If one or more boulders were next to lava and hero broke a wand of digging
next to that location, the boulder(s) stayed over the lava causing a sanity
checking error.
- Add a vision sanity checking routine
- Recalc block point when digging a door for temporary clouds
- Add recalc_block_point after cvt_sdoor_to_door, because doorways
on the Rogue level have no doors, and otherwise the sanity checking
would complain. This doesn't actually change how the Rogue level
vision works, as it uses a different vision system
- Monster using a trap in a secret corridor revealed the corridor,
but didn't unblock the vision unless you saw the location
The ones below marked "left,revisit" might be worthy of code changes
explode.c:removed: arg sx,sy passed to breaks() are coordxy and so are the parameters
invent.c: removed: discover_artifact() parameter is xint16, so change cast to match
mail.c: removed: gv.viz_rmin[] contains coordxy's, and enexto() 2nd param is too
mkmaze.c: removed: x, y are coordxy and so are cc->x, cc->y
mkroom.c: removed: tx, ty are coordxy and so are parameters to occupied()
mplayer.c: left,revisit: x,y declared int, but mk_mplayer() params are coordxy
sp_lev.c 2890: left alone: x,y declared int, but is_ok_location() params coordxy
sp_lev.c 4114: left alone: cast from lua_Integer
sp_lev.c 4123: left alone: cast from lua_Integer
sp_lev.c 4650: left alone: cast from lua_Integer returned from lua_tointeger
sp_lev.c 4765: left alone: cast from int_Integer returned from lua_tointeger
sp_lev.c 4772: left alone: cast from int_Integer returned from get_table_int
teleport.c 366: left alone: cast from int returned from rn2()
timeout.c 2171: left alone: cast from long
vault.c: left,revisit: guardx,guardy declared int, but bestcc fields are coordxy
worm.c 791: left,revisit: wseg.wx, wseg.wy are coordxy, so are nx, ny, but not
ox, oy; why are ox, oy needed at all?
worm.c 955: left alone: wx, wy are coordxy, wseg_at() params are int x, int y
zap.c 5061: left alone: cast from long
Using 'f', if hero is wielding a polearm, and a monster is in range,
don't switch away even if we do have ammo in quiver and a launcher
in the inventory.
gcc has recognized various "magic comments" for white-listing
occurrences of implicit fallthrough in switch statements for
a long time:
The range and shape of "falls through" comments accepted are
contingent upon the level of the warning. (The default level is =3.)
-Wimplicit-fallthrough=0 disables the warning altogether.
-Wimplicit-fallthrough=1 treats any kind of comment as a "falls through" comment.
-Wimplicit-fallthrough=2 essentially accepts any comment that contains something
that matches (case insensitively) "falls?[ \t-]*thr(ough|u)" regular expression.
-Wimplicit-fallthrough=3 case sensitively matches a wide range of regular
expressions, listed in the GCC manual. E.g., all of these are accepted:
/* Falls through. */
/* fall-thru */
/* Else falls through. */
/* FALLTHRU */
/* ... falls through ... */
etc.
-Wimplicit-fallthrough=4 also, case sensitively matches a range of regular
expressions but is much more strict than level =3.
-Wimplicit-fallthrough=5 doesn't recognize any comments.
Plenty of other compilers did not recognize the gcc comment convention,
and up until now the compiler warning for detecting unintended
fallthrough had to be suppressed on other compilers. That's because the code
in NetHack has been relying on the gcc approach, and only the gcc approach.
The C23 standard introduces an attribute [[fallthrough]] for the
functionality, when implicit fallthrough warnings have been enabled.
Several popular compilers already support that, or a very similar attribute
style approach, today, even ahead of their C23 support:
C compiler whitelist approach
--------------------------- -------------------------------------
C23 conforming compilers [[fallthrough]]
clang versions supporting
standards prior to
C23 __attribute__((__fallthrough__))
Microsoft Visual Studio
since VS 2022 17.4.
The warning C5262 controls
whether the implict
fallthrough is detected and
warned about with
/std:clatest. [[fallthrough]]
This adds support to NetHack for the attribute approach by inserting a
macro FALLTHROUGH to the existing cases that require white-listing, so
other compilers can analyze things too.
The definition of the FALLTHROUGH macro is controlled in include/tradstdc.h.
The gcc comment approach has also been left in place at this time.
I recently realized that I've been editing sources in a terminal
window that was widened in order to fit curses borders for testing
something or other. That has resulted in some new wide lines in the
source. There were lots of old ones too.
This updates some source files to try to achieve the goal of 78
characters or less. As in the past, I've been inconsistent about
lines with 79 characters. Lines with 80 or more have been wrapped
or shortened (usually by trimming an end of line comment or removing
redundant parantheses, sometimes just by reducing the indentation
of the continuation portion of an already wrapped line).
I eliminated one instance of warning manipulation for non-constant
format string, and simplified stone_luck() where Ken had a silly
comment about the function argument's name.
> If there's a trap on a no-dig level, the floor beneath it will always be "too hard to dig into", making it impossible to remove the trap.
>
> As you can still dig pits in these levels (just not holes), the floor under the trap itself resisting to become a pit seems inconsistent.
>
> Steps to reproduce:
>
> Go to no-dig level like Mine's End
> Make a trap
> Dig a pit next to it -> works
> Dig on the trap -> does not work
Return more information about the dig_check() results to caller (was
just a boolean).
Move the messaging that was in dig_check() into a separate
digcheck_fail_message() function that uses the expanded return
information.
Resolves#1254
.
The g? structs had a mix of variables that were written to
the savefile, and those that were not.
For better clarity and to distinguish those that end up in
the savefile, relocate some g? variables that get written
directly to the savefile into different structs.
This updates EDITLEVEL, although technically it probably
didn't need to, since savefile contents are not changing.
Details:
gb.bases -> svb.bases
gb.bbubbles -> svb.bbubbles
gb.branches -> svb.branches
gc.context -> svc.context
gd.disco -> svd.disco
gd.dndest -> svd.dndest
gd.doors -> svd.doors
gd.doors_alloc -> svd.doors_alloc
gd.dungeon_topology -> svd.dungeon_topology
gd.dungeons -> svd.dungeons
ge.exclusion_zones -> sve.exclusion_zones
gh.hackpid -> svh.hackpid
gi.inv_pos -> svi.inv_pos
gk.killer -> svk.killer
gl.lastseentyp -> svl.lastseentyp
gl.level -> svl.level
gl.level_info -> svl.level_info
gm.mapseenchn -> svm.mapseenchn
gm.moves -> svm.moves
gm.mvitals -> svm.mvitals
gn.n_dgns -> svn.n_dgns
gn.n_regions -> svn.n_regions
gn.nroom -> svn.nroom
go.oracle_cnt -> svo.oracle_cnt
gp.pl_character -> svp.pl_character
gp.pl_fruit -> svp.pl_fruit
gp.plname -> svp.plname
gp.program_state -> svp.program_state
gq.quest_status -> svq.quest_status
gr.rooms -> svr.rooms
gs.sp_levchn -> svs.sp_levchn
gs.spl_book -> svs.spl_book
gt.timer_id -> svt.timer_id
gt.tune -> svt.tune
gu.updest -> svu.updest
gx.xmax -> svx.xmax
gx.xmin -> svx.xmin
gy.ymax -> svy.ymax
gy.ymin -> svy.ymin
Related note:
There are some pointer variables that are heads of chains that were not
moved from 'g?' to 'sv?', because they are not actually written to the
savefile directly, but the objects/monst/trap/lightsource/timer in the
chains they point to are. That can be changed, if desired.
Examples: gi.invent, gm.migrating_objs, gb.billobjs, gm.migrating_mons,
gf.ftrap, gl.light_base, gt.timer_base
Replace several upstart(y_monnam(mon)) with new YMonnam(mon) to
produce "Your little dog" and such.
Also change one or two Monnam(mon) to YMonnam(mon) and one pline(...)
to pline_mon(mon, ...).
I still haven't found any explanation for the report by a hardfought
player recently that going down some stairs with a pair of leashed
pets got one into a confused state where it was flagged as leashed
but the corresponding leash was no longer in use.
This adds some new object and monster sanity checks regarding leashes,
and it changes o_unleash(obj) to clear obj->leashmon even if/when the
monster can't be found.
It also changes behavior for dipping an attached leash into a potion
of polymorph when that happens to yield another leash--now the new
one will end up being pre-attached.
In 3.6.2 parts of the wakeup code were merged together, and this
caused pets consider any noise made by the hero - such as hitting
iron bars or digging - as whistling for them to come to the hero.
Change it to only consider actual whistling and ringing a bell.
Note: Original change is from xNetHack by copperwater <aosdict@gmail.com>,
but this commit comes from HACKEM-MUCHE by Erik Lunna, with
some minor code formatting.
From xNetHack commit a0a6103bea:
'The original goal: nerf item destruction using a method I initially
proposed for SpliceHack, in which the number of items subject to
damage from any single source is limited by the amount of damage the
effect caused. The intent was to be more fair all around and prevent
aggravating situations where, for instance, a chest shock trap zaps
you for 4 damage and immediately ten of your rings and wands blow up.
Problem 1: no easy way to limit the items destroyed without biasing
heavily towards the start of the invent chain. The old code was able
to get away without bias by just indiscriminately destroying
everything eligible with a 1/3 chance. Here, I had to introduce
reservoir sampling in a somewhat more complex form than I've applied
it elsewhere, since there are a pool of potential items.
Problem 2: destroy_item no longer worked remotely like destroy_mitem,
which still destroyed 1/3 of items indiscriminately. Commence the
process of squishing them into one function that handles both the
player and monsters. (Which required making a lot of adjustments to
destroy_one_item, now named maybe_destroy_item, on nits such as
messaging and when to negate damage. An annoying consequence of the
merge is that in the player case, their HP is deducted and they can
be killed directly, but for monsters they need to add up the
destruction damage and return it.)
Unifying destroy_item and destroy_mitem has some advantages: in
addition to the obvious code duplication removal, it ensures monsters
now take the same damage as players for destruction (previously they
took a piddly 1 damage per destroyed item). Now when you hit
something with Mjollnir and their coveted wand of death breaks apart
and explodes, you at least get the satisfaction of knowing they took
the standard amount of damage from it. Monsters also now get
symmetry with players in having extrinsic elemental resistance
protect them from item destruction, and damage negation from item
destruction if they were appropriately resistant.
Problem 3: a lot of callers didn't preserve the "amount of incoming
damage" that this refactor relies on. E.g. if the defender resisted
that element, the local dmg variable would be set to 0. So I had to
do some wrangling with callers to save that original damage
value. The rule of thumb is: all *incoming* damage counts. So that
includes the player's spellcasting bonus if applicable, but not
things like half damage, negation due to resistance, or extra damage
due to being vulnerable to cold/fire.
Then I figured, while I'm here let's get rid of all those silly cases
where destroy_items is called multiple times for various different
object classes, and cut the object class parameter out of it. This
has a few minor effects:
- Places where different object classes previously rolled
independently for destruction to happen at all now roll
once. (Which, by my calculation, generally means less incidences of
destruction - a fire attack now won't have three separate chances
to hit your scrolls, potions, and spellbooks. On the flip side, a
lucky roll will no longer save an entire object class in your
inventory.)
- Callers can no longer specify different probabilities for
destroying different object classes. The only place this was really
used was to call destroy_item with a slightly lower probability on
SPBOOK_CLASS. With the nerf in this commit, less of them ought to
be destroyed anyway.
- A very edge case of where explosion-vs-monster damage was totted up
differently for golems, which could result in differences of a hit
point here or there.
- All object classes being processed in one go means that less items
are destroyed than would be if they were still processed
independently. This is not really visible compared to the old
baseline of just destroying 33% of everything, but would be a
marked difference versus a copy of the game that still called
destroy_items separately for different object classes. To
compensate, I adjusted my planned damage-to-destruction-limit
scaling factor down from 8 to 5.
Not done: merging in ignite_items(), though that would probably be
really easy now.'
Notes from porting from xNetHack:
- It might be necessary to reexamine at all the conditional checks for
calling destroy_items. Because item destruction is much more
restrained and uses the actual damage from an effect, we might now
need to check 'if (!rn2(3))' and similar in all the places item
destruction occurs.
Simplify the valid-position highlighting via '$' by combining the
tmp_at(start) and tmp_at(populate) steps.
Add highlighting for a couple of targetting operations that would
give valid/invalid feedback via autodescribe but weren't displaying
highlight markers for $.
I made several jumping changes, most of them dealing with picking
hero's own spot.
Instead of just accepting an attribute, it's now possible to
use a color, or both color and attribute, for example:
OPTIONS=menu_headings:inverse
OPTIONS=menu_headings:red
OPTIONS=menu_headings:red&underline
Default is still just inverse.
This lets the player change the menu heading color without
needing to use menu colors for them.
Also makes it so the core uses NO_COLOR instead of 0, for all
the menu lines which don't have any prefedefined color.
Tested for tty, curses, x11, qt, and win32
cg.zeroobj was originally added (under its previous unprefixed name)
for providing a one-line way to zero out the fields of a struct obj.
struct obj tempobj;
tempobj = cg.zeroobj;
initfn(struct obj *otmp)
{
if (otmp)
*otmp = cg.zeroobj;
}
More recently, the address of cg.zeroobj began to be used as a return
flag to indicate some things, but the 'const struct obj zeroobj' wasn't
an ideal fit for the purpose and required a number of casts, including
casting away const.
Provide a better fitting variable (gi.invalid_obj) and eliminate a
number of casts.
Pull request from NullCGT: applying gold from inventory will flip
one coin and report "heads" or "tails". If impaired, that coin will
be dropped, otherwise it will remain part of the stack in inventory.
Can be done via 'a $' or with context-senstive item action via 'i $ a'.
I've used 'git merge --squash' and 'git commit -C <commit#>' to
flatten four commits into one and it seems to have accomplished what
I wanted, including retaining the log message from <commit#>.
I also changed
|You flip a gold piece. The gold pieces slips between your fingers.
to
|You flip a gold piece. It slips between your fingers.
when impaired and applying one from a stack.
Closes#1107
Applying one or more gold pieces now flips one of them, which will
cause it to come up heads or tails. This is NetHack, so there are
special cases for flipping a coin underwater or while fumbling or
greasy.
I've tried to future-proof this commit so that the code will not
need to be modified if other items are eventually added to the
coin class.
If memory serves, there was a patch for this on the bilious patch
database, but I was unable to locate it or who the original author
was. In any case, the code is entirely original.
Using apply to unlight a lit potion of oil makes it unlit, removes
it from inventory, and then re-adds it to try to force it to merge
with other potions of oil. If it was wielded and the other potions
were quivered, the game would panic. When merging, they get forced
into the weapon slot in preference to the quiver slot.
Unwearing it before freeinv+addinv would solve this but also leave
the hero with nothing wielded, even if it didn't merge with another
stack. Instead, don't try to merge if the potion being unlit happens
to be worn.
3.6.x was subject to this too and the fix is small+isolated but the
situation is so uncommon that I haven't bothered backporting it.
Applying a lump of royal jelly and then not picking anything to rub
it on had a similar problem. It also panicked if the applied lump
was wielded and other lumps were quivered. The fix is different
because the stack it gets split from during apply is known. This one
doesn't impact 3.6.x; applying jelly to eggs wasn't implemented yet.
When a hidden pet was relocated via magic whistle, a message was given
about "your <pet> appears" but it didn't show up on the map.
It was being marked as no longer undetected, but that was after rloc()
had moved it and redrawn its location. use_magic_whistle() needed to
redraw its location again after clearing monst->mundetected.
Treat a trap object that has become the focus of trap setting
occupation as if it had already been used up. (No discernable change
in behavior unless someone adds an artifact bear trap or landmine
that talks.)
Shouldn't 'trapinfo' be part of 'context' and be saved and restored?
(If so, it will need to include o_id and undergo pointer fixup during
restore.) When trap arming is in progress hero will be too busy for
player to issue S(ave) command but a hangup save could take place.
Found this running the fuzzer with address sanitizer. After applying an
object, the game checks whether it's a talking artifact so that it can
speak to you afterwards.
If, however, the object got destroyed in the process of applying it,
this will read and dereference obj after they have been freed. I found
this with a cream pie, which is always destroyed when someone applies
it.
To fix this, I tracked the obj pointer for what I think are the only two
cases in the big switch statement that don't track this - royal jelly
and cream pie. Royal jelly is only conditionally destroyed depending on
further input, so its function had to be refactored to take a struct
obj**, but cream pies are unconditionally destroyed so it can just be
set to null.
