The new livelogging of wish results caused a segfault when attempting to
handle the results of a wizard mode terrain wish, since a successful
terrain wish returns a nonzero obj which nonetheless is just a dummy
object. Move the existing check for that further up to skip all the
livelogging stuff entirely, since such wishes will never happen in a
real game and exist purely for debugging purposes.
K3610 reported to devteam:
When you see a monster wield a cursed two-handed weapon,
the weapon "welds itself to the foo's hand" instead of its "hands."
Observed on hill orcs wielding a cursed two-handed sword.
When taking stuff out of a container, specifying a subset count for
an item and getting the pickup_burden prompt, answering 'q' undid the
subset split but answering 'n' did not. If the item in question was
a stack of gold, the container would end up with two stacks. That
action could be repeated as long as any of the stacks was big enough
to trigger pickup_burden confirmation so an arbitrary number of gold
stacks could be produced. (Eventually they would be too small for a
subset to cause an increase in encumbrance, or possibly all reduced
to just one gold piece, then no more stacks could be created.)
Situation occurred for all menustyles; traditional and via-menu needed
separate fixes. It didn't occur for pickup off the floor.
Report was for 3.6.6 but the bug was still present in dev version.
Wishing for "N <size> globs [of pudding type]" produces 1 glob
starting at <size> and then multiples weight by N (so possibly
increasing <size>). When not it wizard mode, N can be replaced by
a random amount to prevent the total weight from being huge. When
N was less than 6, it was possible for that random amount to be
larger than what the player asked for.
Change the way the random amount is calculated so that it won't
ever be larger than what player specified. Also for wizard mode
prompt whether to make the substitution so that the player can
choose to abide by the limit or to obtain a huge glob for whatever
testing is being conducted.
Extend the log event for a wish to include what was produced. It
would be better to show the item as fully ID'd but then #chronicle
gives away information.
The backslash+newline pairs were inserted for this log message. In
the game and in dumplog those two lines are each one wide line. The
turn numbers shown are actually arbitrary since ^W takes no time.
|Logged events:
| Turn
| 1: wizard the chaotic male orcish Wizard entered the dungeon
| 2: made his first wish - "protection", got "a tattered cape"
| 3: made his first artifact wish - "blessed +2 rustproof magicbane",\
got "an athame named Magicbane"
| 4: wished for "master key of thievery", got "a key named The Master\
Key of Thievery"
Pull request from vultur-cadens: if a wish is declined by asking
for "nothing" or "none" or "nil", log that for gamelog/livelog as a
declined wish. Likewise when specifying monster species or class to
genocide as "none" or "nothing" or ESC, log it as declined genocide.
Closes#768
I forgot to do this with yesterday's post garbage collection fix
update. Record lua warnings in paniclog during normal play too, not
just when in wizard mode.
selection_new() returns an address of malloc()'ed buffer.
If ov is null, this value is discarded without freeing the buffer.
To avoid this, move null-checks before calling selection_new().
Also, remove null-check of the return value of selection_new()
because it always returns non-null.
Now that the garbage collection problem has been fixed, record lua
warnings in the paniclog file rather than showing them on the screen.
Move nhl_warn()'s warnbuf[] to struct g in case restart ever gets
implemented so that it can be cleared if the restart occurred while
a warning message was under construction.
Writing lua warnings to paniclog (coming soon; tested without the
garbage collection fix in order to have test data) could crash on
the last pair. Those are written after the 'nomakedefs' structure
had been freed so version_string was Null.
The NAO PANICLOG_FMT2 code triggered a warning about the test for
g.plname; it is array so will never be Null.
realloc(NULL, size) is legitimate usage and nhrealloc() shouldn't
log a "< 0x00000000 __FILE__ __LINE__" entry for it. heaputil
would complain about freeing Null.
If homebrew is detected, and if the homebrew ncurses package
is installed, the set things up to use it.
Hopefully, this works okay for everyone right away, but it may need
some tweaking.
Add new routine 're_alloc()' that functions as MONITOR_HEAP-aware
libc realloc(). 'nhrealloc()' is the version that passes source
file and line info if built with MONITOR_HEAP enabled. The heaplog
data might now contain '<' (freed by realloc), '>' (replacement
allocation by realloc), and '*' (resized by realloc) entries in
addition to the previous '+' (allocated) and '-' (freed) entries.
heaputil has already been updated in the NHinternal repository.
Move FITSint_() and FITSuint_() from hacklib.c to alloc.c so that
they can be accessed by miscellaneous utility programs.
Remove three or four copies of FITSint_() that were duplicated in
utility programs like dlb and tile2bmp due to those not having
access to src/hacklib.o. They do have access to src/alloc.o (and
util/panic.o).
Pull request from entrez: fixes the combination of A and P for
menustyle:Full. For menustyle:Traditional, it fixes selecting P for
putting stuff into a container. Using P for multi-drop (D) already
worked and I haven't tried to figure out why the two commands behave
differently with just-picked.
Closes#774
When using 'A'/autopick with the 'items you just picked up' category,
instead of autoselecting all items within that category, it selected
every item in your inventory (like it used to work before 3.7). Just
blew up a bag of holding because of this.
While testing the fix for that, I noticed 'P' wasn't working at all
with menustyle:traditional -- you could select it as a filter, but it
didn't actually get applied to anything, so it would end up prompting
you for every item in inventory. Fix both those things.
Add the patch from entrez to describe the tower of flame effect from
a scroll of fire as "the blast" rather than "your spell" if it reveals
a secret door.
This will be an annoyance for wizard mode until someone actually
figures out and fixes the problem. The complaints from lua during
garbage collection aren't new, they were just being ignored before.
Remove a ton of tabs in nhlua.c and add missing whitespace to a bunch
of 'if(test){' lines and to a few casts.
Also simplify? obj handling during garbage collection (does not fix
the current gc problem) in nhlobj.c.
This parameter appears to have been in the code for a very long time,
but never used, since no version of NetHack I can find had mazes with
randomly placed fountains in them. Certainly isn't used now, so this can
be reduced to the same call to find_okay_roompos used by similar
functions such as mksink.
Andrio pointed out at some point that the "below 25% HP: pet will not
attack at all" mentioned in this comment was wrong. It will not attack
*peaceful monsters* at all, but will still attack hostile monsters.
Also, the math behind the balk variable has confused several people,
thinking it's off by one and allowing the pet to attack one level higher
than stated. This is not the case, since it's the lowest level they
*won't* attack. Clarify that.
