (
Characters 8 (backspace) and 12 (form feed) are now handled in the ANSI viewer, and skipped in the letter window. Ctrl-Z is stripped in the ANSI viewer.
The BBS name and Sysop name fields in the info window of the area list were being unnecessarily truncated.
In the PDCurses versions, the cursor remained disabled when shelling to an editor. Reported by Dane Beko. (Due to the way I implemented this fix, PDCurses 2.2 is no longer acceptable for MultiMail; you must use 2.3.)
When creating a .REP, a zero-length message body would cause a segfault.
Blank space at the end of the little area list is now omitted (as in the letter list).
An attempt to send Netmail when no Netmail area is defined (as is always the case in QWK mode) now pops up a nonFatalError window.
The “Save lastread pointers?” message now comes up less often; the “any read” flag is set less aggressively. (If you re-read a message that’s already marked as read, the any-read flag will not be set, as it was in previous versions.)
When entering areas which are fully read, the highlight bar is no longer moved to the last message. Similarly, if the entire packet is marked as read, the active area is not set to the last non-empty area, but the first. Suggested by J.H.
Hopefully, a better subject-line sort in the letter list: If two subjects have the same content for the entire length of the shorter of the two, the shorter one is no longer automatically placed first; instead, they’re put in message number order.
The ENTER key on the numeric keypad did not work in the PDCurses versions. Reported by Russell Tiedt.
Added “/m” to the parameters passed to LHA in the MSDOS version. Without this, it gave an “Extension is not .lha. Continue? [Y/N]” prompt when archiving replies.
The long-awaited support for new file lists and bulletins! They’re displayed automatically on opening the packet. Currently, this is a bit of a kludge – although I’m starting to like it. Memory usage is excessive. (See the man page for more info.)
Support for XCurses (the X port of PDCurses). See README.xc for details.
Added a line counter to the ANSI window, as in the letter window.
If a tagline was displayed on screen, it would replace the tagline selected for a reply via the tagline editor. This bug dates to 0.20.
TAB characters that advanced past the end of a line were causing problems in the ANSI viewer.
Aborting a netmail reply left the netmail address set when a regular reply was attempted afterwards.
The translation toggle (‘c’) now works in all screens. (In particular, the ANSI viewer.)
Proper printing of character 127 (DEL) under ncurses.
The SIGWINCH (resize) handler now works acceptably even when a SIGWINCH occurs while in a WarningWindow() or getstring() call.
The code to generate the default tagline file was broken in 0.20, which meant that all first-time users got a segfault. Argh! Also, since I was messing with the taglines anyway, I changed the default list.
In the PDCurses versions, the “Could not uncompress packet” and “Packet type not recognized” messages would pop up without first restoring the packet list screen. Also, some extra parentheses have been added around color names for the benefit of PDCurses/Win32 (still an incomplete port).
Updated the README.{DOS,OS2} files to reflect the new style of default MMAIL.RC in use since 0.19.
”/” and “.” added to pop-up help window.
Search function. Case-insensitve searching is available everywhere. It’s still a bit limited, in that it only works within the current list (or letter); i.e., you can’t yet search multiple letters at once. Hit ‘/’ to bring up the text entry window and start the search, or ‘.’ to repeat the last search. (Yeah, I know – it needs better keys.) Searching is from the current line on down.
Fido “hidden” message lines (marked with a ^A) can be displayed; ‘X’ toggles this function.
New features in the tagline editor and address book: New entries in the address book are checked for dupes; also, tagline dupes, which were already checked, are now reported. The address book and the tagline editor allow you to Kill entries. And the tagline editor can be brought up directly, via ^T, in the same way as the address book. Old features: The reLoad function has been removed from the tagline editor.
Rot13 function in letter window – toggle with ‘D’.
If a new letter is not edited (as measured by the time stamp), you’ll be asked if you want to cancel it on returning to MultiMail. Partly inspired by Gregory Paksi.
Redundant colors can be omitted from the ColorFile. See colors/README.col for details. Also, there are some new colors, and some old ones are gone; be sure to check your ColorFile if you’ve customized it.
Messages to or from you are highlighted in a different color in the letter list. After Ingo Brueckl (though this implementation is a little different from his).
Read-only support for Blue Wave’s .XTI files (last-read markers). This means that any old packets you read with Blue Wave will be marked in the same way when first opened in MultiMail; however, any changes made in MultiMail will not be seen by Blue Wave. Partly inspired by I.B., but this is a different (and more correct) implementation.
“Home” and “End” keys are now available in text entry fields (i.e., ShadowedWin::getstring()). After I.B.
The FMPT kludge line is now supported, which means that in Blue Wave mode, the netmail addresses of points will be recognized.
Aborting a header edit from the letter list could cause a segfault. (The fix in 0.19 was incomplete.)
Some bogus ANSI codes (mainly, “ESC[?7h”) are now semi-interpreted, so they no longer clutter the screen. Also, the ANSI background color is now hardwired to white on black, instead of being set by Main_Back.
The highlight method for the active line was no good on a light background with ncurses 4.2 (was ok with 4.1). Some other changes for light backgrounds, too.
High-bit, “low-bit” (below 32) and certain special characters are now handled better by MultiMail, being actually printed instead of interpreted as control codes. :-) Particularly useful for ANSI.
TAB characters are now rendered as the appropriate number of spaces in the letter and ANSI windows.
In text entry fields (getstring()), high-bit characters were showing up with weird attributes (due to a sign error). Reported (indirectly) by I.B., this bug has been present since about 0.10! I’m suprised no one from Fido zone 2 mentioned it to me before.
Packets can now be renamed without MultiMail losing track of its last read markers. (This will, however, still result in a second .red file being created.)
When entering a tagline manually via “E” in the tagline editor, the entry field is now placed correctly, instead of appearing to overwrite the last tagline. Also, bogus taglines are no longer generated from message lines which are just “…”.
In the Reply area, the help menu for the letter list was showing a ‘K’ where it should’ve been an ‘S’. Also, capitalization in help menus has been standardized (only the keys that activate commands are in caps in the descriptive text), and the pop-up letter help window is now in two columns.
Killing an item (in any menu) now moves down to the next entry, regardless of the position in the list (unless at the end).
The ‘L’ command now remains available in the address book after a screen resize (SIGWINCH). Also, taking an address via ‘L’ automatically sets the active line to the last position (where the new entry is).
Netmail addresses are now included in Saved letters.
The resize (SIGWINCH) handler was causing a segfault in some circumstances. Although I just noticed it, this bug dates to 0.18. Curiously, it occurred for me only when running under “screen”. The fix was to change Win::inkey() to remove the loop.
Changed the .mmailrc separators from “ = “ back to “: “, which makes them compatible with 0.18 and earlier. I’ll probably change them again. (I was trying to achieve a more “classic” style of .rc file, but in fact, it still wasn’t quite right.)
The cursor was not being disabled in PDCurses versions.
“Read” indicator in the letter window was not updated after Saving.
Changed interface/Makefile and mmail/Makefile to allow “make clean” to work in DOS. Suggested by I.B.
And of course, various internal changes.
The .mmailrc has been redesigned. New versions of MultiMail will update it automatically; comments will be lost, but old data will be preseved. Note: the new form is not readable by 0.18 or earlier, because of the use of ‘=’ as a separator. Partly inspired by Ingo Brueckl.
Non-fatal errors. Currently, this is used only for the previously fatal “Could not uncompress packet” and “Packet type not recognized”.
You can change the colors MultiMail uses without recompiling. See the automatically generated “~/colors” for details; also, some sample color schemes are included in the 0.19 archive. This feature is due largely to Ingo Brueckl.
“Marked” messages are kept in the active letter list, whether read or unread. In combination with the “Save” changes (see below), this means that marking is now actually useful. :-)
Extensive changes to the “Save” dialog: When you Save from the letter list, you’re given a choice of “All”, “This one”, or “Quit”; if any messages are marked, you also get the option “Marked”. Saving “All” now saves all messages in the active list, rather than all in the area. (To get the old behavior, just toggle the full list on first.) Saving a message sets the “Read” flag, and turns off “Marked”. The last-entered filename is retained (separately) for “All”, “This One”, and “Marked”. If no name is entered manually, the automatic name is regenerated with each use; otherwise, the manual name is kept until you exit MultiMail.
Segfaults in saving All, in some cases.
Certain ANSI pics, with codes in the form “
After the cursor was turned on, e.g. by the Save window or an aborted message header edit, it was left in the middle of the screen on some terminals. It is now forced down to the corner.
The header editor was being left on screen after exiting via ESC.
Removed the text “Fatal Error” from fatal errors. :-) It was sometimes redundant, and other times wrong.
When the last packet was deleted from the packet list, a segfault could occur. Reported by Alan Ianson.
Functions which redrew the letter window (e.g., toggling character set translation) were setting the Read marker each time.
The tagline window now looks better on screens other than 80 columns.
Various internal changes, and added comments.
MultiMail can now read QWK packets without .NDX files. Partly inspired by Simon Callan. Note: It’s still preferable to include them, since it takes about three times longer to open the packet without them. (But that’s not counting the unpacking time, which is much longer still.)
The letter list now allows toggling between showing all messages, and showing only those which are unread; when entering an area, it defaults to unread (unless all the messages are marked read). I’ve found that this makes a BIG difference in the feel of the program – it’s more like a newsreader now. ;-) Use the ‘L’ key to toggle between the long and short lists (as in the area list).
Removed a lot of redundant code from the interface section, and reorganized it… There are many new features that need implementing, but so far I’ve been concentrating on cleaning up and simplifying the existing code, so as to have a solid base to build on. I think I’m almost there. ;-)
Found a couple small memory leaks, in AnsiWindow (the statbar was not being deleted) and main_read_class. Also, the tagline file was being left open after reading it.
Now gives a fatal error (instead of segfaulting) if the screen is smaller than 60x20.
The address book and tagline window now respond correctly to SIGWINCH.
In the tagline window, the key for rereading the file has been changed from ‘F’ to ‘L’. This allows ‘F’ to be used as an alias for PgDn, as on other screens. (The real purpose was to merge the code into the rest of the keyboard-handling code.)
If there was only one area (i.e., REPLY), and the right arrow was pressed, an infinite loop would occur. Reported by Tamminen Eero.
The ‘N’ key, for Netmail, is now disabled when no Netmail area is available. Partly inspired by Francois Thunus.
Sample filenames in the default .mmailrc now conform to the OS (i.e., 8.3 (FAT) format for DOS and OS/2), and the unused “ReadDir” line has been removed. Inspired by F.T.
Stripping and adding of blank lines at the end of messages works a little better now, but still needs work.
The addressbook was crashing, when empty (0 items). Reported by Marc D. Williams.
Netmail addresses, when present in Blue Wave packets (or reply packets), are displayed in the From: or To: fields in the letter window, as appropriate; and when entering a netmail message, the address can now be edited. (Note: Netmail is largely untested yet.)
In the header editor (From, To, etc.), you can now use the up and down arrows to move between fields, or press ESC in any field to abort the edit. Press ENTER on the last field to exit normally.
Reading a message in the QWK Personal area marks it as read in its original area.
A full area list is now available in both QWK and Blue Wave modes; it can be toggled by pressing ‘L’ in the area list. (The compile-time “shortlist” definition has been removed.) One benefit is that you can enter a message in any area, even when only the short list is being displayed.
When re-editing a message, you can now edit the header as well as the text.
ANSI animation. While in the ANSI viewer, press one of the ANSI activation keys again (or ‘A’) to see an animated view. Press any key to abort the animation.
The ANSI viewer supports the screen-clearing code.
The cursor is turned off, wherever possible. Less visual clutter. :-)
True scrolling instead of redrawing; minimized the redrawing done for letter window, ANSI viewer, and all ListWindows. It’s now MUCH faster on slow machines and slow terminals.
MASSIVE rewrite… Almost everything that was crufty in previous versions is now done the right way. :-) General fixes include: plugging memory leaks (there were a lot of these, I’m sorry to say), elimination of one-based arrays, adding more limit checks, and removal of all kinds of kludges and redundant code. The interface is more fully separate from the specific drivers; adding new packet types now requires changing only mmail/driverl.cc.
The netmail area is found by attribute rather than name, and (owing to the full area list now available in Blue Wave mode) can always be found, instead of only when you have received netmail. :-)
Character set translation is applied to the area list, and to area descriptions in the letter list and letter window.
The number of replies shown in the REPLY area is updated immediately when it changes.
The method of adding blanks before taglines and tearlines is changed; it should no longer be necessary to type an extra CR at the end of replies.
QWK replies are automatically word-wrapped at 80 columns, if they’re not wrapped already. (In Blue Wave, this is not desirable. QWK doors seem to prefer individual lines; Blue Wave, paragraphs.)
MultiMail now ensures that QWK replies end with a line-ending character, instead of letting the last line run off into the padding area. Without this, certain QWK doors would strip off the tearline.
The path is no longer displayed in the Save window, and the suggested filename is conformed to FAT (8.3) standards. Note that you can still enter any pathname, and any filename that will be accepted by your system.
Specifying pathnames that end in a slash or backslash (e.g., in .mmailrc) should now work correctly.
The extraneous “bw” prefix found on many of the .mmailrc variable names is no longer needed, though it will still be recognized for backwards compatibility.
Space bar now works in the little area list.
The QWK “Personal” area no longer appears in the little area list, nor can messages be entered in it from the area list.
Unrecognized packets no longer cause a segfault (just an exit).
Added trivial SIGWINCH handlers for AddressBook and TaglineWindow (they just exit that menu).
Fatal errors are now reported correctly in the OS/2 version, even when they occur before curses initialization.
In the ANSI viewer, there are no more segfaults in OS/2, and no more stray characters on the status line in either PDCurses version.
ANSI Reverse attribute now works with PDCurses.
ANSI viewing works in reply area.
A SIGWINCH can no longer result in the active (highlighted) line being displaced off the screen.
New default LHA parameters for MSDOS version (the Unix ones didn’t work for it).
The currently selected address is preserved between calls to the AddressBook.
This version compiles for OS/2, with EMX. This entailed substantial changes to the Makefile (q.v.), as well as some code.
The environment variable MMAIL can now be used instead of HOME, to specify the directory of the .mmailrc (or mmail.rc) file. If neither variable is defined, MultiMail will use the current directory (this is also new).
The DOS-to-Latin 1 character table has been improved, for translation of graphics characters, by stealing most of the table from DOSEmu.
Truly automatic packet recognition, by packet contents instead of name. Patch by Robert Vukovic.
Internal ANSI viewer. Works on all platforms (except that it sometimes segfaults when scrolling, under OS/2). Much better than the “less” viewer, and it will form the basis of new file list and bulletin viewers.
Another big bug in the DOS version (argh!): Last read markers were not being saved. This bug was the result of an error in the bug fix in 0.15. :-/ I’m surprised no one has reported this.
In the previous DOS versions, under some arrangements of directories, the temporary directories would not be removed.
The man page has been renamed to “mm.1”, and the title changed, so that “man” and “apropos” will work more appropriately. The install will make a link to the old name (mmail.1).
The “Replied” flag is now set by O- and N-type replies, as well as R.
Some code reorganization.
Compatible with MSDOS. All source code files have been changed to single-case, 8.3 form, and #ifdef’s are used on filenames internally where necessary. (Most other issues are taken care of by DJGPP, though some Makefile changes were necessary, and the starting directory is now restored on termination – needed in DOS, but not in Unix.)
Compatible with PDCurses (as distributed with DJGPP), and SysV curses (specifically, Solaris). Note that the shadows on windows are opaque if you compile with SysV curses. (I also added “#define USE_SHADOWS”, which you can comment out or remove to get rid of the shadows altogether.)
After the reorganization of resource.C in version 0.11 (if not before), any change to the value of mmHomeDir in .mmailrc would leave the bwPacketDir, bwReplyDir, etc. variables unaffacted.
memError() now works correctly, even before initialization of the interface. Fatal errors in general will be reported better in this version.
Some additonal bounds checking on sprintf() calls.
tmpnam() checked. In principle, as few as 26 temporary filenames might be available from it (which should still be enough, unless you write a whole lot of replies).
Area and system descriptions added to saved messages.
I removed the last remaining command-line option, “-m”, on the grounds that it was useless. (This is a new UNfeature.) After testing it on my 9600 bps terminal, it doesn’t appear to improve the speed; and the alternate color scheme, which it was originally designed to select, was never implemented. (Instead, I’ve added inversing to the top and bottom bars in the letter window, and made a few other changes to make it more monochrome-friendly.)
resource.C and resource.h rewritten and simplified to allow compilation with gcc 2.8.1. (It’s not yet clear to me why the old version didn’t work, but this is better anyway.) There are still a few warnings, but it works OK.
When entering the From:, To:, and Subject: on replies, the full width of the fields had not been available. Under some circumstances, this could even cause a segfault. These lengths still need to be checked more stringently.
Automatic “Re:” adding now takes place before subject-line editing, instead of after. This allows users to see that it will be added, and to override it if desired. (It also saves a few bytes of code!) I don’t know why I did it the other way before.
In the event of a “Fatal Error:” exit, MultiMail will now clean up after itself, as with normal exits. Also, failed memory allocation is now explicitly checked.
Default paths for “zip” and “unzip” removed. (These would only show up if the paths were not defined in the .mmailrc – as they are with the default .mmailrc.) Thanks to Carey Bloodworth for pointing this out.
If the message numbers exceeded 5 digits (i.e., 100000+), they would mess up the display in the letter list. Now, it takes 6. :-)
Much internal reorganization.
Fixed a bug with my website – the new-style URL (/~wmcbrine/) messed up some relative pathnames. :-)
Shadowed windows! Tell me if you like them.
SIGWINCH support. You can now resize the terminal (e.g., maximize the xterm) while MultiMail is running, and it will adapt to the new size.
Under Solaris, the help menu area was not being cleared when changing from one menu to another.
One more space available in letter list (now consistent with the other lists).
On some systems (such as Linux w/ glibc, and NetBSD), if you attempted to read a packet that didn’t already have a “bbsid.red” file (mm’s read markers) in it, MultiMail v0.9 would segfault right after unzipping a packet, due to a very stupid attempt to call fclose() with a NULL pointer. Under Linux w/ libc.5, this actually worked OK; and I didn’t notice it on my NetBSD test system because the packets there already had .red files in them. Thanks again to Cesar Cardoso for reporting the problem.
Due to the seriousness of this bug, I’m releasing v0.10 early.
Added alias, “!”, for F2. Added Tab as an alias for right arrow (next unread) in letter list.
Better handling of screen widths other than 80; better use of available screen space even in 80 columns. Still to do: handle SIGWINCH.
QWK now has the option (on by default) to show only those areas which have messages in them, instead of the full area list. I set this on by default to match the behavior of the Blue Wave side; you can change this in the top-level Makefile. Currently, it’s only a compile-time option. (In the future, I’ll add the ability to do a full list in Blue Wave mode, and to toggle the mode at runtime.)
.REP and .NEW filenames are now forced to lowercase. Should be easier to type. :-) But if you have any uppercase-named reply packets from previous versions, you’ll have to manually rename them before 0.9 will recognize them. Sorry.
Blue Wave mode now works on big-endian systems! The full functionality of MultiMail is now available on all platforms where it compiles. And “-fpack-struct” is no longer needed in the Makefile – which I hope may mean increased portability.
system() calls to rm and sed have been eliminated.
Strip spaces from Blue Wave subject lines; fixes sorting in some cases.
In Blue Wave mode, MultiMail was using LF as a paragraph delimiter. The correct behavior is to use CR as the delimiter and ignore any LFs. Thanks to Marc D. Williams for submitting a packet that required this fix.
Messages in Blue Wave packets are supposed to have a leading space, but packets produced by the ReneWave door lacked them. They also contained nulls in messages (a no-no). MultiMail now deals with these problems without flaking out. Thanks again to M.D.W.
With Blue Wave packets, MultiMail now scans for “*.inf”, instead of assuming that the packet name minus the extension is the basename. Normally, that is the case, but the “welcome!.000” packet that comes with the Blue Wave reader – which uses “welcome” internally – is a counterexample. MultiMail can now read this packet. :-)
Startup for QWK packets is much faster, especially on slow systems. Previous versions would attempt to open the .ndx file for each area defined in the control.dat (and would do so several times for each area), whether it existed or not; 0.9 scans to see which .ndx files actually exist first.
Changed ANSI viewer to work better with certain messages. Really, I should make it user-definable, instead of being hard-wired to “less”; but I’m probably going to make it into an internal function anyway.
Added “#include <sys/types.h>” to mmail/mmail.h. Usually I wouldn’t mention a change of this type, but in this case, it was done to get MultiMail to compile with glibc (libc 6). Thanks to Cesar Cardoso for reporting the problem and testing the solution.
Saved messages are now wrapped at 80 columns, regardless of screen width at the time of saving; and the date is now added to the saved header.
Replying to a message no longer messes up the right margin of the original when viewing it immediately after replying.
Paths are no longer stored with the .red file when using LHA. In previous versions, this could cause the storing of multiple .red files into a packet.
Next/previous unread in letter list now work correctly, even with messages that have manually been marked Unread.
Miscellaneous minor internal fixes and optimizations.
“Re: “ is stripped from subject lines, for sorting and display purposes, and added automatically on replies (unless doing so would truncate the subject). Subject sorting is now case-insensitive.
Letter window now displays “bbsnum (x of y)”, for consistency with the letter list and to provide more information.
Slightly expanded the area description length to make better use of the available space. Also, in the letter list, more of the subject is shown.
Restored the “line/lines” counter found in early versions of MultiMail. This shows the length of the message in lines, and the number of the top line on screen.
Temporary files are now cleared at the end of a session. Also, since MultiMail now generates a unique temporary directory for each session, a single user can run multiple concurrent sessions. (Just don’t try to read the same packet in each one!)
System uname added to Blue Wave tearline (if it will fit).
Character set translation can now be toggled at runtime, by pressing ‘c’, instead of at compilation time.
Packet type is now recognized automatically, based on the filename’s extension. (You can still force the other type.) Consequently, the “-b” and “-q” options have been removed.
The Blue Wave area list now uses the description instead of the echotag. This is much more useful, and conforms to the Blue Wave reader.
The number of the original message is now passed to the door for reply linking.
In the letter window, the top and bottom bars are now inversed when in monochrome. In the various lists, the highlight bar is now drawn with stdout(), for greater contrast.
Many changes to the top-level Makefile; it’s now more portable, and commented. You can now set the location of the ncurses header file here, instead of editing the source.
The packet list is now sorted, and shows file sizes as well as dates.
The backquote character (`) is converted to an ESC when using the ANSI viewer. (Some systems recode ESC characters this way.)
You can now Kill packets from the packet list.
If all preexisting replies were deleted, attempting to create a new reply during the same session could cause a segfault.
Terminal newlines stripped from replies. Conforms to the Blue Wave reader; suppresses gaps after the tagline with some doors.
QWK subject fields had been truncated at 24 characters. (The field is 25 chars.)
The tagline file can now be hand-edited without introducing blank lines.
In the letter window, PgDn, End, and the down arrow now stop at the actual end of the text.
No more extra junk in Blue Wave reply packets.
Blue Wave replies can now be safely reedited. (Previously, the line endings could be messed up – left in Unix format.) Stray characters no longer appear at the end of Blue Wave replies while viewing them.
Blue Wave mode, like QWK mode, now shows the BBS message numbers in the letter list, instead of the messages’ position in the packet.
Trailing “, Sysop” stripped from sysop name in QWK mode.
Opening reply packets that were generated by the Blue Wave reader caused a segfault, due to case mismatch; it now works correctly.
Version number “encryption” for Blue Wave replies fixed (so the number appears correctly in tearlines). (IMO, the Blue Wave specs are in error in describing this feature – bluewave.h refers to addition when it should say subtraction.)
Left and right arrow keys in the area list now work much faster when skipping empty areas, and can take you to the first and last areas, instead of second and next-to-last.
Killed the stupid repeating-REPLY-area bug in Blue Wave mode.
When creating Blue Wave replies, MultiMail now makes the proper choice between “real name” and “alias”, depending on the area flags.
Right and left arrow keys now select next/previous unread message in an area, similarly to the way they function in the area list (jumping to non-empty areas).
ANSI viewer. Hit ^A while reading a message to view it in color, if it has ANSI codes embedded. This is still rudimentary; it uses “less” as the viewer, and it depends on the terminal to interpret the codes.
Private flag support. Complete, except for Blue Wave area flag checks.
Character set translation is essentially complete. The only further changes I envision are the ability to turn translation on or off at run time, instead of at compile time; some possible changes to the translation table; and maybe additional character sets.
Stolen taglines are checked for dupes, included only once.
With some minor changes, I got it to work under SunOS (Solaris). No longer just a Linux program! :-) So, the tearline is now derived from the uname. Blue Wave mode still requires a little-endian system.
QWK “Personal” conference.
Many key aliases added, mainly for use on terminals that don’t support keys like PgDn and F1. Also, Space Bar now functions as a combination PgDn/Enter in the letter window, to allow paging through a conference.
Marking and read/unread toggle now work from the letter list, simplifying bulk marking. In the REPLY area, Kill now works from the letter list.
Automatic creation of .mmailrc and the mmail directories; no more “make install_dirs”.
Changed bluewave.h to version 3. (No related feature changes yet.) I now use it in unmodified form; “-fpack-struct” is sufficient.
Suppression of “hidden” text lines added for Blue Wave (already present for QWK).
Append saved files instead of overwrite.
If saving without a path specified, save in the “save” directory.
Better adaptation to nonstandard screen sizes.
Really random taglines (previously, there was no call to srand).
“Save lastread pointers?” and “Reply area has changed…” messages now come up only when appropriate.
“Personal” column in area list removed in QWK mode. Maybe not a bug, but it was unused.
Fall back to login name if no alias name defined. (It was using only the alias name, sometimes leading to blank From: lines in Blue Wave mode. This feature needs more work – the alias should not always be the default.)
Various date fixes (QWK and Blue Wave).
Print the right area type for QWK Replies.
Area names padded out with spaces when necessary to correct a cosmetic defect in the REPLIES/PERSONAL letter lists’ “Area” fields.
Version number references made consistent.
Remove extra bytes from ends of messages (QWK and Blue Wave).
A big one: Kill the correct reply messages with ‘K’. (!)
A BIG one: QWK reply area numbers fixed. It was using the internal area number, rather than the QWK number; so replies would go to the wrong areas! (Under specific conditions – a packet with conferences that were numbered serially, starting from zero – this would actually work right, which is doubtless how the bug snuck in to begin with.)
Stop truncating replies. (This bug complimented the extra byte bug, preventing segfaults in the old version.)
I changed the default directories from “bwdown”, etc., to “down”, etc. This wasn’t a bug, but it was too Blue Wave-centric for a dual-function reader. (I believe the original intent of K.T. & T.I. was that the QWK implementation would have its own directories, e.g., “qwkdown”; but as implemented by J.Z., a common directory was used. I may revert to the putative original design at some point. Internally, the “bw” prefix is still used on many shared structures.)
Make .REP packets from the BBSID, not the base packetname.
“Unread” now counts all messages marked unread, not just those which are also unmarked and unreplied.
Letter sort fixed (now sorts by number within each subject), and faster.
Eliminated the defaulting to “reply” when an otherwise undefined key was pressed.
Calls to todos replaced with internal code.
QWK .ndx parser replaced with faster, non-endian-dependent version.
Many ncurses changes. Most importantly, shells and (normal) exits now restore the screen mode. (Error exits still need fixing.)
Makefile changes (including centralizing options in top-level Makefile). Could do with a bit more changing, I think.
A big one (since it prevented me from even using 0.6, as it was): Fixed segfaults on opening packets with uppercase names within (i.e., those from most or all DOS-based BBSes).
Many minor changes to suppress warning messages during compilation.
Many, many more. :-) These are just the user-visible ones.
– William McBrine
Oops! The signature was in the wrong place. Right order is: signature, tagline, tearline ;-)
Tagline adoption is supported! (Doesn’t check dupe tags yet.)
QWK-reply packs contained ‘\n’ as a line terminator. It’s incorrect! We have to use softCR (char#227).
Character conversion is somewhat in! For ISO 8859-1 <–> CP437 (DOS) translation, #define ISOCONVERT in /interface/interface.h
ISO conversion doesn’t do the from/to/subj yet
Fixed another date bug (when reloading QWK replies, the date would be mangled).
You can now use a signature file after each letter (specify its filename in .mmailrc)
Says error, if ~/.mmailrc isn’t found. (The program needs this file!)
File open errors printed, inside a BW/QWK packet, it tries four variations of the filename (eg. mybbs.dat, MYBBS.dat, MYBBS.DAT, mybbs.DAT).
The annoying bug, which occurred when quitting, has disappeared! ;-o
Fixed the date handling with QWK packets. Should do it well. That part wasn’t even implemented before.
Colors should be fine now, everywhere.
One small, but ANNOYING bug introduced: the program doesn’t exit properly, sometimes you have to kill it! Sorry, couldn’t trace this bug yet. Maybe you can help.
Took out many more bugs. This version now is almost usable ;)
You can now (re)Edit your reply!
Major design and color changes… Still not done, but doing good!
Added message Marking, Read/Unread toggle.
– John Zero
– Kolossvary Tamas and Toth Istvan