Fixes#114
Report and contributed fix described lack of support for room type
"ant hole" in the code that loads special levels (and mentioned that
none of our des files attempted to use that room type so it isn't
noticeable in unmodified version of the game). The fix overlooked
a couple of other missing room types (leprechaun hall and cockatrice
nest) so I didn't use the pull-request's commit (so not sure what
github's automated updating will make of 'Fixes #114').
Fixes#111
Casting stone-to-flesh at a random statue animates it as a monster
(created via direct call to makemon()) at an adjacent or nearby spot
if there is already a monster at the statue's spot, but doing so on
a statue of a petrified monster (create attempt via montraits() which
called makemon() without the ADJACENTOK flag) turned it into a corpse
instead. Pass an extra argument to montraits() so that it behaves
the same normal statue animation for stone-to-flesh without changing
how it behaves when reviving corpses for undead-turning.
Handle suppression of 'score' from the 'O' menu for viewing and
setting hilite_status rules differently so that the count of the
number of rules will always match the number of rules shown by the
'a - View all hilites in config format' choice. If there are score
rules in the config file for a game running under !SCORE_ON_BOTL
configuration, list 'score' normally but if player picks that menu
entry, the followup menu will have the existing rule(s) and 'X -
remove selected hilites' but not 'Z - Add a new hilite'. If there
aren't any score rules--or there were some but they've now all been
'X'd--then 'score' won't be listed as one of the status fields.
When built without support for SCORE_ON_BOTL (the default), suppress
'score' field from 'O' menus for adding/removing hilite_status rules.
Also, add "encumbrance" as field name synonym for "carrying-capacity"
and "experience-points" for "experience". Relevant for rules set in
config file or via NETHACKOPTIONS.
The formatted value for attributes of condition highlights was
reporting 'normal' (aka no attributes) even when the highlight
rule specified some other value. It initialized a bitmap variable
to ATR_NONE, which is not 0, and then or'd other bits to that.
Then during formatting it checked whether the ATR_NONE bit was set
and returned 'normal' without examining the other bits. Actual
highlighting wasn't affected, just the strings in the rule set
shown by the 'O' command.
This is a separate issue from the earlier 'bonus fix' where the
attributes of previous condition rules got clobbered if a
hilite_status:condition/any-color&normal rule was added.
Take a step towards eliminating merging hilite_status rules during
highlighting by creating a single rule instead of multiple ones
when specifying multiple attributes for the same highlight via the
'O' command's menus.
Old:
(pick_one menu to pick a color) + (pick_one menu to pick an attribute)
| hilite_status:title/always/red&bold
(pick_one menu to pick a color) + (pick_one menu to pick an attribute)
| hilite_status:title/always/red&blink
New:
(pick_one menu to pick a color) + (pick_any menu to pick attributes)
| hilite_status:title/always/red&bold+blink
At present, rule selection during highlighting still merges multiple
applicable rules instead of finding the best one first, with the
problems that entails.
Bonus fix: a hilite_status rule for status conditions which specified
"no attributes" would clear attributes for all previous condition
rules rather than just the one(s) in that "no attributes" rule.
Fixes#110
NetHack dumped core while qsort was executing for sortloot. Fix a
logic error introduced by adding filtering capability to sortloot()
which could result in a sparsely populated array instead of having
the number of elements be less than the list size.
I don't know why this didn't show up sooner.
Fixes#109
Spells of healing and extra healing cast at monsters were handling
monster blindness differently from other forms of healing. (Potions
also work differently when drunk by monsters but I haven't changed
that since it seems to be intentional.)
Hero:
potion of healing cures blindness if blessed; spell of healing
cast at skilled or better now behaves likewise;
potion of extra healing cures blindness if not cursed; spell of
extra healing is inherently not cursed and already behaved
likewise;
potion of full healing always cures blindness even if cursed.
Monsters quaffing potions:
plain healing cures blindness if not cursed;
extra healing and full healing always cure blindess.
Hero casting healing spell at monster:
plain healing behaves like the hero plain healing case: cures
blindness as if blessed when cast at skilled or expert level;
this is a change in hehavior--it used to cure timed blindness
even if unskilled and not cure 'permanent' blindness at all;
extra healing cast by hero is inherently not cursed so always
cures blindness.
Wishing for "<size> glob of black pudding" worked, and wishing for
"black pudding glob" worked, but wishing for "<size> black pudding
glob" didn't work. Fix that.
Also, remove a bit of spaghetti introduced by the previous patch.
And once we know we're wishing for a glob, we can skip a big chunk
of wish parsing and special case handling.
Fixes#108
"small"/"medium"/"large" prefix was being stripped off during wish
parsing so that it could be used to control glob size. But those
are also prefixes for monster and/or object names. Wishing for
"small mimic corpse" or "large mimic" corpse failed with "nothing
matching that description exists" when it tried to satisfy "mimic
corpse". (Asking for "giant mimic corpse" worked as intended.)
Not mentioned in the report: wishing for "large {dog, cat, kobold}
corpse" produced the corpse of corresponding normal sized critter
instead of that of a large one.
Noticed while testing the fix: wishing for "glob" failed rather
than pick a random glob type. Wishing for "glob of grey ooze"
failed even though "grey ooze" is recognized as a variant spelling
for the gray ooze monster. Wishing for "<monster type> glob" also
failed even when the monster type was viable for globs. This fixes
all of those even though no one will ever notice....
Wishing for "small box" (and "medium box") no longer yields a large
box, it fails with "nothing matching..." instead. I was ambivalent
about the earlier change which had the unintended side-effect of
making them synonyms for "large box" so haven't tried to revive it.
The earlier change for 'fix github issue #106' could result in a
polymorphed weapon being worn in multiple weapon/alt-weapon/quiver
slots. Reorganize the relevant code more thoroughly this time.
Fixes#106
If dipping a worn amulet into a potion of polymorph turns it into an
amulet of change, the game panics while trying to use up that amulet
when the new one hasn't replaced the old one in inventory yet. Simply
reordering the relevant code isn't sufficient to fix things: once it
is in inventory and can be successfully used up, later code would end
up deferencing a stale pointer because it was unaware of the deletion.
Build fix, avoid use of 'class'
include\hack.h(199): error C2236: unexpected token 'class'. Did you forget a ';'?
include\hack.h(199): error C2332: 'class': missing tag name
include\hack.h(199): error C2027: use of undefined type 'sortloot_item::<unnamed-tag>'
Implement the suggestion that since teleporting away from the vault
while being confronted by the guard results in a shrill whistling
sound, the vault guard ought to have a tin whistle in his inventory.
I also added a check that he does have the whistle and to give an
alternate message if not, but after half a dozen tries to have a
squad of beefed up monkeys steal the whistle, they never accomplished
that. At least three times they took everything except the whistle
but I never succeeded in verifying the alternate message.
The code that formats an object for use in alphabetic comparisons
during sorting is forcing off wizard mode to avoid any alternate
formatting that might produce. Add a guarantee that doing this can't
be used as a backdoor to create a normal mode panic file if someone
figures out a way to make xname() panic.
When objects are in the same class, sortloot orders them by their
formatted name. It was reformatting each object every time it got
compared to another object. Change that to remember the formatted
name so that any given object is formatted at most once (during the
current sort; future sorts will need to format it again).
Armor and weapon classes are subdivided into smaller subclasses
and the formatting plus alpha compare is only done for items in
the same subclass, so helms come out before cloaks and don't get
their names compared, for instance. [That was from my 'revamp'
rather than the original implementation.] This adds a couple more
subclass sets: food (named fruit, 'other' food, tins, eggs, corpses,
globs) and tools (containers, pseudo-containers [bag of tricks and
horn of plenty once those have become discovered; prior to discovery,
bag of tricks is classified as a container and horn of plenty as an
instrument], instruments, 'other' tools).
The main difference, aside from the formatting efficiency improvement,
is to change the previous sort order
| pink potion
| potion of enlightenment
| purple-red potion
to be
| pink potion
| purple-red potion
| potion of enlightenment
by grouping undiscovered items before discovered items when class and
subclass match. So discovery state is essentially a sub-subclass and
formatting plus string comparison is only done for members of the
same sub-subclass. There are actually four state values: unseen
(which applies to particular objects rather than to their type),
unknown (not discovered and not named), named (not discovered but has
player-assigned type name), and discovered (either fully discovered
or considered not interesting to discover [no alternate description,
not nameable]).
My testing was primarily done with pickup ('m,' with menustyle:T)
and sortloot:Loot (the default) plus !sortpack (not the default and
not a setting I ordinarily use, but less verbose without the class
separators). It won't astonish me if oddities crop up with other
usage combinations.
Yesterday's sortloot() overhaul didn't include some cockatrice corpse
handling for pickup. If there's an object class filter in place and
pickup has been told to care about cockatrice corpses, have sortloot()
include them in the loot array even if food class isn't accepted by
the filter. In the pre-sortloot days, and in 3.6.[01] which didn't
attempt to deliver a filtered subset of loot, the check for such
corpses was done before pickup checks the filter. They need to be in
the loot array to retain the same behavior.
H7205 - full-pack identify might skip items if perm_invent is on
because updating the inventory window might reorder 'invent'
while the identify code is in the midst of traversing it;
H7120 - pickup that doesn't pick anything up can change the glyph
shown on the map because the pile might be reordered such
that a different item is on top;
H5216 - performing a sortloot operation on a pile and then switching
back to sortloot:none doesn't restore pile's original order.
The 'revamp' that changed the contributed sortloot feature to switch
to simpler usage (object list itself was sorted rather than having a
parallel array that needed to be constructed, sorted, traversed, and
discarded) turns out to have too many problems. This reverts to a
hybrid solution that constructs an array for traversal, leaving the
linked list in its original order, but hides most of the details of
that from sortloot() callers. The 'revamp' benefit of being able to
use normal list traversal is lost, as is the potential to skip
sorting when the list turns out to already be in the desired order.
This could stand to have a lot more testing than it's had so far.
Noticed while investigating the report about sortloot interacting
with persistent inventory window when identifying all of invent and
possibly skipping some items. [This doesn't fix that.]
End of game disclosure was using makeknown() on inventory. It is a
jacket around discover_object() which passes the flag to exercise
Wisdom. That's useless at end of game [now; conceivably wrong if
disclosure of characteristics exercise ever got added], so call
discover_object() directly to suppress exercise of Wisdom.
discover_object() was also calling update_inventory() for every item
being discovered. That's not useful when looping through inventory
at end of game.
While deciding which highlights to apply, give 'percentage' and/or
'absolute' rules that match precedence over 'always' rules regardless
of order within the config settings.
When using 'O' to add 'up/down/changed' rule, don't include 'down'
as a choice for field 'time'.
When using 'O' to add rules, don't squeeze out spaces if adding a
'textmatch' rule for title (to support "field worker", "high priest",
"student of stones", and so forth).
While deciding which highlights to apply, ignore double quotes when
testing whether a 'textmatch' rule matches the current text of a
field. This allows rules to specify string values as '"value"'
instead of just 'value'. It not does validate them to ensure quotes
are paired at beginning and end, it just ignores them. New rules
created via 'O' for rank title include them when displaying what the
new rule would look like as a config file option. Other text fields
haven't been changed to show quotes but ignoring such applies to all
'textmatch' comparisons.
Expand the menu for adding 'textmatch' rules for title. When a rank
has separate male and female titles, list three entries instead of
just one
"male rank"
"female rank"
"male rank" or "female rank"
(the order of the first two entries and of the two titles in the
third entry is reversed if the current character is female). If the
user picks the third entry, two rules are added instead of just one,
identical to each other except for the text to match.
Further expand that menu with
"none of the above (polymorphed)"
at the end. When deciding which highlights to apply, "none of the
above" and "(polymorphed)" and the full string are treated as
equivalent (with spaces, quotes, and parentheses ignored). Rather
than comparing anything against the title text, it matches if the
hero is polymorphed (where title will be "<hero> the <monster-type>"
instead of "<hero> the <rank>"). Note that the user can have config
file 'textmatch' rules for title to match specific "<monster-type>"
values but the 'O' menu doesn't offer any opportunity for that.
(I've just realized that rules for specific monster types should be
given precedence over "none of the above" but at present that isn't
done; the order of the rules will determine which wins out.)
Simplify the string comparison done when checking 'textmatch' rules
to decide whether to highlight something.
Fix the menu titles when setting up a textmatch via 'O': the title
for color referred to attribute and the one for attribute used the
default. The two tiles are set up in advance; the one for color was
set correctly but then the one for attribute was written into the
wrong buffer.
When using 'O' to manipulate hilite_status rules, if there are any
when you're done and the 'statushilites' option (iflags.hilite_delta)
is 0, give a message reminding that it needs to be non-zero for
highlighting to be activated.
The fact that the index to the array of hunger strings is an unsigned
field in 'struct you' is unimportant as far as its usage for status
highlighting. Since it is the only ANY_UINT field, change BL_HUNGER
to plain 'int' so that there'll be no need for ANY_UINT handling.
And some more validation when setting up highlight rules. For 'O',
in the menu to choose a relationship after supplying a number N,
don't include "less than N" and "N or less" for percentage or
absolute--other than AC--unless N is greater than 0, and don't
include "N or more" and "more than N" for percentage unless N < 100.
Also, when 'O' prompted for a number, if you entered <X or =X (for X
not a sequence of digits), it remembered the '<' or '=' (or '>=', &c)
when reprompting for a valid number. If the 'X' portion is invalid,
discard the relationship operator before asking for another number.
Add threshold relationships <= and >= so that the change to make <
and > perform their expected comparison can be resolved. "Point
release shouldn't force players to update their config files" does
not carry sufficient weight given that they already had to do that
to turn on status highlighting when going from 3.6.0 to 3.6.1. The
3.6.2 release notes can warn them about the need to update their
status highlight options if they're currently using '<' and/or '>'.
Entering new hilite rules via the 'O' command accepted '=' prefix
for numbers, but rules from config files did not. Now they do.
The '=' prefix is optional in both situations.
With 'O', percent rules and absolute rules had separate menu entries
so picking one was already choosing the rule type, but entering a
numeric value without percent sign (for percent) or with one (for
absolute) would change the type on the fly. If someone has already
picked percentage they shouldn't be required to append '%' to the
digits, so that is now optional. If explicitly included with the
number after having picked absolute, the value is rejected. It is
trivial to back up in those menus and choose the alternate type if
someone changes his/her mind part way through.
If a status field has both persistent (percent, absolute, always)
and temporary highlights (up, down, changed), give the temporary one
precedence when the value has changed. To do that with 3.6.1, the
rules for temporary had to follow the ones for persistent highlights
since whichever matched last was the one used. Now their order
relative to each other doesn't matter. If a value increases and
there is both an 'up' rule and a 'changed' rule, the more specific
'up' takes precedence, regardless of their relative order; likewise
for decreases and 'down' vs 'changed'.
There were a couple more tweaks needed to support negative values;
I overlooked the 'O' menu handling before. >-1% and <101% now work
for both the config file and interactive adding via 'O' methods of
defining highlight rules, although new >=0% and <=100% will be
clearer to anyone examining a rule set.
'enum relationship' was forcing LT_VALUE to be -1 but that fact was
never utilized anywhere, and the code was using magic number -2 to
mean "no relationship yet". This adds NO_LTEQGT to replace the
latter and gives it value -1. EQ_VALUE is still 0 so effectively
the default if a highlight hasn't been fully set up yet. LT_VALUE
is now just another positive value along with GT_VALUE, LE_VALUE, &c.
The Guidebook hasn't caught up with the code yet.
The rule choosing code used when deciding how to highlight something
only supports 'int' fields and relies on 'long' having the same bits.
It needs to be extended to support 'long' properly. Fixing should
be straightforward (except maybe for the initialization of min/max
best fit handling) but this doesn't address that. Also, data type
for encumbrance/carrying-capacity should be changed from unsigned to
plain int so that no extra handling for just one field will be needed.
Negative AC needed one extra change to support >-N since there was
a place in the code that assumed 0 was the lowest possible value.
(My earlier testing was with <-N which didn't have that issue.)
Make '/<N/' work as 'val < N' instead of 'val <= N', and />N/ work
as 'val > N' instead of >=. The <= and >= behavior might have been
intentional but the only support for that I could find was that
the 'O' menu used "N or less" for '<' and "N or more" for '>' when
setting up 'absolute' rules. If we actually want <= and >= (and we
probably do...), we should add them as more relationship operators
instead of misusing < and >.
Simplify the is_ltgt_percentnumber() case when parsing options
since input has been fully validated by the point that that test
passes. Among other things, /<-0/ and />-0' are now accepted (as
synonums for 0; -0 doesn't mean anything special) instead of being
silently rejected and then discarding the rest of the config file.
(That bad behavior is a separate issue not dealt with here.)
The code to choose a likely target when applying a polearm was
basing its decision on visible spots which contained monsters,
so could expose the location of a hidden monster if there was
only one such spot within polearm range. Not mentioned in the
report: it also wouldn't pick remembered, unseen monster unless
there was a monster still at that spot.
I've changed it to choose candidate location based on the glyphs
shown rather than on the presence of monsters.
Orc heroes get an extra food item ("to compensate for generally
inferior equipment") and it could randomly be lembas wafers (or
cram rations), and Ranger heroes always started with cram rations
even when they're orcs. Fixing the latter was simple, but the
normal race-based substitutions weren't applied to randomly
generated items, so the fix for the former required a bit of code
reorganization in ini_inv().
Elf heroes already get lembas instead of cram; do the reverse for
dwarves (although I don't think this case can happen--no role gets
lembas wafers and only orcs and always-human tourists get random
food); give orc heroes tripe instead of either lembas or cram.
> [1. perm_invent is kept in flags so persists across save/restore, but
> perm_invent capability can change if player restores with a different
> interface--or same one running on a different-sized display--so it
> ought to be in iflags instead.]
Not addressed here.
> 2. perm_invent window does not get updated when charging a wand (or
> other chargeable item presumably), with a scroll of charging.
Most scrolls rely on useup() -> update_inventory(), but charging uses up
the scroll early so that it will be gone from inventory when choosing an
item to charge. It needed an explicit update_inventory() after charging.
> 3. update_inventory(), is called from setworn(), which is called from
> dorestore(), when loading a save. Segfaults have been observed in
> variants based on this code (though not yet in vanilla 3.6.1), so it's
> possible this may be unsafe. The update_inventory() call in setworn()
> could be protected with "if (!restoring) ..."
tty doesn't support perm_invent, so this might be a win32 issue.
I've made the suggested change, but a better fix would be to turn off
perm_invent as soon as options processing (new game) or options restore
(old game unless/until #1 gets changed) has finished setting things up,
then turn it back on at the end of moveloop()'s prolog when play is
about to start.
= =
Most of the read.c change is reordering prototypes to match the order
of the corresponding functions. I did this when adding a new static
routine, then ended up discarding that routine.
