Fix the bug From a bug report.alt.org server, where killing a monster by closing the
castle drawbridge resulted in a panic after the dead monster's possessions
were dropped into the moat and a potion of acid exploded in the process.
water_damage() deleted the object but had no way to tell flooreffects()
that it was gone, so flooreffects() couldn't tell its own caller not to
place and stack the object. After that, a chunk of freed memory became
part of the floor objects chain and eventually triggered a panic which
tried to make a save file but whose reason didn't get logged properly.
<email deleted> wrote:
> * You can't sell a candelabrum (with candles attached) to a lighting store.
You still can't sell the candelabrum after this, but acknowledge
the situation by having Izchak say something unique.
Provide a common routine that always does the right
thing with respect to timers and weight when altering
obj->corpsenm, and use it throughout the code.
From a bug report.c was first creating a corpse
complete with revive timer if it was a troll, or without
a timer if it was a lizard or lichen corpse. Then it might
change the corpse type, leaving strange corpse
types that revived, or that lasted forever.
We've been getting numerous complaints from people
about "dungeon failure", often related to attempts
to start NetHack from within various zip utilities
that present a folder-like view.
The dungeon failure was actually misleading. The
real problem was a dlb file open failure, but the
return value of dlb_init() was not being checked
in pcmain.
This moves the dlb_init earlier in the startup,
checks for failure, and provides some feedback
around the common zip utility problem for win32.
This fix should work for both DLB and non-DLB without forcing binary
mode for the latter. And it should continue to work even if we later
decide to force that mode (which I think we should do...). Use actual
file positions to calculate the size of the true and false sections,
rather than counting the bytes of each rumor (that counting has been left
in place but it gets overridden now.) Those two sizes will be bigger on
platforms which use CR+LF line ends and maintain dat/rumors as non-DLB
text file. But due to the way that nethack uses the sizes, such size
differences don't matter.
The branch variant shouldn't need any corresponding fix. It uses
the counting method when the output is binary, where the accumulated
value is accurate, and the check end-of-file shortcut when using text,
which should not be affected by stdio converting CR+LF into \n for text
input and back to CR+LF again for text output. (The original bug was due
to starting out with the EOF shortcut when input had CR+LF line ends, but
then writing less data due to binary output keeping just LF instead of
putting CR back.)
I've forgotten who pointed this out recently, but the hero was having
wisdom exercized (if true rumor chosen) or abused (for false one) whenever
level creation made a random floor engraving which used a rumor for its text.
To fix the gibberish rumor observed by at least a couple of players on
Windows (where if rumors.tru had <cr><lf> line ends and DLB was defined,
makedefs built a file that caused nethack to end up in the wrong place when
it reached EOF and tried to jump back to the first false rumor), change
the way that makedefs builds the rumors file. It now will collect more
information about the true and false rumors. Instead of one hex value on
the second line, there are nine values (three groups of three).
1,2,3;4,5,6;7,8,9
1 = number of true rumors; not previously collected and not currently used
2 = size in bytes for all true rumors (in decimal, as is #1)
3 = offset to first true rumor (in hexadecimal; not previously collected)
4,5,6 = same as 1,2,3 but for the false rumors section
7,8 = always 0 (imaginary third section has no entries, no size)
9 = offset to end-of-file (could be used for sanity checks; currently isn't)
Adding #2 with #3 yields #6; adding #5 with #6 yields #9. The old format's
lone entry was the same as the new format's #6. #2, #3, and #5 are values
which nethack was previously calculating on the fly after opening the file.
This also extends rumor_check() a little bit (displays the last rumor
for both sections in addition to the first), but I think it can probably
now be demoted to ``#if 0'' and removed from the extended commands list.
With makedefs changed to open the files as binary, the
rumors generation and retrieval seems to work consistently.
win32tty
T start=000067 (000043), end=020658 (0050b2), size=020591 (00506f)
F start=020658 (0050b2), end=043641 (00aa79), size=022983 (0059c7)
T 000067 A blindfold can be very useful if you're telepathic.
F 020658 "So when I die, the first thing I will see in heaven is a score list?"
win32gui
T start=000067 (000043), end=020658 (0050b2), size=020591 (00506f)
F start=020658 (0050b2), end=043641 (00aa79), size=022983 (0059c7)
T 000067 A blindfold can be very useful if you're telepathic.
F 020658 "So when I die, the first thing I will see in heaven is a score list?"
Provide a command to easily verify that the rumor true/false
boundary offsets are correct for the rumors file.
If the boundary is pointing mid-line, the rumor at the boundary
won't decrypt properly.
some rather complex boolean operations needed more parentheses to avoid
warnings. I think I put them in the right places.
A couple other items: naked assignments in if stmts, and an extra function decl
Axes can't dig through the floor, but attempting to dig down with one
shouldn't just "scratch the floor" if there is an armed land mine present.
Have it set off a land mine or bear trap, in addition to the possibility
of chopping your way out of being stuck in a bear trap (a recent change).
Add Hojita Discordia's Dungeon Map overview as
conditional code for experimentation and testing.
Everything is guarded by
#ifdef DUNGEON_OVERVIEW
#endif
The notes that accompanied the original patch follow.
Dungeon Map Overview Patch for Nethack 3.4.3
Version 3
=============================================================================
Changelist:
v3: Changed #level to #annotate to avoid #levelchange collision. Fixed
handling of elemental planes and astral plane (oops). Changed
formatting to be slightly closer to print_dungeon()'s. Should be
"final" version for 3.4.3.
v2: Added tracking of trees. Changed ctrl-m command to ctrl-o. Portals
displayed as "sealed" instead of "closed".
v1: First release.
(Note: all versions are mutually save compatible.)
=============================================================================
This patch creates a dungeon map overview that is recorded as the player
explores the dungeon. I was tired of returning to a game a few days later
and having no idea what the dungeon looked like. Trying to name pieces
of armor with shorthand didn't work so well as an intermediate solution
either, especially around nymphs.
It can be assumed that this map is in the mind of the hero and thus
can't be stolen, can be read when blind, or when buried, or when the hero
doesn't have any hands, or eyes, or hands free, or...etc. On the other hand,
this implies that the hero doesn't remember all of the details ("a fountain",
"some fountains", "many fountains") and that the map is subject to amnesia
when applicable.
This overview tracks fountains, altars, stores, temples, sinks, thrones,
trees, and dungeon branches. It attempts to not spoil the player nor
reveal more information than the hero knows. For this reason, it only
tracks dungeon features found in the guidebook and dungeon branches.
This patch breaks save file compatibility. Sorry.
Added commands
=============================================================================
#overview (ctrl-o, if not in wizard mode) - displays overview
#annotate (ctrl-n, if using numpad) - names current level
Example Output From #overview
=============================================================================
The Dungeons of Doom: levels 1 to level 15
Level 1:
A fountain
Level 3: (My stash.)
An altar, some fountains
Stairs down to The Gnomish Mines
Level 7:
Many fountains
Level 8:
Stairs up to Sokoban, level 7
Level 15:
A general store
Sealed portal to The Quest
The Gnomish Mines: levels 4 to level 7
Level 7: <- You are here
Many stores, some fountains, a temple
More Details
=============================================================================
The overview shows only levels that have anything interesting to display and
doesn't show branches that don't have any interesting levels.
To avoid the map revealing more information than the hero knows, the overview
only displays things that the hero has seen or touched. (If the hero
blinds herself, levitates above a known fountain, and obliterates it with a
wand of digging, the overview will still say that there is a fountain.)
This is done, sadly, by adding 6 bits to the rm struct to track the last
known dungeon type. On the other hand, this change could potentially allow
a window port to do something like drawing an item and a fountain on the same
square.
Things That Could Be Better And Maybe Some Feedback Would Help
=============================================================================
"<- You Are Here" is pretty goofy
-...but an indicator of some sort is nice.
=============================================================================
Many thanks to all the kind folks on r.g.r.n. who had very good feedback
about this patch, in particular L (for the trees), <Someone> Papaganou (for the
#annotate suggestion and some formatting feedback), and <Someone> (for the suggestion
of just overriding ctrl-o instead of using the very broken ctrl-m.)
=============================================================================
20060311. Hojita Discordia. (My usenet email is bogus. Sorry.)
From a bug report, a steed hit by disintegration
breath survived via life-saving, but the program was confused about whether
the hero could still ride and issued some impossible warnings. The code
to disintegrate a monster's inventory didn't bother to unwear worn items as
it destroyed them, presumeably assuming that the monster was sure to die,
so steed was left with a flag bit set claiming it was saddled even though
the saddle was gone. This fixes that, and the rider will end up being
dismounted as the saddle gets destroyed, regardless of whether the steed
ultimately survives. [Since the steed is still present at the time of
dismounting, the hero will get bumped to some other location, possibly to
the next spot about to be hit by the same black dragon breath attack which
just vaporized the steed. That's suboptimal, to put it mildly....]
I couldn't test the circumstances of the original report. Post-3.4.3,
ki-rin has been marked no-hands and is no longer capable of wearing armor
or amulets. I looked to see whether any other potential steed could wear
an amulet of life-saving and couldn't find one. But the original bug also
applied to conferred properties of special armor worn by non-steed monsters
too, so the fix was needed anyway.
The branch and trunk patches are different. For the trunk, I moved a
big chunk of code out of buzz() into a seperate routine. That actual fix
is the same in both variants.
Move part of the recent "munstone fixes" patch to the branch code
since one of those fixes prevents accessing freed memory. The part that
lets monsters eat tins of lizard meat or tins of acidic monsters in order
to get the same benefit as the corresponding corpse has been left out.
There were routines that were passed the
object name as an argument. Before the oextra
patch, ONAME() always returned a valid pointer
to a location within the obj struct. The oextra
patch worked around those cases by
using a temporary variable that was either set
to ONAME (if the obj passed the has_oname() test),
or to "" (pointer to an empty string) if no name was
present.
Since that might be a common thing to do, provide
the safe_oname() routine that you can use as a
function parameter without having to worry about
about whether ONAME(obj) is valid, and without
the need for the temporary variable.
It seemed inappropriate to allocate and tack on an oextra
structure just to mark it as unmergable, and the oextra
struct itself wouldn't be released until a save/restore took
place.
This uses one of the freed up oattached bits to prevent the merge.
The revised newmail() wouldn't compile (Strncpy doesn't exist, `buf'
was an array of pointers rather than of char). Simplify it substantially,
and adjust the one caller (vms) that relied on the old convoluted bit.
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.
<email deleted> wrote:
> * #chatting to hostile prisoners gives "This will teach you not to disturb me!"
> This only makes sense for djinn and water demons.
I don't have a clue what they should say, this only deals with
what they shouldn't say.
<Someone> wrote:
> When applying a drum of earthquake, should you "fall into a chasm" if
> you're already at the bottom of a pit? (Likewise monsters.)
monster already in a pit won't "fall into a chasm" from drum of earthquake
<Someone> wrote:
> When applying a drum of earthquake, should you "fall into a chasm" if
> you're already at the bottom of a pit? (Likewise monsters.)
hero already in a pit won't "fall into a chasm" from drum of earthquake
From a bug report, a monster who eats a lizard
corpse in order to cure confusion was treated the same as one who did so
to cure petrification, losing intrinsic speed in the process. In the same
report by <l>, monsters wouldn't eat lizard corpses to cure being stunned,
and those who ate them for another reason weren't cured of stunning, even
though the hero gets that benefit. While fixing those, I added some code
to let monsters who are carrying tins of lizard or acidic monster use them
if they're also carrying a tin opener, dagger, or knife. I don't think
any monsters except for nymphs are willing to pick up tins, so it won't
have much effect. It now works for nymphs though.
Examining the code while testing showed that mon_consume_unstone()
has been accessing the potion (acid) or corpse (lizard or acidic monster)
after the item had been used up, so that has been fixed too. I never saw
any detectable problems due to this, but folks using a debugging malloc
implementation which overwrites freed memory may have not been suffering
collateral acid damage or receiving intended confusion cure, or perhaps
did get either or both of those effects when they shouldn't have. Since
it only applied to monsters it wouldn't have been easy to observe.
<email deleted> wrote the following on April 10, 2006:
> When blind and levitating, you can produce a message - "You try to feel what
> is lying here on the ice. But you can't reach it!" - which reveals too much
> about the terrain below you.
<Someone> suggested that digging down on a land mine with a pick-axe ought
to set if off. I agree; this implements that and also for bear traps. In
the bear trap case, if you dig down once trapped, you will destroy that
trap explicitly rather than replace it with a pit, so it's now possible to
escape from one without leaving another trap in your wake. Once the bear
trap is gone, further digging there will make a pit as usual. While stuck
in one, digging down poses a modest risk of harming yourself.
|You now wield a pick-axe.
|You start digging downward. A bear trap closes on your foot!
|You start digging downward. You destroy the bear trap with your pick-axe.
|You continue digging downward. You dig a pit in the floor.
|You start digging downward. You dig a hole through the floor.
|You fall through...
[It seems a bit strange that finishing a pit discards all digging context,
so that resuming within the pit in order to make a hole "starts" digging
rather than "continues" it.]
Digging down with a wand or spell will disarm these two types of traps
and then leave the corresponding object (which may or may not fall through
the resulting hole, like any other object there). Digging to the side via
magic while trapped in a pit will also disarm such traps when it encounters
them. (When not in a pit, a digging beam which simply passes over an armed
bear trap or land mine won't have any effect on the trap.) Digging to the
side via tool behaves somewhat oddly ("no room for the rubble"?) and will
probably need some tweaking before eventual release; at present it doesn't
reach adjacent traps so didn't need any land mine or bear trap handling.
I put the fixes entry in the new features section.
<Someone> reported that thitu() was adding d20 damage for silver object
hitting silver-hating hero even though all the callers were using dmgval()
which also does that, resulting in doubled silver bonus/penalty. This
fixes that (including for boomerangs thrown by player, which weren't using
dmgval(), to handle a hyptothetical silver boomerang). While testing it,
I noticed that there was no "the silver sears your flesh" message when a
monster hit you with a wielded silver weapon, so this fixes that too.
(How did we miss that? And how did <Someone>? :-)
