Make the same simplification to save and restore of obj->oextra as
was done yesterday for monst->mextra: no need for a set of all 0 sizes
when the whole thing is null. Bumps EDITLEVEL again.
A mimic posing as a statue was displayed as a tengu statue (and
recognizeable as such now that statues are displayed as the corresponding
monster rather than rock-class back tick), but the lookat code described
it as a giant ant statue (since there was no obj->corpsenm available to
indicate the monster type, it defaulted to 0). This adds monst->mextra
field `mcorpsenm' so that mimics have a place to remember what sort of
statue or corpse they are mimicking. And it picks a random monster type
when they take such forms so that the old tengu hack becomes irrelevant.
newmextra() and newoextra() initialized pointers via memset(...,0)
which is not portable; switch to explicit assignments. The wizard mode
code to display memory used for monsters and objects added in amounts
for the miscellaneous things pointed to by monst->mextra and obj->oextra
structs but didn't include memory for those structs themselves; add it.
Simplify monster save/restore slightly; there's no need for extra zeroes
to represent monst->mextra->X sizes when monst->mextra is null.
Update the startup banner for 2009. I should have done this with a
separate patch but I'm taking a shortcut. :-]
I almost abandoned this when Michael beat me to it, but besides
handling the fruit rename bug it also moves `current_fruit' into the
context structure to eliminate separate save/restore for that.
There was an issue reported where save files between different
versions of a manufacturer's compiler were incompatible because the time_t
ubirthday field was changed from 32 bits to 64 bits.
32 bit time_t implementations will break at 19:14:07 on January 18, 2038.
64 bit time_t implementations will break at 23:59:59 on December 31, 3000.
This removes the dependency on the size of time_t from the save file.
The ubirthday field is no longer embedded in struct you.
This also adds two general purpose routines to hacklib.c, one to convert a time
value to a 14 character char representation and the other to convert that
back to time_t. Those are used by the save/restore routines.
This is a savefile breaking change, so editlevel in patchlevel.h was
incremented.
From a bug report, mimics
which were exposed at the time the hero leaves a level remain unhidden
upon return no matter how long the hero is away. It was actually expected
behavior since the old level is stuck in stasis and hiders only hide when
it's their turn to move, but it was noticeably odd. This makes unhidden
hiders attempt to hide when hero returns to a previous level or enters a
bones level. I reorganized the monster handling in getlev() because the
relevant part was taking place before floor objects got restored, so
hidesunder() monsters had nothing to hide under at the time.
Reported to the beta-testers list by <Someone> last April:
restoring a normal game save file in explore mode let you keep the file,
then after exploring and quitting without saving, you could restore it
again in normal mode and take advantage of whatever information you'd
gained. This makes nethack -X (or playmode:explore) defer the switch to
explore mode when used while restoring a normal mode save file. It now
performs a normal restore (with save file deletion) and then acts as if
the user had given the 'X' command interactively, requiring confirmation
to actually switch into explore mode.
Reorganize the recent wizard mode control: move set_playmode() from
xxxmain.c to the core, and have it call new authorize_wizard_mode() to do
the port-specific part. If the set_playmode() call during startup doesn't
result in running in wizard mode (either because not allowed or user
didn't request it), it will be called again during restore if the save
file is from a wizard mode game.
For ports which check character name for authorization, players will
have to use `nethack -u whatever -D' (or options for name and playmode) to
restore a wizard mode save file if WIZARD has been changed from "wizard".
plname[] from a wizard mode saved game will always have that value, so if
it's not the right one players will need to get authorized by the startup
code before loading the save file.
Wizard mode or explore mode can be forced on (via -D or -X on the
command line, or now via OPTIONS=playmode:debug|explore) when restoring
a saved game; explore mode handling was confined to restgamestate(), but
wizard mode handling was replicated in every main(). Treat `wizard' the
same as `discover'. Also, prevent a new game started when restore fails
from using the old game's option settings if partial restore attempt got
far enough to load the flags struct. And update bemain.c and macmain.c
to catch up with the others modified by the playmode patch.
From a bug report: hangup during
screen updating at tail end of successful restore didn't create a new
save file when disconnecting. Use his suggestion for moving the setting
of program_state.something_worth_saving sooner, before the save file is
deleted. To do that, restlevelstate() needs to come sooner too. I think
this is safe enough to include in the branch code.
For the trunk, I'm not sure whether the SAFERHANGUP config will work
well here. It has to survive long enough under autopilot to enter moveloop
before the chance to save kicks in.
From a bug report. (Michael forwarded a newsgroup posting about it back
then, but I had trouble reproducing it and didn't figure it out until
trying again now.) If hangup occurred while entering the quest, the magic
portal could be rendered inactive for the hero but still work for monsters.
That's because the hangup save stored the old value of u.uz0 before
goto_level set it to the new u.uz, and a magic portal won't operate when
u.uz0 differs from u.uz (to prevent a pair of portals from getting stuck
sending the hero back and forth). The problem could also occur going from
the quest back to the dungeon, or either direction for Ft.Ludios, but the
--More-- prompt when the quest entry text is being displayed makes hangup
during level change most likely to occur during initial quest entry.
This is just a bandaid, and the SAFERHANGUP config wouldn't be hit
by this situation.
There's no need to restrict the menu of restoreable games to 52 in
order to stay within a-zA-Z for entries; just let add_menu() do its own
selector assignments (reusing letters on each page).
Move the code to use a nethack menu for restoring a saved game. It
was inline in tty_askname() but is now a separate routine, restore_menu()
in restore.c. There was no port-specific code and only a small amount of
tty-specific code; it should be useable by anyone (but Qt doesn't have to
switch over if it doesn't want to).
The original behaved strangely if there were exactly 26 saved chars;
the "start new game" menu entry ended up using "{" as selection character.
There wasn't any comparable problem at 52; it was limiting the menu to 51
games. Now it will allow 52 (with "start a new game" bumped into "#" if
there are that many), and adds an explicit quit entry (unless there are
52 or more games so that # is already used by new-game, then quit remains
implicit rather than resort to some other none-of-the-above character).
Pat wrote:
> <Someone> has a patch (we've added a couple of
> his earlier ones) which changes the statue display from a single
> one size fits all "`" to a gray monster symbol instead.
> But I think the idea is a good one, and along with the
> bouldersym option could make the fairly hard to
> distinguish back-tick character go away.
Sources tagged before applying NETHACK_PRE_STATUE,
and afterwards with NETHACK_POST_STATUE for easy
rollback.
- Instead of checking for the Rogue level, check which
graphics are engaged (PRIMARY or ROGUESET) in the
SYMHANDLING() macro.
- track which graphics are active through 'currentgraphics'.
- Instead of symset and roguesymset and symhandling and roguehandling
variables, have symset and symhandling be arrays of two, with the
following indexes:
PRIMARY
ROGUESET
That reduced the amount of repeated code.
(Not to be confused with the 'symset' and 'roguesymset' config file options
both of which still exist)
- the symbol routines were adjusted to pass
the index , rather than 'rogueflag' and coding to roguesymset etc.
Other than fixing bugs that are encountered, this is probably
the last of the symbol stuff, with the exception of
making the symset and roguesymset config file options
accept the keyword value "default".
Saving the game while punished, not carrying the attached ball,
and while swallowed by a purple worm resulted in losing the
ball and chain.
Since the required information was not being written to the
save file at all, I couldn't come up with a clean way to do this
for the branch, and preserve save file format. I could think
of lots of kludgy ways to do it (insert ball and chain into
the hero's inventory prior to saving, and remove it on restore, etc.)
move oattached and oname and other things that vary
the size of the obj structure into a separate
non-adjacent oextra structure, similar to what has
already been done for mextra. The obj structure
itself becomes a fixed size.
New macros:
#define ONAME(o) ((o)->oextra->oname)
#define OMID(o) ((o)->oextra->omid)
#define OMONST(o) ((o)->oextra->omonst)
#define OLONG(o) ((o)->oextra->olong)
#define OMAILCMD(o) ((o)->oextra->omailcmd)
#define has_oname(o) ((o)->oextra && ONAME(o))
#define has_omid(o) ((o)->oextra && OMID(o))
#define has_omonst(o) ((o)->oextra && OMONST(o))
#define has_olong(o) ((o)->oextra && OLONG(o))
#define has_omailcmd(o) ((o)->oextra && OMAILCMD(o))
changed macros:
has_name(mon) becomes has_mname(mon) to correspond.
The CVS repository was tagged with
NETHACK_PRE_OEXTRA
before commiting these, and
tagged with
NETHACK_POST_OEXTRA
immediately after. The diff
between those two tags is this oextra patch.
The associated mail daemon changes to use an oextra
structure instead of a hidden command located in the
name after the terminating NUL, have not been tried
or tested.
Move some internals-related code out of port-specific main so that
it isn't duplicated a bunch of times. One minor side-effect of this
change is that if you auto-pickup something at the very start of a game,
it will happen after any full moon/new moon/Friday 13th message rather
than before. There's a second change for some: the shared main() used
by several of the micro ports had a small difference in game play--if you
saved a game while on an engraving, it would automatically be read when
you resume--that will now occur for everybody [Elbereth weenies rejoice!].
pcmain() was also calling update_inventory() at start of play. That's
unnecessary for new games, where inventory initialization triggers a call
to it for each item added to your pack; but I wasn't sure about restored
games, so everybody gets it there now.
The Mac and BeOS ports evidently haven't been touched it some time;
they still referenced flags.move which got replaced by context.move quite
a while back. The Windows GUI code has a declaration for mswin_moveloop()
which appears to be non-existant, but I left it alone. I assume that the
Qt interface uses the existing main() routines; at least I couldn't find
any start of game code specific to it. vmsmain's revised main() is the
only one which has been tested.
Add new_mname/free_mname functions to make monster name handling be
more like the other extended data and to hide mextra details a bit more.
Add some casts where int and unsigned are being intermixed. Simplify
christen_monst(); it ought to be changed to have type `void' but I wanted
to avoid modifying another ten or so files.
Note: The CVS repository was tagged with NETHACK_PRE_MEXTRA
prior to application of this patch to allow easy withdrawal if necessary.
Adds a new mextra structure type that has a set
of pointers to various types of monster structures
including:
mname, egd, epri, eshk, emin, edog
Replaces the mextra bits in the monst structure
with a single pointer called mtmp->mextra of type
(struct mextra *).
The pointer can be null if there are no additional
structures attached. The mextra structure is not
adjacent to the monst structure.
Reduces the in-memory footprint of the monst that
has no other structures attached, at the cost
of adding 6 extra long ints per monster to
the save file
The new mextra structure has the mextra fields
independent of each other, not overlapping as was
the case with previous NetHack versions.
This patch doesn't do anything to capitalize on
that difference however.
Consolidates vault.h, epri.h, eshk.h, emin.h and edog.h
into mextra.h
Adds a macro for checking for whether a monster has
a name:
has_name(monst)
This fixes the magic trap panic
expels() -> spoteffects() -> dotrap() ->
domagictrap() -> tamedog()
because the monst no longer varies in size so no
replacement is required.
Remove several duplicate includes I discovered while reconciling the
vms Makefile. All of these are already being brought in via hack.h so don't
need to be explicitly included after it.
o Add support for zlib compression via ZLIB_COMP in config.h (ZLIB_COMP
and COMPRESS are mutually exclusive).
o rlecomp and zerocomp are run time options available if RLECOMP and
ZEROCOMP are defined, but not turned on by default if either COMPRESS
or ZLIB_COMP are defined.
o Add information to the save file about internal compression options
used when writing the save file, particularly rlecomp and zerocomp
support.
o Automatically adjust rlecomp and zerocomp (if support compiled in)
when reading in an existing savefile that was saved with those options
turned on. Still allows writing out of savefile in preferred format.
o In order to support zlib and not conflict with compress and uncompress
routines there, the NetHack internal functions were changed to
nh_uncompress and nh_compress as done in the zlib contribution received
in 1999 from <Someone>.
I tagged the sources NETHACK_3_5_0_PREZLIB prior to applying these
changes.
- always write plname into save file, no longer conditional
- add 'selectsaved' wincap option to control the display of
a menu of save files for ports/platforms that support it.
- add support for win32 tty using normal nethack menus.
- the win/tty/wintty code is generalized enough that any
tty port could support the option if the appropriate port-specific
code hooks for wildcard file lookups are added to src/file.c
specifically in the get_saved_games() routine. There is posix
code in there from Warwick already, and there is findfirst/findnext
code in there from win32. Warwick has the posix code only
enabled for Qt at present, but with wintty support, that could be expanded
to other Unix environments quite easily I would think.
Here is what the tty support looks like:
NetHack, Copyright 1985-2005
By Stichting Mathematisch Centrum and M. Stephenson.
See license for details.
Select one of your saved games
a - Bob
b - Fred
c - June
d - mine3
e - Sirius
f - Start a new character
(end)
The following files existed in the NetHack SAVEDIR directory
at the time:
ALLISONMI-Bob.NetHack-saved-game
ALLISONMI-Fred.NetHack-saved-game
ALLISONMI-June.NetHack-saved-game
ALLISONMI-mine3.NetHack-saved-game
ALLISONMI-Sirius.NetHack-saved-game
Note that despite the file names, the actual character name
is drawn from the savefile.
The WIN32CON support passes
USER-*.NetHack-saved-game
to findfirst/findnext where USER is your login name of course.
<Someone> wrote:
>> If the previous character was non-neutral, the unicorn would have
>> started off as hostile. When a bones file is loaded, I don't think
>> hostiles are made non-hostile, although the reverse is certainly true
>> (pets of the deceased are usually hostile).
>
> In the general case, they are, or rather their hostility is
> re-evaluated with respect to the new character; see the peace_minded()
> call in getlev(). However, co-aligned unicorns always being created
> peaceful is a special case in makemon(), _not_ part of peace_minded(),
> so they'd just have the usual chance of being made peaceful or not
> depending on alignment strength, as for any other co-aligned monster
> not explicitly declared always peaceful or always hostile.
on Sunday, April 4, 2004 at 20:27:06:
> On occassion when restoring a game where the
> character is wielding Sting, floor glyphs
> will show up before the --more-- prompt.
> These floor glyphs usually correspond to the
> location of monsters (sometimes they are just
> cavern features such as walls). Some of these
> floor glyphs are not in the character's line
> of sight upon restoring.
Also in this patch is a restore of Sting's ability
to glow blue.
This patch increments editlevel making existing save and bones files useless.
Add polywarn() code to grant the ability to detect certain monster
types while polymorphed into other specific monster types.
If you polymorph into a vampire or vampire lord, you are able to
sense humans.
And just for fun, if you polymorph into a purple worm, you are able to
sense shriekers :-)
Move the counter for the next attribute check to the context
structure.
This increments patchlevel so previous save and bones
files are unuseable after applying.
On September 11, 2003 "<Someone>" wrote:
> When we're going to have a different save file format, could
> the last messages in the message history be saved as well, so
> ^P would work the same before and after saving (possibly
> including a few less messages to make room for the startup
> messages?).
This seemed like a reasonable request. This patch:
- adds the core support required.
- adds the tty supporting routines.
Introduce a new set of functions to manage delayed killers in the trunk, used
in addressing the various reports of delayed killer confusion. Since existing
delayed killers are related to player properties, the delayed killers are
keyed by uprop indexes. I did this to avoid adding yet another set of
similar identifiers.
- the new delayed_killer() is used for stoning, sliming, sickness, and
delayed self-genocide while polymorphed. Some other timed events don't
use it (and didn't use the old delayed_killer variable) because they
use a fixed message when the timeout occurs.
- A new data structure, struct kinfo, is used to track both delayed and
immediate killers. This encapsulates all the info involved with
identifying a killer. The structure contains a buffer, which subsumes the
old killer_buf and several other buffers that didn't/couldn't use killer_buf.
- the killer list is saved and restored as part of the game state.
- the special case of usick_cause was removed and a delayed killer list
entry is now used in its place
- common code dealing with (un)sliming is moved to a new make_slimed function
- attempted to update all make dependencies for new end.c -> lev.h
dependency, sorry if I messed any up
Pat Rankin wrote:
> collect them all into some new struct and
> save that separately rather than jamming more non-option stuff
> into struct flags.
This patch:
- collects all context/tracking related fields from flags
into a new structure called "context."
It also adds the following to the new structure:
- stethoscope turn support
- victual support
- tin support
Bones loading was only checking to see if a
monster was marked extinct, it wasn't adding
up the born count of a species in the current
game with the number of that species on the
bones level being loaded. That made it possible
to exceed the correct number of nazgul and
erinys via bones.
This adds a common routine called propagate()
that makemon() and restmonchn(ghostly) share,
for incrementing the born count and checking for
extinction, etc.
When a bones level is loaded, restmonchn()
will flag an illegal monster (duplicated unique,
or too many of a species) by setting the
individual monster's mhpmax to the cookie
value DEFUNCT_MONSTER. Before getbones() finishes
loading the bones level, it will purge those
monsters from the chain.
<Someone> wrote:
>This _must_ be a bug: if a character leaves a pet corpse in a
>bonesfile, someone getting those bones will receive
>"So this is how you repay loyalty?" should he sacrifice it, even
>though the loyalty wasn't shown to _him_."
Clear the appropriate fields from the attached monst structure
when loading bones.
This patch gives game and savefile compatibility
whether GOLDOBJ is defined or not.
You can build with GOLDOBJ defined or not, and
still load your saved games. Rebuild with the
opposite, and load the same game.
That way GOLDOBJ can be experimented with
more easily.
1. Leave the "you" struct and the "monst"
struct the same under the hood between
GOLDOBJ and !GOLDOBJ.
2. Always write out gold as an
object on the player and monster
inventory chains.
On a restore of the savefile with GOLDOBJ
not defined, take the gold objects out of
the inventory chains and put it into u.ugold
or mtmp->mgold as appropriate.
On a restore of the savefile with GOLDOBJ
defined, nothing special is done.
- Version change from 3.4.x
- timed_delay feature ignore in makedefs
- several flags from iflags to flags
- use offsets from mons array entries in save file rather than storing
the ptr and calculating the distance from beginning of array
Incorporate a fix from <Someone> that resets the timestamp on
stinking clouds left in bones files. This does allow the cloud's ttl to
restart. Extended it to reset the player_inside flag.
Then noticed that the only placed that called in_out_region was domove().
Added calls when changing levels (which causes player_inside to be set
correctly if the cloud covers the location where the player starts on the
bones level), teleporting, and inside hurtle_step. There are still other
places that fail to call in_out_region, which others may choose to tackle.
I split out the remaining stinking cloud bug reported in the same e-mail
into a separate betabug. That one probably cannot be addressed without
changing the level file format to track the creator.
1) consolidate all core usage of `errno' in files.c;
2) give more feedback for any failure by create_levelfile or open_levelfile,
similar to what was being done for problems during level change;
3) include trickery info in paniclog (many instances of "trickery" seem to
be due to disk or quota problems rather than user misbehavior...).
The create_levelfile call in pcmain probably ought to be changed to use
error feedback, but in the meantime this should continue working.
Perhaps error() should be modified to update paniclog too, but I didn't
want to go through all its port-specific incarnations making changes.
Prevent the pardoning of trickery in wizard mode from attempting
to continue when there's no longer any current level. Also prevent
the ZEROCOMP configuration from trying to read from file descriptor -1
in case there're any other places which still let that slip through.
And fix an oddity in the VMS port's error() routine which has gone
unnoticed for years.
Several flags added since 3.4.0 were destined for flags
(to be saved with the game) but were placed in iflags for
savefile compatibility. These include:
boolean lootabc; /* use "a/b/c" rather than "o/i/b" when looting */
boolean showrace; /* show hero glyph by race rather than by role */
boolean travelcmd; /* allow travel command */
int runmode; /* update screen display during run moves */
This patch has no effect unless you define this in your port's
XXconf.h file.
#define SAVEFILE_340_CONVERT /* allow moving of some iflags fields to flags
without destroying savefile compatibility */
Without it, the new flags remain in "iflags." With it, the flags are moved to
"flags" and the structures are converted when the save file is read. There
is no reverse compatibility. If you save the game after conversion, you
can't load the savefile on 3.4.0, only 3.4.1.
Move get_saved_games() functionality to files.c
Use moved get_saved_games() functionality in Qt windowport.
[also some non-enabled perminv code in Qt windowport]
