the message history saver needs the windows to stick around until save
is completed. This usually the case, but end_of_input() was calling
exit_nhwindows prematurely. This is actually an old bug (calling it early,
but wasn't harmful until other recent changes).
Use a new flush_screen(-1) call to toggle 3rd screen update in goto_level.
This keeps the 3rd screen state unchanged, no matter what happens at higher
levels, ensures the map window cursor stays on the hero, ensures the
hero's showrace colors remains bright white, and so on.
<Someone> suggested and I agree (mostly, if you're a Monk poly'd to a
strict carnivore, I think it should still smell delicious):
> "You smell the odor of meat." (if herbivorous)
> "You smell a delicious smell." (otherwise)
> I think that if your character is a monk or still veg(itari)an, you should
> also get the first message. Even if you're not intentionally vegitarian,
> the first message is still appropriate.
Reported a really long time ago (June 2001) by <Someone>:
- stand on the upstairs and engrave Elbereth with a /oFire
- create Demogorgon and e.g. a tame dragon
- cause conflict
- Dragon will kill Demogorgon and Demogorgon will never ever attack Dragon.
All monsters could still attack Demogorgon without response via fightm().
Modified fightm() to include a bit of code in m_move and dog_move, allowing
response to an attack. Testing this in action, Demogorgon still usually
did things detrimental to the player, mostly summoning nasty monsters.
To reproduce: disable lit_corridor, enter a corridor, read a scroll of
light, then read a cursed scroll of light. The corridor won't darken again.
A test of flags.lit_corridor in newsym was at fault. I'm guessing this was
never noticed because lit_corridor is on by default on Windows and the
default lit and unlit corridor symbols are the same on most other platforms.
Cutting a shopkeeper poly'd as a long worm would generate strange messages
and could result in a crash. cutworm didn't deal with all the intricacies
of duplicating a monster. Fixed by changing cutworm() to use clone_mon()
to do most of its dirty work. It seems to me that without this change,
cutting a tame long worm could also have similar bad effects.
Other side effects of this change:
- clone_mon now takes x,y coordinates, 0,0 results in previous behavior
- clone_mon no longer always makes cloned monsters tame/peaceful if player
caused the clone, using the same formula previously in cutworm. Someone
else may wish to tweak this for gremlins.
- clone_mon will christen the new mon with the old shopkeeper's name, even
though clones are never shopkeepers (game can't handle 2 for a shop)
- cutworm can now be called during conflict or pet combat, although I
added no such calls (yet)
I noticed a few panic messages contained newlines, and one included a naked
carriage return. panic() adds a newline itself, and also generally ensures
the message will be on a new line (the initial "oops" ensures the message
itself will be on a new lines). This patch removes the unneeded characters.
Incorporate a slightly modified version of a patch submitted by <Someone> back in June. The basic problem, which I noticed again this past
weekend, is that pets with a high apport value can still try to get to
objects when there's no path to the object. The patch extends the
can_reach_food function to be used for any object, renaming it to
can_reach_location, and adds a could_reach_item function for doing point
checks. This also removes any chance of a pet dragon, for example, eating
something underwater.
When changing levels, the state of rndmonst() is reset, causing the monster
choices to be recalculated. However, the frequency counts for initial
uncommon() monsters were never cleared. Thus, if the first non-extinct
monster were a hell monster, and you returned to the main dungeon, the hell
monster will remain in the list and could be selected.
Calling rehumanize directly when u.mh > 0 doesn't consider Unchanging
(perhaps it should?). But, it's probably better to call losehp anyway.
Also, part of a buglist item: .5PD didn't affect rust trap damage on iron golem
I wrote to the devteam early last week:
> Given my understanding of travel, it's supposed to be somewhat intelligent,
> and "convenient", and should, therefore avoid walking into water, lava,
> traps, or other things that distant movement would avoid, even if you're
> right next it. Unless... the travel destination is the "bad" location
> next to you when the travel starts.
To that end...
- add a context (iflags in 3.4.3 to maintain savefile compat) flag to
differenciate the first travel step from later steps, to allow the
detection of the final sentence, above.
- several changes to set/reset the travel1 flag as needed
- add code to findtravelpath to treat the first step specially if it's
the only step, allowing forced travel into a "bad" location
- correct the "don't travel over traps" code, which was getting confused
because hero's starting location was being avoided
- add code to avoid traveling into water and lava, duplicating
checks used for non-travel running
- fix some strange "guess" travel behavior: avoid zigzag paths when there's
a more direct path (even though the number of moves is the same)
- trunk change adds a new DISP_ALL tmp_at type, and uses it in some debug
code for travel, debug changes not added to the 3.4.3 branch
This hack simply causes ^C to be ignored completely by the gnome interface.
Gnome really cannot handle interrupts, and the window port interface does
not currently provide a way for the gnome interface to just quit. Allowing
the gnome interface to prompt would require a complete overhall to the the
way window ports work.
Allow migrated objects to break on arrival. Added code to obj_delivery to
cause this, along with a flag to keep breakage from occurring. The new
flag isn't used yet, because all the current object migration involve
objects that were moving/dropping. To help make this change, rloco now
returns whether the object was placed or not, so caller can know if an obj
pointer is still valid or not.
Making the breakage messages for MIGR_NEAR_PLAYER objects show up after the
new level is displayed required some effort (rather than while the old level
was still displayed, which was confusing), due to the needs of goto_level.
- obj_delivery now has 2 passes, one for before player arrives, another after,
allowing the two cases to be treated differently
- goto_level calls obj_delivery twice (run_timers is not called twice,
since the run required before the level is displayed will have already run
any timers on migrating object)
- kill_genocided_monsters now kills eggs on the migrating_objs list too
Use fqname buffer 1 for restoring the save file (just like save does when
creating it) so the value won't change out from under the code in unixmain.
- Also moved a tty-specific hack in docompress_file that was causing
the 'y' response to the "keep the save file" prompt to be echoed twice.
Not all of the extended commands worked in the gnome interface because
'#' just caused the next character to be treated as a meta character (this
was a hack I added a while back when none of the extended commands worked).
Resolved by finally adding an extended command menu to the gnome interface.
- updated some formatting so I could read the code
- fixed startup player selection menus so accelerators work
- added necessary calls to make sure selected menu item is visible
- also removed some dead code
The tty menu code wasn't totally reentrant, causing it to free memory
that wasn't on the heap if you had your inventory displayed, ^C then
ask to see the inventory again. Solved this by converting the buffer used
by process_menu_window to be heap-allocated. If the Quit code could return at
this point, this would still be very bad, but since it doesn't, this is OK.
Internals of gnome_yn_function should treat key inputs as 'int' until
returning them, to avoid truncating and not sign extending as needed on
platforms where char is unsigned.
Add a menu item for Quiver. While testing this, found the Quit menu
item still said that you quit using shift-q, so I updated that too.
Trunk change includes some other exit related fixes that don't apply to
the 3.4.3 branch.
cosmetic changes to allow qt_win.cpp to compile cleanly using gcc -Wall.
This corresponds to a debian report that doesn't mention a debian bug ID.
The changes are not identical to the suggested debian patch.
Too much code was being shared between the extended command window
and the popup dialogs for player selection, causing an uninitialized
variable to be referenced, often resulting in a core dump.
Add support for hilite_pet to X11 text map mode (hilite_pet was already
supported when tiles were enabled). While testing this, I found a missing
newsym() in the code implementing the creation of a "tame" monster.
<Someone> suggested that a dragon, lacking hands, shouldn't be able to
throw things. Dragons can pick things up, and it seems they can throw
things the same way. However, it does seem that a notake() monster, which
cannot pick things up, shouldn't be able to throw things either.
Reported a while back, a (stonable) hiding monster will hide at a location
containing only a cockatrice corpse. While it would be interesting to
allow monsters to try, and stone themselves as a result, I chose the
simpler fix which is to not have monsters hide in such situations. I found
the hiding code was duplicated in several places, so I moved it into a new
hideunder() function that works for both the hero and monsters.
When moving the code to bill for tins before eating them (to charge for
them at the pre-eating hunger status), several cases were able to bypass
billing. Moved the bill checking code to a new costly_tin() subroutine so
it could be called in all the necessary places.
Prompted by a question from Pat a long time back, this change finally allows
tiles or text map mode to be chosen dynamically at runtime (using the
"tiled_map" option) rather than having to pick it via an X resource and
keep your selection until you exit. This brings map mode selection up to a
level similar to most other graphical window ports.
In addition, the map mode automatically switches to text on the Rogue
level, also like other graphical window ports.
The default mode for the X11 binary is now tiles, once again, like most (all?)
other graphical window ports.
The patch also removes some dead X11 code that is unlikely to be useful again.
<Someone> reported that randomly placed aquatic monsters can end up in
lava. The placement code allowed lava whenever the WET flag was passed to
it. This was so passing (WET|DRY) would match all locations, but it's not
appropriate for when only the flag WET is used. Since we have no levels
currently affected by this bug, I fixed it only in the trunk.
<Someone> suggested that nagas should eat. This seems reasonable.
I made all but the black omnivores. The black are similar to the AD&D
spirit naga, which are listed as carnivores, while AD&D's other nagas are
omnivores, so I took the same approach.
Change the code involved to use a new #define for the field, like other
overloaded uses of corpsenm. Also remove one check for this field from
potion.c: there are no blessed potions from sinks.
There was some unreachable code in dokick related to drawbridges. Since I
liked the current "Ouch!" behavior, I moved the drawbridge test inside the
IS_STWALL code but made sure to update the maploc so kickstr would return
the right thing. Since there may be more than one drawbridge (perhaps it
should test for Valkyrie?) changed the kickstr prefix for drawbridge to "a".
If you try to put a saddle on a cockatrice, you turn to stone,
even if wearing gloves. This seemed inconsistent with other cases,
and I couldn't think of a reason why you'd touch the cockatrice with
another body part, so I added the uarmg test.
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
It makes sense that if you cannot normally move to a location containing a
boulder, dismounting there should be avoided if possible. It seems
that it should be possible to land there if there's no other choice.
Perhaps doing so should result in damage, but that's not in this patch.
<Someone> pointed out the inappropriate appearance of the 3.4.2 message when
fog clouds have you engulfed if you happen to be polymorphed into an aquatic
creature. It does seem that a fog cloud should not cause added damage to
amphibious creatures, which includes breathless (but non-flaming) monsters.
I also thought another message was more appropriate for flaming creatures
(see on_fire()), and added a special case for that.
gulpum was missing symmetric code and I added it there too.
This change adds a new flaming() macro and uses it in several places
where the list of flaming monsters was tested. on_fire() didn't list
salamanders as already being on fire, but should have. A couple other
cases were not updated to include flaming sphere.
While wearing the Eyes of the Overworld, one could use a hook or pole thru
walls. The bug report thought they should work past a boulder, but I
disagree, given a pole or hook's rigid nature, and did not special case that.
couldsee() is not affected by the Eyes, so use that after checking cansee().
Incorporate various killer message grammar fixes suggested by <Someone>.
Mostly these deal with using the proper killer_format and prefix ("the" or
no "the") for unique and type_is_pname monsters, or death to to eating
their corpses.
One case is handled by a general fix to name_to_mon to allow it to deal
with "Croesus' corpse".
The pre-3.4.1 topten behavior for "starved to death" messages is also restored.
Some changes for standard C platforms, to avoid declaring errno explictly.
Such platforms should declare errno in errno.h, which is already included
in the files in question.
Nethack's manes are based on AD&D manes which are in turn based on the
manes of Roman legend. They are supposed to be spirits of the dead.
To that end, added them to the nonliving() macro. The biggest behavioral
change is that death spells no longer effect them, which does technically
make them a bit tougher but also makes sense. Also, they're so wimpy, it's
hard to believe anyone would use a death/disintegration on them anyway.
The choice was to avoid one of the two sellobj() calls in in_container.
Since I liked the message ordering "put gold into... credit" better than
"credit... put gold into", the code now avoids the 1st sellobj call for coins.
This is an initial round of SAFERHANGUP hangup changes. It introduces
SAFERHANGUP, provides the core framework, and enables it for UNIX.
Window-port changes are provided for win/tty, win/X11 and win/gnome. Qt
changes should be forthcoming after having Warwick look at them.
window.doc is updated so windowport maintainers have an clue what needs to
be done to support SAFERHANGUP.
As suggested in a message from <Someone>, add a rumor and an oracle
regarding priestly donations. Keni's suggested wording was incorporated.
A spelling error in another oracle is also fixed.
This patch introduces a change to yname() and Yname2() that avoids the
possessive "your" for the hero's normal, fully identified artifacts.
Quest artifacts still get the possessive, as do all other objects and all
objects not in the hero's possession. shk_your()/Shk_Your() are used in
many places with a specific, generalized name for the object, so I didn't
introduce the artifact behavior there, although I did change them to append
a space, which simplified some other code. Through added use of yname(),
there may be some places that used to just say "corpse" that will now be more
descriptive via yname()'s use of cxname(). I'm sure <Someone> will point
out any such places that are too onerous, although nothing obviously is.
I took the opportunity to inspect many uses of "your" and even Your(). Two
new functions are also introduced, yobjnam() and Yobjnam2(), which work
like aobjnam() and yname() combined, because I found that many uses of
aobjnam() were preceeded by "your" and I couldn't generally provide the
desired behavior for artifacts (or future artifacts) without a combined
function. In some cases, this change allowed better sharing of code.
rust_dmg() still takes a string as input which is sometimes initialized
from xname() and often prepends "your" to it. Currently, this isn't a
problem since there currently are no normal, armor artifacts. If/when any
are introduced, rust_dmg() will need to be addressed.
The patch is for the trunk only. A lot of research was required and I
didn't feel the upside was there for repeating it in the 3.4.3 branch.
A while back, I delayed applying this patch after a discussion about
quivering coins, because I didn't want to change the behavior of GOLDOBJ
vs non-GOLDOBJ games. I'm tired of seeing this diff in my tree, and I
recall there was some sentiment that I should have checked it in, so I'm
adding the feature in the trunk.
Quoting the buglist: "If shk.c does get modified, I'd
like to see a call to rouse_shk() added at the start of rob_shop().
If the shk wakes up for buying and selling, he ought to wake up for
robbery--even when it's covered by the customer's credit--too."
A recent report noted that if you are invisible, teleport into a shop,
steal something, teleport out, and then return, the shopkeeper will become
pacified. This is because the pacification code occurred even if the
shopkeeper already knew your name, as long as the "Welcome" message had not
yet occurred. Implement <Someone>'s proposed fix: Before pacifying, either
the visitct must be zero (as per old code) or the customer name must
already be known. And, of course, the customer name must differ from the
current setting. This skips the pacification code in those situations
where the shopkeeper learned your name but was not happy about it (which is
what visitct > 0 implies, at least initially).
This also deals with an older bug report where you attack a shopkeeper
while outside the shop and then later enter the shop while the shopkeeper
is in it. hot_pursuit() sets the customer name, and the new check avoids
pacifying the shopkeeper for the same reason as noted above.
rewrote the message so it doesn't talk about his body dissolving into gas,
since there's a possibility that he'll leave a corpse. It now doesn't
mention a body at all, so it's not so bad when he doesn't leave a corpse
either. The contents are completely changed, and I think it's more in line
with the sort of thing you'd expect from Master Kaen.
remove_rooms() was trying to be clever and truncate the maximum room bounds
in partially overlapping rooms. In the process, it would truncate the
bounds of L shaped rooms incorrectly, leaving some locations outside the
bounds with the roomno field set. Since the maximum bounds really do not
have to actually match the largest extent of the room in the case of
irregular rooms, it was easiest to just remove the code that was performing
the incorrect truncation. Due to the way remove_room() is coded, this
could result in shopkeeper messages (or, less likely, priest messages,
although that has not been reported yet) occuring for locations to the
right (or the upper/lower right corner) of the MAP on the Bustling Town level.
correct the region where the down stairs can be placed so they don't show up
in a cut off cave. If/when the level generator is improved to avoid generating
such caves, the down stairs region could potentially be changed back.
<Someone> forwarded from rgrn that a monster using Stormbringer would
lose hp when hitting the player. The sign of the "gain" was reversed due
to subtracting the difference in the wrong order.
Build and install the NetHack.ad file, and change the nethack.sh to append
the HACKDIR to the XUSERFILESEARCHPATH. This means users do not have to
do anything special to use the NetHack X11 resources, nor does it require
teaching nethack where the X11 app-defaults directory is. The change also
updates the X11 and linux doc files in a corresponding way.
I think this was From a bug report, but I cannot find the report
anymore. Perhaps I noticed it while doing something else. In any case,
every turn you're engulfed in, say, a dust vortex wearing the Eyes of the
Overworld, you get the "can't see"/"vision clears" message pair. This
patch modifies the behavior to only display this the first turn. Other code,
partially related to B01003, already deals with the puton-remove cases.
Since B01003 is only in the trunk, this fix is also only in the trunk.
Correct the unreachable "snores" message From a bug report.
Applied to trunk only in hopes that beta8 might be the last for 3.4.2.
The fixes entry is generic since I'm sure other similar messages will be fixed.
Digging a pit in the location where you're stuck should always free you.
This would not occur if you used a wand of digging to do the digging,
especially if you did so while levitating.
Back in April, <Someone> reported to the list that when you are
polymorphed into a mimic and #monster, and return to human form while still
mimicing, your appearance does not change. This was due to a change
between 3.4.0 and 3.4.1 that caused all mimicing to go thru one case and
the code to change appearance was actually never executed.
Several bug reports have been filed regarding the use of attached iron balls to
move in ways that are otherwise disallowed: moving diagonally from a
doorway, moving through boulders, squeezing through small openings, passing
over traps and diagonally between boulders on sokoban levels. Modified
bhit to deal with this. All these cases now cause the iron ball to stop if
appropriate for your form, level of burden, and so on. Since both boulders
and iron balls are big, boulders now always stop thrown iron balls,
even if not attached. Since iron balls don't cause any damage as they pull
you along, I didn't add any damage for the new "jerks to a halt" cases.
Undo the change to test_move and instead, don't even test moving in diagonals.
This opens the possibility for approximating the diagonal while polymorphed
into a grid bug in the future.
the cmdassist message shows all the valid directions even if you are
polymorphed into a grid bug. I noticed this when I typed a diagonal
direction as a grid bug and got the dialog, telling me the key I just typed
was valid, but it wasn't. Limit the keys to those valid for grid bugs.
Since only developers know that "13" is EACCES, try to include the text
message. I'm not 100% sure the ifdefs are complete, but it can be tweaked
as needed. This was the only common message in files.c that included errno
so it's the only one I changed. Of course, "13" is only one of several
possible errno values that might reasonably show up here.
- If you have Gnome installed on solaris, the GETRES support wouldn't build.
I don't have access to a solaris system with Gnome installed, but hacked
unixconf.h to force the GETRES code itself to be compiled. So, I believe the
unixres.c change will work for folks really using Gnome on Solaris.
Whether the rest of the gnome code will build there is beyond me.
- I accidentally left TIMED_DELAY defined and the Solaris build failed.
While solaris has usleep(), this is not part of SVR4 as far as I can tell.
But, SysV does have poll, so I implemented msleep() for SysV systems in
terms of poll. So, you can now define TIMED_DELAY on any SYSV build.
<Someone> thought the incorrect grammar in the messages was due to scatter()
changing the quan of treefruit, but in fact, it was due to treefruit
referring to the wrong object (there is also a theoretical possibility that
treefruit can refer to a merged object with the wrong count). Create a temp
object for use in the message to avoid these possibilities.
If you dug in a pit next to a sleeping, angry monster, you'd stop every
turn due to a complex check at the end of dochugw. It turned out this
was due to a long-standing bug in the special case vision code that failed
to set the COULD_SEE bit for the locations where it set the IN_SIGHT bit.
It looks like the underwater code had the same problem (it didn't set the
bit, obviously there are no pits underwater). However, the same could
occur if you see the angry, sleeping monster with Xray vision. In this
case, setting COULD_SEE is not appropriate, so added a mcanmove check to
the complex check in dochugw.
Avoid ever putting an "I" on the hero's location by checking it in
map_invisible(). It appeared there were a few other special cases that
could call map_invisible() for actions involving the steed, so checking there
catches them all.
Finally got around to installing OpenBSD (rev 3.3) in a vmware partition.
Found that several #if BSD's were inappropriate for modern BSD's. Haven't
installed FreeBSD or NetBSD, but based on reading their man pages,
these changes are needed there too. Mostly due to POSIX time() signature.
Since cansee() is false for all locations while swallowed, need to test if
the monster being hit is the one that swallowed you to ensure that various
artifact hit messages, including the DRLI message, are printed.
cohrs