Change log
From 0.6.7 to 0.6.8
+ 0.6.8 was finished at about Tue Jan 12 09:53:42 CET 2010.
- Bugfix release: Interpreter and interface information is now correctly
written into the header after restarting, additionally the input length
for transscript and command input/output-stream filenames has been
fixed. Thanks to Samuel Verschelde.
- Added "sync-transcript" option: When enabled, it'll flush the wordwrapper
and file output buffer as soon as possible.
From 0.6.6 to 0.6.7
+ 0.6.7 was finished at about Mon Dec 14 22:28:51 CET 2009.
- Fixed trunction of babel data input. So far, the last byte was cut off
from the input, resulting in an error if the file was encoded using
unix newlines, but no errors as long as DOS newlines were used. This
fixes missing metadata issues in the story browser. Thanks again to
Zachary Kline.
From 0.6.5 to 0.6.6
+ 0.6.6 was finished at about Mon Dec 14 19:10:57 CET 2009.
- Fixed story browser to remove ".zblorb" suffixes from story titles
which are not backed by babel information and evaluated from the
filename. This will only work for new stories which are added to
the list, to re-index your current story-list the old file has to be
removed and re-built. "rm $HOME/.config/fizmo/story-list.txt" should
be run first, followed by the command "fizmo -u". Thanks to Zachary
Kline.
- Minimal changes to the seeding of math.c's "srand" and "init_by_array"
in hopes of increasing input entropy.
From 0.6.4 to 0.6.5
+ 0.6.5 was finished at about Tue Oct 13 07:27:56 CEST 2009.
+ Bugfix release: Recommend for MacPorts, will make compiling work for
parallel builds in MacPorts version 1.8.0+.
- Transcription will be kept active after restarts, thanks to Samuel
Verschelde.
- Fixed Makefiles for parallel builds (make -j).
From 0.6.3 to 0.6.4
+ 0.6.4 was finished at about Sun May 24 13:45:31 CEST 2009.
+ Bugfix release: Definitely recommend for read_char fix.
- Fixed interface to supply default colors in case ncurses' "pair_content"
returned invalid colors (thanks to Samuel Verschelde).
- Input from read_char is no longer re-echoed to the screen (thanks to
Samnuel Verschelde).
From 0.6.2 to 0.6.3
+ 0.6.3 was finished at about Sun Apr 26 08:26:31 CEST 2009.
- The top-level Makefile may now be invoked with "CONFIG_FILE=<filename>".
- Adapted "config.macos.mk" to MacPorts portfiles and ports-filesystem
so that fizmo is now easily MacPorts-buildable.
- Added license information to every file.
- Improved GNU coding compliance: install now goes to "($DESTDIR)/usr".
From 0.6.1 to 0.6.2
+ 0.6.2 was finished at about Sat Apr 25 00:59:41 CEST 2009.
+ Bugfix release: Corrected save opcode and prepared for debianization.
- Adapted manpage to make lintian work.
- Corrected "src/c/interface.h".
- Fixed typo in manpage.
- Fixed save opcode to not wrap data into IFF and to disable compression
when only a table from memory is saved (operands 2 and 3 for address and
length are set). This fixes "Comp96.z5".
From 0.6.0 to 0.6.1
+ 0.6.1 was finished at about Mon Mar 30 22:15:04 CEST 2009.
+ Bugfix release: minor addition of several features (may now compile
without libsndfile and libSDL alone, implemented XDG base directory
specification).
- Removed unused references to SDL-mixer (which isn't used for anything
anymore).
- Improved and documented default "config.*.mk" files and added config
default for cygwin with sound support (to my own great surprise, sound
really works under Cygwin).
- Added more information to "INSTALL.txt".
- Implemented new ENABLE_AIFF_FOR_SOUND_SDL variable in config files. This
allows the sound-sdl-interface to at least play Infocom .snd files when
no libsndfile is available.
- Added config-file options to manpage.
- Changed location of fizmo config directory from fixed "$HOME/.fizmo" to
XDG_CONFIG_HOME standard and implemented use of XDG_CONFIG_DIRS. Thanks for
suggestions and debugging help from Eric Forgeot. Specifications and more
information found at http://freedesktop.org/wiki/Software/xdg-user-dirs.
- Fixed broken space allocation, forgotten NULL-checks and fixed some issues
to make frontispiece work better on GenToo. Thanks for these patches from
David Leverton (see also line #50 in "src/ncursesw/Makefile"). This fixes
possible startup crashes.
- Added FIZMO_BIN_DIR variable to Makefile.
From 0.5.3 to 0.6.0
(find c cgi cpp fizmo ncursesw simple-c simple-cpp single-turn sound-sdl
-type f -name '*.c' -or -name '*.cpp' -or -name '*.h' | xargs cat | wc -l
resulted in 29953 lines of output)
+ 0.6.0 was finished at about Wed Mar 25 21:47:06 CET 2009.
+ This marks the point for the first public beta release. Killed all known
bugs as far as I could find them and made fizmo work on Linux -- tested
with a 64-bit distribution of Debian/Lenny -- and Mac OS X 10.5 with a
32-bit Intel machine and MacPorts for SDL. New features since version 0.5
include an SDL-based sound interface, support for charachter font via
unicode translation, partial (z)blorb-support -- runs code from blorbs, can
display frontispiece images in XTerms and play AIFF-sounds, a story file
navigator, interface margins, restore from command line, a much improved
streams implementation, a manpage and many other small features as well as
a large amount of bugfixes.
- Implemented conversion of font 3 (charachter font) to unicode. This
appears to make Beyond Zork's map appear at least readable.
- All directories in the "ZCODE_ROOT_PATH" are now recursively searched for
Z-machine games.
- Implemented recording and replaying of timed input.
- Implemented new "*.snd" search method: Generalized for all files, not
only Lurking Horror and Sherlock, tries upper- and lowercase.
- Fixed scan_table opcode. This fixes Beyond Zork window output.
- The filename input for save and restore may now be cancelled using the
escape key.
- Pressing arrow up/down during filename input no longer displays the
command history.
- Implemented wordwrapper / ncursesw margins.
- Flags are now correctly restored during undo and restore opcodes.
- Added progress indicator for story-list update.
- Fixed a bug that would damage the story list in case a game had a release
code with a length equal 5.
- Fixed an issue that would cause bits > 0xffff from routine address to be
lost on timed-routine-calls.
- Implemented "no-update" command line flag and config option to avoid
long startup scan times (for example on slow notebook drivers with many
story files).
- Implemented "--search" and "--recursively-search" command line invocations.
- Fixed bug in scrollback parapgraph position cache.
- Fixed a memory leak from getcwd(NULL, 0) in "filelist.c".
- Fixed scroll-down display for scrolling to bottom.
- Fixed crash that occured when a read-instruction was recursively called
from a timed input verification-routine.
- Modified LOADB / LOADW opcodes to store 0 in target instead of doing
nothing when trying to read from an illegal address. This appears to make
the russian version of "All Roads" ("AllRoadsR.z5") work.
- Fixed a bug in the story-title evaluation from filename when the story
name did not contain any slashes.
- Fixed display error on refresh of preloaded input from history.
- Implemented IGNORE_TOO_LONG_PROPERTIES_ERROR config define which
makes "property.c" ignore too long properties (fixes "HugeCave.z8").
- Implemented forced predictable mode (ignores seeding random generator).
- Fixed story-list for directly invoked files with relative pathname.
- Many, many other minor fixes and extensions.
From 0.5.2 to 0.5.3
(find c cgi cpp fizmo ncursesw simple-c simple-cpp single-turn sound-sdl
-type f -name '*.c' -or -name '*.cpp' -or -name '*.h' | xargs cat | wc -l
resulted in 28499 lines of output)
+ 0.5.3 was finished at about Wed Mar 11 21:34:59 CET 2009.
- Re-built SDL-Sound system (implemented sound effect stack, better sound-
has-finished-detection and many, many fixes).
- Re-wrote Makefile system. Releveant module defaults are now kept in
separate .mk-files.
- Added support for AIFF-sounds in blorb files.
- Added support for extra-blorb files (these are specified on the command
line directly after the story file name).
- Added "-Wextra" flag to CFLAGS and cleaned up new warnings caused by
the new flag.
- Cleaned up config file, fixed a minor bug and made boolean vars work
"more" correctly.
- Fixed "VERIFY" opcode.
- Several fixes: Patched "savegame.c" to re-enable use of "CMem" save method,
removed a superfluous free in streams.c which tried to free a static
string.
- Many minor fixes (option system, running X11-enabled fizmo in text-only
console, etc).
From 0.5.1 to 0.5.2
(find c cgi cpp fizmo ncursesw simple-c simple-cpp single-turn sound-sdl
-type f -name '*.c' -or -name '*.cpp' -or -name '*.h' | xargs cat | wc -l
resulted in 27123 lines of output)
+ 0.5.2 was finished at about Sat Jan 17 17:07:07 CET 2009.
- Implemented filelist (which remembers all files invoked and files in
searched directories).
- Added use of libxml2 to parse metadata from the babel XML file.
- Implemented story selection menu in ncursesw interface.
- Implemented parsing of rc file.
- Added bold-for-bright-foreground and blink-for-bright-background terminal
options.
- Implemented dont-use-color and force-color option.
- Fixed interface for > 121 colors.
- Added "help" command.
- Implemented "savegame-path" configuration option which specifies the
directory for savegames.
- Implemented text paragraph position cache to speed up scrollback for
large amounts of text (not elegant, but works).
- Savegames are no longer restored in case serial, release or checksum
don't match.
- Removed non-standard "FILE" chunk again, the storyfiles for directly
invoked savegames (from the command line) are now located via the
filelist.
- Implemened "force-quetzal-umem" config option from the command line.
- Added manpage.
- Fixed "verify" opcode.
- Basic implementation of input stream 1.
- Implemented basic blorb support (meaning that blorb files are detected,
information about pictures and sound is parsed and the first found "ZCOD"
chunk is executed).
- Fizmo now always keeps a FILE* to the story file open, eliminating the
need for a lot of open / close calls (and simplyfing blorb handling).
- Stories may now be started from the command line using their "real name"
in case they are stored in the story-list. Example: "fizmo sorcerer".
- Many, many small fixes (scrollback, winch-redisplay, negative score,
undo, crash on large scrollbacks, etc).
From 0.5.0 to 0.5.1
(find fizmo simple-c simple-cpp c cgi single-turn cpp ncursesw qzinspect
snd2aiff sound-sdl
-type f -name '*.c' -or -name '*.cpp' -or -name '*.h' | xargs cat | wc -l
resulted in 23265 lines of output)
+ 0.5.1 was finished at about Di 4 Nov 2008 21:48:29 CET.
- Implemented SDL-sound-interface. To make SDL work in Mac OS X I've used
MacPorts to install SDL ("port install libsdl"), on Debian I've been using
"apt-get install libsdl-sound1.2-dev" (maybe "apt-get install alsa-base
alsa-utils" is also required).
- Fixed scrollback and i18n-exit-translation.
- Fixed libfizmo to also use setitimer/sigaction (makes fizmo work better
on linux).
- Implemented foreground/background color command line parameters for
ncursesw interface.
- Fixed get-cursor-[column|row] for ncursesw (makes PRINT_TABLE and
Sherlock work).
- Added "UMem" support for quetzal.
- Added "snd2aiff" commandline utility (this was the basis for the first
sound interface experiments).
- Added "qzinspect", a simple commandlinetool which uses libfizmo to
list the contents of a quetzal savegame.
- Added non-standard "FILE" chunk to savegames which contains the absolute
file name of the story file which saved the game.
- Implemented restore from the command line: In case a quetzal-savegame
containing the non-standard-chunk "FILE" is given on the command line,
fizmo will try to restore the savegame using the story filename stored
in this chunk.
From 0.4.1 to 0.5.0
(find fizmo simple-c simple-cpp c cgi single-turn cpp ncursesw
-type f -name '*.c' -or -name '*.cpp' -or -name '*.h' | xargs cat | wc -l
resulted in 21963 lines of output)
+ 0.5.0 was finished at about Do 30 Okt 2008 11:42:27 CET.
+ This version is now definitely usable to play all non-v6 games. Did
extensive testing using Borderzone, "LostPig.z8", "Zokoban.z5", "crashme.z5",
"etude.z5", "paint.z5", "random.z5", "reverzi.z5" and "unicode.z5". Only
two minor known bugs remain: Scrollback sometimes miscalculates the
current row after a lot of scrolling back and forth (which is always
"fixable" to pressing any-key which correctly rebuilds the current output
page, and a display anomaly on the frontpage of "vampire.z8" which I
intended to keep after a lot of code-inspection (since fizmo appears to
be implementing the screen modell correctly and fixing this display
problem breaks a lot of other games). This version has been tested on
Linux, Darwin (Mac Os X) and a little bit on XP/Cygwin (using a self-built
ncursesw).
- Added "z_ucs_rchar" to "z_ucs.c".
- Re-implemented word-wrapper to not compress multiple spaces, keep spaces
after newlines and buffer more than on line. The fixes the map display
problem in "Enchanter".
- Changed opcode_restore to read dynamic memory and stack of the restored
game into a freshly allocated memory block instead of directly into the
game data. This makes it possible to now continue the current game instead
of dropping back to the command line.
- Fixed opcode_save to not exit in case an errors occurs during saving, but
to merely give a warning (and correctly evaluate the branch). Now the
currently active game isn't lost anymore in case a save goes wrong.
- Added a LOT of error checking in "iff.c" and "savegame.c". There's possibly
a better way, but checking every single return value also works.
- Added table-saving extensions to save- and restore-opcodes.
- Fixed tokenizing for version 5+.
- Fixed timed-input detection for "read" opcode. This fixes the crash in
ZTUU when the lantern goes out.
- Ran fizmo through (sp)lint, fixed memory leaks, minor bugs, many typecasts,
some double and inconsistent definitions, cleanup up code.
- Fixed Makefile dependencies.
- Fixed a possible SIGSEV when using fprintf with "%ls" output format. This
tended to crash when used on wchar_t on darwin.
- Implemened scrollback for ncursesw interface, added support method for
scrollback in "history.c".
- Added last missing v5 opcodes.
- Fixed end-of-stack detection so that stack is now correctly enlarged when
required.
- Added module-based localization, removed unused localization entries.
- Fixed space detection in wordwrapper.
- Fixed metadata-handling (memcpy) in history.c
- Added "single-turn" interface.
- Fixed operand handling in zpu.
- Added configuration system.
- Re-built save/restore, should be finally 100% quetzal compatible.
- Implemented output stream 4.
- Fixed output stream 2.
- Some split-window / set_window / set_cursor fixed for ncursesw-interface.
- Substitued ualarm/signal with setitimer/sigaction since the first one
won't catch SIGALRM on linux and manpage says that ualarm is obsolete.
- Fixed color management so that ncurses color pair is avaiable for reading.
- Many, many minor bugfixes.
From 0.4.0 to 0.4.1 (16373 lines in *.h and *.c files)
+ 0.4.1 was finished at about Wed Nov 14 08:23:19 CET 2007.
- Fixed a bug in "parse_utf_8_char", this fixes i18n-messages.
- Added C++ wrapper.
- Implemented demo C++ screen interface.
- Separated code in core, interface, c and cpp interfaces.
- Improved upper window handling (trinity sundial) and blockbuffer (zork1.z5).
From 0.3.1 to 0.4.0 (13718 lines in *.h and *.c files)
+ 0.4.0 was finished at about Tue Mar 20 07:57:01 CET 2007.
- Moved to this new version on March 3, 2007. The old version has been
lying around untouched since November 15, 2006.
- Abandoned the internal use of UTF-8. It is just much too complicated to
handle in the interface. I'm not even sure that with all the code necessary
to handle UTF-8 handling that there is much space wasted. The important
point for me is that using UCS-4 -- fixed, 32-bit sized characters -- is
much more easier, and thus safer to use.
- Re-implemented everything using wchar_t. The barebones C-Interface is the
first converted. Everything took only a few hours.
- Compacted code a little bit (tracelog.h, etc.).
- Re-implemented the history-buffer in "streams.c". It is controlled by the
Z_HISTORY_INCREMENT_SIZE and Z_HISTORY_MAXIMUM_SIZE constants in file
"config.h". The buffer is re-alloced until it reaches the given maximum
size. At that time, wrap-around is used to keep the buffer up to date. A
debug function named "log_history" exists.
- Converted the error handling to wchar_t.
- Implemented SIGWINCH-handler for interfaces and associated handling in
"ncrws-if.c".
- Implemented metadata (color and style) storage for the history.
- Implemented set_font opcode.
- Fixed two minor issues in the wordwrapper.
- Fixed a missing string-terminator in the zchar_to_ucs_4-function. Over
two years that thing was working without it ... unbelievable.
- Fixed a token-length problem in parsing.
- Implemented correct UTF8-filename-handling for savefiles.
- Implemented CATCH and THROW opcodes, both untested.
- Re-built interpreter commands with new wchar-base, prompt-repetition
from history and user-configurable prefix.
- Re-implemented scripting with UCS-4.
- Implemented buffering of metadata -- colour, styles and font -- in
the history buffer. Windows may now be resized and keep their styles
intact.
- Fixes in property-management for version 4 games.
- The sound-effect opcode is accepted now for effects 1 and 2, although it
doesn't do anything yet.
- Improved split-screen in ncurses-interface: In case the upper window becomes
smaller, its content is copied over into the lower window (making reading
the sundial in trinity.z4 working).
- Fixed set_window to set cursor to (0,0) when selection upper window (this
makes the sonarscope in Seastalker work).
- Fixed wordwrapper to accomodate for styles which are sent after normal
text which is followed by a space -- so far the style was activated before
the space was output.
- Implemented command history.
- Fixed a bug that would cause screen resizes in versions before 5 try to
set colours.
- Implemented specialized blockbuffer for upper window.
- Threw away wchar_t- and locale.h-usage. The problem is that they are not
suported on older compilers, embedded systems or in emulations like
cygwin. According to specifications, wchar_t may even only have 8 bits
width and is compiler-implementation dependant, so it's practically not
useable to store unicode-data. So i've implemented a z_ucs type which
is always used for internal processing of characters. Thus every interface
gets its output in z_ucs chars (which are simple int32_t types) and can
happily typecast it into wchar_t if desired. Interfaces may still use
wchar_t (in fact, ncursesw does), but the core is now completely autonomous.
- Fizmo now compiles fine without warning on Cygwin. Both the c- and the
ncursesw-interfaces seem to work nicely.
- Fixed a bug in the wordwrapper which caused the zchar-converter-state and
the linebuffer to overlay in a single byte.
- Added "z" parameter to i18n string, representing z_ucs-strings.
- Fixed stream-3-output: Current length is now always updated, not only when
closing the stream.
- Fixed input to make it work all across the screen and not only on the last
line.
- Fixed cursor movements and screen splits for version 4 and 5.
- Splitted files into fizmo-core and interfaces. The interpreter itself is
now put into a static library and linked in from the interfaces.
- Moved "extern" statements into header files.
From 0.3.0 to 0.3.1 (12276 lines in *.h and *.c files)
- Added a few ZSCII/ASCII/ISO-8859-1/UTF-8 conversion functions in "text.c".
- Improved the ncursesw-interface to handle US-ASCII and UTF-8 directly.
- Implemented detection if a character has a suitable ZSCII representation
before adding it to the input line.
- Fixed a bug which prevented correct assembly of multi-Z-characters.
- Fixed a bug that caused multi-Z-character constructions to have a wrong
word length.
- Extended ncurses interface to handle scrolling on the input line.
- Once input of a new line is finished, it is erase and then output again
using the interface's output_utf8 function. This allows output which is
longer than a single line to be word-wrapped.
- Implemented correct delete and insert in the ncurses interface.
From 0.2.3 to 0.3.0 (13302 lines in *.h and *.c files)
+ 0.3.0 was finished at about Fri Nov 10 20:48:54 CET 2006.
- Abandoned the concept to pack absolutely everything in UTF8f. Having to
think though the fork-8f is simply more complicated than necessary. In
the name of "Keep it simple, stupid" I'll revert to my old idea of
having a function in the interface for every output function like
color setting, changing font style and so on.
- Removed wordwrapper, pager, decoder, utf8conv, fork8f and fpipe. Reverted
the system to the plain-style-C interface. I'll use the break in the
development chain to clean up the code and take a second look at most
things which has proven quite benefical in similar occasions.
- Cleaned up most of the code. I've now got a clean, simple version without
any real interface around running again. Better yet, since my last tries
with ncursesw I've finally found out how to reliably produce UTF-8 output:
By simply calling the setlocale function from #include <locale.h> and
setting LC_ALL to "". Once the system locale is now set to something
using UTF-8, ncurses delivers correct output.
- Implemented a new, much simpler wordwrapper that should now also wrap
UTF-8 correctly.
- Fixed quetzal saving. The main FORM chunk has now the correct length so
that both frotz and Zoom now correctly load savegame files.
- Implemented a whole new wordwrapper. This one correctly wraps UTF-8 and
is capable of storing "metadata" -- colors and styles -- into the text.
This allows color and style changes in the middle of a word without
having to flush buffers.
- Implemented colors.
- Implemented text style.
- The etude.z5-test is running again. So far I've got everything working
except the character output.
- Implemented v3 window handling.
- The random.z5 test is now running okay.
- Implemented [More] prompt.
From 0.2.0 to 0.2.1 (...)
- Improved command line parameter handling.
- Fixed a REAL nasty bug in the line-history resulting in buffer overflows
in very rare cases.
- Fixed another minor bug that would let the word wrapper start a new line
after input with a space (Wishbringer's "read message and move leaves") by
added the wordwrap_reset_wrapper method (which correctly reset the
number of consecutive spaces).
- Fixed a bug that would cause a crash when games were saved in a restarted
game (forgotten store_empty_stack_frame).
- (Successfully completed Wishbringer -- 96/100 points, one wish used. I
just have to find a way to unlock the chains down in the tower...)
- Fixed paging module, long lines causing wrap-arounds are now correctly
accounted for. Modified the module to keep one more line than required
on the screen to ease continued reading.
- Tested compatibility to Zork I, Z-Machine version 1, release 2, serial
number AS000C. This version appears to have a bug at $585b. It occurs
when the command "open mailbox and read leaflet" is executed right at
the beginning. To circumvent this, a skip-underflow-check can be
enabled in variable.c.
- I've posted a related newsgroup-message at http://groups.google.de/group/
rec.arts.int-fiction/browse_frm/thread/c77652b28be69f2d/
a52b9f652fc75928#a52b9f652fc75928
- Added the "@info" command.
- Added automatic detection of "problem" games. This way, the "AS000C"
hack is automatically activated at start-up.
- Activated hacks are displayed using the "@info" command.
- (0.20.1 finished)
- Enchanhed the pipe/filter system. Each filter will now receive the
destination function (and "object-struct") directly. Thus, no more
extra forwarding functions are required.
- Rebulid the windowing system in the curses interface.
- No more flushes are sent from the fizmo "core" any longer. The interface
will flush it's filters by itself if required (sending a flush into
the first filter in case input is required or the interface is closed).
- ABSOLUTELY FASCINATING to discover how many can bugs can still reside in
a program just to creep up occassionally ... or is that just a sign of
a bad programming style? Now I begin to grasp even more the significance
of Java's Array-out-of-bounds-exception. I found a bug in the utf8x-buffer
(forgot a multiplation by factor 3, resulting in a much too small buffer.
I must have had a lot of overflows without ever noticing them in the
past). Found another bug in the UTF8x-Converter where I simply
referenced the wrong variable.
- Fixed a forgotton "converter->utf8f_parameter_bytes_to_copy = 0" in
utf8conv.c which SOMETIMES (grrr!) caused an error which caused the
screen output to vanish.
- Reorganized piping in the curses interface. Every window now has its
own combination of wordwrapper, converter, pager and decoder. General
window commands like split_window or erase_window are caught by a
enhanced type of decoder named fork8f and passed along to the correct
window structure. The random.z5 seems to work fine and I hope it's
sufficiently correct to finally try my luck with Seastalker.
- (0.20.2 finished)
- Fixed status line display.
- Fixed ncurses color recycling.
From 0.1.4 to 0.2.0 (12794 lines in *.h and *.c files)
+ 0.2.0 was finished at about Sat Apr 15 02:06:22 CEST 2006.
+ Version 0.2.0 adds an unix interface, allowing for timed intput, and an
[n]curses interface including color, better wordwrapping and accented
characters. Zinc now runs random.z5 and passes all of the etude tests.
I completed Moonmist without encountering any strange behaviour.
- Put the conversion of UTF-8 to [ASCII|ISO-8859-1|UTF-8] in utf8conv.[c|h].
That way, c_if.c is now down to a minimum size, and all it's "outsourced"
functions -- wordwrapping and UTF-8-conversion -- can now be used without
any problems by other functions or interfaces.
- Cleaned up Makefile.
- Implemented call_vs2 and call_vn2 opcodes.
- Made PRINT_NUM output signed instead of unsigned numbers (found it thanks
to etude.z5).
- Corrected OPCODE_MOD to correctly handle signed operand 0 values (etude
again).
- Corrected a REAL nasty mistake: Replaced sizeof(<constant>) in a
utf8conv.c malloc-call with a simple <constant>. This made the system
crash under OS X PPC (though not on linux and some other unix-variant).
- Re-organised the makefile a bit. Interface dependencies are now
correctly resolved and build dependencies are in order again.
- Renamed the interpreter from the project name "zint" -- meaning
Z-Machine-Interpreter -- to "ZInC", meaning Z-INterpreter-Code.
- Corrected the Makefile again.
- Implemented the Unix-Interface. This one assumes that it's running
on a unix machine terminal without any specials (not even curses).
It uses select and tc[g|s]etattr to implement timed and preloaded
input and a get_char implementation.
- Fixed a bug in unix_if.c that would cause a segfault on startup.
- Some more i18n, cleaner tc[g|s]etattr in unix_if.c
- Implmented correct input line restoration after timed-input output.
- Modified text.c so that the streams_echo_input command is executed
no matter whether the first word could be found in the vocabluary or not.
- Implemented preloaded input in the Unix-Interface.
- Implemented correct READ_CHAR in Unix-Interface.
- Fixed a LOT of tiny bugs and itches.
- Implemented [MORE] paging via additional pager module.
- Fixed german translation of i18n messages.
- Fixed ZSCII-to-UTF8 encoding, ZInC now passes etude's accented character
test.
- Implemented a curses interface. At the moment raw constructs for the
upper and lower window and a status line. Basic color management. Adapted
wordwrap.c a little bit in order to achieve corrent linebreaks.
- Completely re-wrote wordwrapping in order to get rid of trailing spaces
and other problems evident when using curses.
- Completely re-wrote the entire output stream system. It turned out that
the usage of flags caused problems buffering text: In order for
everything to work correctly the flags would have to be buffered along
with the text output. Text-relevant formatting data like color is now
passed directly along with the UTF-8 text: Internally I'm now using
an encoding that can encode special formatting sequences into the UTF-8
stream -- see UTF8X.txt for more information. These conversions added
the utf8x.* and decoder.* files. Although quite a major change, the
encoding helps to make the whole system a lot cleaner.
- Numerous fixes to the [n]curses interface, addition of status line, better
color management with careful consideration of availiable color pairs.
Using the ncurses interface, the interpreter now passes the etude.z5
test with the exception of the text styles and undo capability.
- Modified the i18n_translate_and_exit and close_streams functions to
support a message-on-exit. This way, error messages are supported without
regard for what's left in the UTF8x output pipe and error output can be
seperated from the other valid stream output. This way, the ncurses
interface can correctly endwin() and output the message to read on stderr.
- So far, multiple invocations of READ or READ_CHAR with timed input casued
a "nested timed input not availiable" error. This is a problem for the
random.z5 test which appears to use this feature. The nested behaviour
has been altered to an overwrite behaviour. In case a second timed input
is requested, the old timer and routine are erased from memory and replaced
with the new data.
- Fixed interface commandline choice.
- The random.z5 tests now apperas to run correctly.
- Fixed a tiny bug in wordwrapping directly after read_line, fixed command
line parsing, i18n-string-length and a few tiny localization issues.
- Added text styles to the curses interface.
- Finally added modifiable names for game saving and restoring.
- (Successfully completed the red storyline of Moonmist).
- Renamed ZinC (name already exists) to fizmo: Fizmo Interprets
Z-Machine Opcodes.
From 0.1.3 to 0.1.4 (8848 lines in *.h and *.c files)
+ 0.1.4 was finished at about Fri Dec 9 03:36:20 CET 2005.
- The 0.1.4 complies to the strictz.z5 test, allows transscripting (even
on startup via command line switch). A few minor bugs were fixed and
a history of the last output is kept, allowing for a faithfully
original prompt after an interpreter command has been finished.
- Compacted locate_dictionary_entry (so that the zscii_string_length function
is no longer needed) and adapted it to version 4+ files.
- Implemented the object-number-unequal-zero-checks (in order to run
strict.z5).
- Adapted read opcode to version 4+.
- Renamed c_iface.c to c_if.c in order to gain more space for other interface
names.
- Fixed zchar_to_utf_8 so that multibyte-characters are handeled correctly
between two zchar_to_utf_8 function calls (when continuing on the first
call's source string due to output buffer shortage).
- Implemented basic transcipting in streams.c and moved the trace functions
there. All output is now channeled from text.c through stream.c into the
active interface. The input is seperately written to to the streams, in
order to be able to distinguish it from regular output (and to write it
only to the streams requested).
- Implemented "-s" / "--start-script" startup option to start scripting
right away. This allows to capture really all input emitted by the game.
- Implemented input of file name (supplying a default filename) for
transcripts.
- Fixed a bug in the parsed word position which would cause the following
garbeled Moonmist-output:
>get out of the cat
[I don't know the word " cat."]
>oops car
[I assume you mean: get ou o th ca]
You're not in it!
- Implemented CALL_VN opcode.
- Added STRICT_Z definition in config.h and implemented a lot of checks in
object.c together with a warning mechanism in i18n.c in order to make this
interpreter compliant to strictz.z5.
- Fixed a bug in the length-code-size determination in properties.c (replaced
"& 80" with "& 0x80").
- Added a lot of STRICT_Z tests to properties.c.
- Implemented READ_CHAR opcode (not very well, since in stanard C without
any Operating-System-specific calls there's no single read of a char, thus
you still have to press enter).
- Zinc now passes the strictz.z5 test and looks close enough to the frotz
output stored at http://www.ifarchive.org/if-archive/infocom/interpreters/
old/frotz/frotz_zstrict_result (with other warning messages, of course).
- Implemented wordwrapping as a seperate re-usable module in wordwrap.[c|h].
The C-interface and the function for output to stream 2 currently use it
both.
- Implemented line_history_buffer, a circular buffer that will record the
last output sent to the streams. It's supposed to be used for
reconstruction of the current prompt in case an interpreter command
is entered an the original prompt should be restored after executing it
and later for easier handling of preloaded input.
- Interpreter commands will now prompt with the last line that was output
by the Z-program after finishing command processing.
From 0.1.2 to 0.1.3 (6013 lines in *.h and *.c files)
+ 0.1.3 was finished at about Sun Nov 27 15:06:18 CET 2005.
- This 0.1.3 version adds localization for error- and startup messages,
corrects a few minor bugs abd adds UTF-8 support.
- Made mod (remainder-after-divison) signed (due to 2.2.1).
- Corrected opcode_div so that the divisor, not the dividend, is verified
not to be 0. Added divisor-not-0-check to mod.
- A few paranoid sanity checks in opcode_read.
- Modified opcode_read so that the whole line of input is always read, no
matter how much can actually be stored so that the next line of input
does not begin with left-over-text. The non-used input is discarded.
- Removed use of input_to_zscii-function in text.c, since the input does
not have to be converted.
- Extended store_ZSCII_as_zchar in text.c to accept not only characters in
the current alphabet, but also to encode other chars as multi-z-chars,
thus also allowing the '@' as input which I want to use to identify
interpreter commands.
- Implemented the "@predictable" command to switch interpreter in or out of
predicatable random mode as suggested in the spec.
- Made random generator more spec conforming by using negative values to
seed the generator to this value.
- Implemented parsing of command line options. Currently only used to start
up interpreter in predictable random mode.
- Implemented UTF-8 handling. The data traffic between the core interpreter
and the interface(s) has been modified to use exclusively UTF-8. It is
now also possible to output a whole string via the utf_8_output function.
- The C interface can now be put in three output modes -- ASCII, ISO-8859-1
and UTF-8. Characters that cannot be printed in the ASCII and ISo-8859-1
modes are substituted by a '?'.
- Implemented UTF-8 output as defined by the Z-Machine specification v1. So
far only implemented without testing.
- Implemented loclization. In order to avoid special cases in which it is
not possible to tell the user that the localized files cannot be loaded
since the files have not been loaded yet in his own language :-) I'm
using the prepi18n binary to convert the locales/xx_XX.txt into the
file locales.c which can be directly complied into the interpreter binary.
The localization files are supposed to be encoded in UTF-8 so they can
be easily output using the usual interface functions.
- Split the activate_interface function into activate_interface and
link_active_interface_to_story. The first call actually initializes the
interface which is used to make output possible (in case the story file
can't be loaded), the second call performs the necessary initializations
with the story file's header.
- Extended localization to use parameters the like "Could not open \{0}.".
So far, parameters are only allowed to contain ASCII values.
- Moved last remaining function in output.c, read_zscii_string, into text.c.
- Implemented forgotten lowering of case after reading input from the
keyboard. Now it's possible to answer "YES" to Moonmist when it asks
for "YES or NO".
- Cleaned up stack.c a bit and fixed a bug that would have caused the
stack not to enlarge in case it was supposed to.
- Extended localization to use type identifiers. In the beginning I
decided to skip them, but to make error messages useful to the user
and handier in the code I decided to give it a shot. Parameters are now
specified by using \{0s} for a string and \{0d} for a decimal. Together
with a few supporting wrapper functions, the main file already looks
much cleaner. Due to the buffering of the Z-Machine output, messages
are even correctly word-wrapped.
- Implemented switching between ASCII / ISO-8859-1 / UTF-8 into zinc.c.
Logically it would have been better to put it in to c_iface, but since
it will be required more often and don't want any copied code I decided
to leave it this way.
- Implemented echo option in the C interface in order to be able to read
the input from input piped into the interpreter.
- Converted all current 51 conditions-to-translate from FATAL_ERROR and
[s|f|]printf to i18n_translate[and_exit]. The FATAL_ERROR macro has been
removed.
- Compacted the zchar-storage-functions.
- Completely replaced read_zscii_string with zscii_to_utf_8. The new
functions will output z-char-strings of an arbitrary length into an
limited-size output buffer. Multiple calls to this function for the
same source string will continue with the output where the last call
of the functions had to stop.
From 0.1.1 to 0.1.2 (5094 lines)
+ 0.1.2 was finished at about Thu Oct 20 23:45:34 CEST 2005.
+ 0.1.2 fixes my currently last known problems with playing the 840726
Zork I revision. Word wrapping makes it look a little bit nicer and the
barebones-C-interface is now probably as good as it can get. Did quite
a lot of code-cleanup.
- Implemented NOT opcode.
- Implemented NOP opcode.
- Implemented RESTART opcode.
- Implemented POP opcode.
- Implemented SHOW_STATUS opcode.
- Allowed output of tab -- ZSCII 9 -- for version 3 (due to the guidebook
in the Dam Lobby).
- Implemented VERIFY opcode.
- Implemented PIRACY opcode (we're all honest, so an evaluate_branch(1)
sounds okay, right?).
- Fixed a bug in INC_CHK (wrong cast to unsigned value instead of signed).
- Totally forgot about the 8.3-filename-limits on DOS machines, files were
renamed and merged together -- /(.*)_opcodes.c/ joined /$1.c/.
- Fixed a bug in the random opcode which allowed 0 as a minimum result value
to the random call (should have been 1). Fighting with the thief now
results in much less garbled text output.
- Cleanly wrapped TRACE_LOG calls in curly braces, meaning disabling
tracing now works correctly. The zinc binary now has a size of 82308
bytes with logging enabled, and 54700 bytes without logging.
- Created Makfile-dependency for zinc.h from iface.h (by using the touch
command, is the okay?). Made all and clean .PHONY targets.
- Removed globals.c, moved globals into zinc.c and created "extern"
references in other source files.
- Moved globals to corresponding source files (which eliminates a lot of
non-needed "extern" references).
- The linux manpages tell me bzero is outdated. Faithfully replaced bzero
with memset calls.
- Replaced memcpy call with a loop that assembles uint16_t values byte by
byte. This will avoid problems on little-endian machines (i386 and the
like), since the Z-machine uses a big-endian model.
- Used the Mersenne Twister as random generator. Only the genrand_int32
function is used, the rest of the functions, with exception of the
two init calls, were deleted from the mt19937ar.c file. The generator
is seeded by an init vector consisting of the number of seconds since
1970, a standard C random generator number, the number of milliseconds
since 1970, followed by another standard C random generator number.
- Splitted zinc.h in a header file for each .c file. Adjusted Makefile
and #includes accordingly.
- Changed **z_opcode_procedures into *z_opcode_functions and moved it from
zinc.c to zpu.c (thus avoiding external reference and setup). Now
z_opcode_functions is a one-dimenstional array indexed by the value
instruction_form + instr_code. This allows saving the space for 32
pointers in the smaller opcode classes (with only 16 instead of 32
opcodes).
- Implemented buffering / word-wrapping in the C Interface. With this
change, the C interface should be complete. Comparing to everything else
it's still very raw, but hey, it works.
- Implemented BUFFER_MODE opcode.
- Moved all the configuration options from zinc.h to config.h.
From 0.1.0 to 0.1.1
+ 0.1.1 was finished at about Fri Sep 23 03:16:56 CEST 2005.
- 0.1.1 was the first version that was able to save and restore games.
- Implemented OR opcode.
- Implemented MOD opcode.
- Changed get_prop_len so that get_prop_len(0) returns 0 (for 1.1 standard).
- Updated property.c, splitted up property search in three functions in
order to allow easier implementation of the GET_NEXT_PROP opcode.
- Implemented GET_NEXT_PROP (you can now pick up the elven sword).
- Implemented CALL_2S.
- Implemented CALL_2N.
- Implemented SET_COLOUR (but does not have no effect in c_interface).
- Implemented all remaining 2OP-opcodes with the exception of THROW.
- Implemented DEC opcode.
- Implemented PRINT_ADDR opcode.
- Implemented CALL_1S opcode.
- Implemented REMOVE_OBJ opcode.
- Implemented LOAD opcode.
- Modified the tokenising routine to correctly return position and length
of a word even if it was not found in the dictionary.
- Implemented dummy stack frame storage for quetzal compatibility.
- Implemented SAVE opcode.
- Implemented RESTORE opcode.
From 0.0.4 to 0.1.0
+ 0.1.0 was finished at about Wed Sep 14 00:45:00 CEST 2005.
+ 0.1.0 is the first version that allows basic playing of version 3 games.
The only game used for testing so far is Zork 1 which was drawn from
the well-known site of Peter Scheyen.
- Moving around the house, object movement (taking and dropping of the
leaflet) all seems to work quite okay now.
- Implementing more missing opcodes.
- Fixing a few not-so-easy to find bugs (like the property addresses etc).
- Fixed a bug in the branch-address calculation.
- Fixed a bug in the object-unlinking mechanism that would not correctly
remove the object-to-be-unlinked from an object further up the sibling
line.
From 0.0.3 to 0.0.4
- Implemented the c_interface, which allows very bare-bone input / output
provided by standard C facilities (meaning fprintf / fgets).
- Implementing of user input, tokenising and dictionary parsing.
From 0.0.2 to 0.0.3
- Major cleanup of the code.
- distribution of the old interpreter.c file into multiple .c files.
- First appearance of the interpreter.h interface.
From 0.0.1 to 0.0.2
- Further implementation of the basic capabilities, including property
management etc.
To 0.0.1
- First basic implementation of memory, variables, stack and primary
opcodes.
