There is a lot of code affected by this, and Pat Rankin correctly
observes that it would be better to store roguelike as a level flag
rather than just using Is_rogue_level. A note for the future.
This change has two parts.
First, the Book of the Dead now has a special message in lava. This was
originally added by Steve Melenchuk (aka Grunt) as a precaution against the Book
burning up. While I'm pretty sure that the Book can't burn up in the current
codepath, the message is a nice touch and it doesn't help to add a safety check.
Second, always print the inventory destruction messages. The reason for this is
fairly subtle, but relates to the timing. By not printing the messages when
dying, the visible effect to the player is that if they have lifesaving, the
items burn up before they die, but if they do not, they burn up after they
die---the fact that they do burn up is visible in the dump or bonesfile.
This avoids a player being left standing on lava if their boots are burned up and
they survive the damage.
This change is originally due to Steve Melenchuk, fixed up by Derrick Sund.
Simplify retouch_equipment (post-3.4.3.code, called when hero
changes alignment or turns into a silver-hating critter; inventory
traversal must cope with the possibility of losing object->nobj when
processing object) by switching to recently added object->bypass code.
From a bug report, dropping a lit
(burning) potion of oil while levitating can produce an explosion which can
destroy inventory. If in the process of dropping multiple items, the ones
after the oil might be gone, resulting in use of stale pointers and possibly
triggering an "extract_nobj: object lost" panic or even a crash. While
testing my fix, I discovered that being killed by an exploding potion of oil
could produce an "object_is_local" panic if bones are saved (and reproduced
with unmodified 3.4.3).
From a bug report, being teleported onto
a sink while busy putting on levitation boots triggered a crash when
Boots_on() was called (as '(*aftermv)()' on the next turn) because
'uarmf' would be null by then. Not mentioned, but the same problem was
encountered by Boots_off() if the teleport happened while you were busy
taking boots off. It could be fixed by having having dosinkfall() call
cancel_don() if donning(uarmf) yields true, but this patch does a little
more than that: cancel donning/doffing of any multi-turn armor if you
fall onto a sink. It also prevents you from falling if you end up
flying (which will have been blocked while levitating).
The situation when putting on levitation boots has a sequencing
issue: setworn() causes you to be flagged as levitating immediately,
but the float_up() feedback doesn't occur until Boots_on() gets called
a turn later. Teleporting to the sink will tell you that you crash
onto the sink and and that you stop putting on boots, without having
been told that you've floated up into the air. It's suboptimal but it
doesn't seem to actually be incorrect.
After newsgroup discussion of C343-162, I noticed that it had fixed
the verb usage in the message for confused and/or stunned, but not in the
one for resisting. (If a monster name ends in 's', or it has been named
"you", 3.4.3 used the wrong verb because vtense() was fooled about plural
or second person.) No new fixes entry needed....
When hiding as a monster, say so during the #monster command and
also list being hidden in the status section of enlightenment/^X.
Also, prevent hiding on the floor or ceiling on the planes of air
and water. (Didn't apply to monsters, who only hide on ROOM spots.)
Half the change to dohide() is just revised indentation.
In the newsgroup recently, someone mentioned that when polymorphed into
a flying monster, he couldn't retrieve items out of pits, even with the '>'
command. I thought that we had fixed that, but the fix apparently only
covered walk-through-wall creatures (xorns and earth elementals), not flyers
and ceiling clingers. Now those can also deliberately enter pits via '>'.
The fixes entry is in the new features section in order to be next to
the xorn one. '>' at pit locations is new, but it handles something which
was missing so feels more like a bug fix than a new feature to me....
[Keni, the bug page entry C343-12 about '>' (for xorns?) is general enough
to cover this fix, so we don't need to add a new one for flyers.]
I noticed an unrelated fixes35.0 entry mentioning '>' which was duplicated;
this removes one of the copies.
From a bug report, the message displayed to the screen by
impossible() might be different from the one written into paniclog, if it
had argument subsitution/formatting. I couldn't reproduce that myself,
but stdarg.h/varargs.h is tricky stuff and I think that passing the va_list
to a routine which steps through it requires that va_start be called again
if you're going to use the va_list a second time. This changes impossible()
to handle its arguments only once, like panic().
Message tidbit when engraving with a weapon which becomes too dull to
finish the whole text: add missing final period to
You are only able to write "<partial text>".
I think that proper usage puts the period in front of the quote rather than
after, but that could make it look as if the period ended up inside the
partial engraving.
From a bug report, you get the message "this shop
appears to be deserted" when entering a shop whose shopkeeper is not there,
even if you can't see at the time. Change it to be
'this shop is deserted'
when you can see that the shop is empty,
'this shop is untended'
when there's no shopkeeper but you can see some other monster(s) inside,
'this shop seems to be deserted'
when you can see but there is a monster that you can't see (probably a
hidden mimic, but possibly an invisible monster or one standing behind
a boulder that's ended up in the shop), or
'this shop seems to be untended'
when you can't see (and don't have active telepathy or persistent
monster detection).
From the newsgroup: inventory dropped by a monster killed by system
shock when zapped by a wand of polymorph is protected against being poly'd
by that same zap, but worn/wielded equipment that's forced to be dropped
due to being unusable in a transformed monster's new form was not. Not
mentioned in the newsgroup, but also fixed: boulders dropped when a giant
turns into a non-giant were also subject to the zap instead of protected.
Introduce some variation in monster movement by adding and substracting
small random amounts to the fixed increment from the main monster table. The
amount added is potentially bigger (average is +2) than the amount substracted
(average is +1), so monsters will tend to be slightly faster. Probably not
noticeable except for super-slow monsters not being so glacially slow, since
net +1 is a bigger relative increase for such critters' low movement rates.
In practice, the noticeable effect will be that ordinary speed monsters will
occasionally get an extra move even if/when player keeps the hero unburdened.
Possible extension (although I'm not planning to do it): sort the
monster list by pending movement points, so that faster monsters move before
slower ones. The random variation would become noticeable because groups of
same-speed monsters would alter their movement order depending upon who got
a bigger increment or smaller decrement on that turn.
From a bug report, when the hero has
precisely 24 movement points, fast monsters will move twice and then the
hero will move twice, instead of being interleaved. The hero will get 24
additonal movement points approximately one turn out of three when wearing
speed boots (or when using spell of haste self or potion of speed) and can
also get 18 additional points twice in succession, with a net of 24 after
12 of them are used on the first turn. But hero might not be starting at
zero at the time, so I'm not sure how often this is actually noticeable.
He apparently got a report about this for <some variant> and had some
user-contributed instrumentation in place that made it easier to replicate.
This fix is the one he suggested.
From the newsgroup, while rehashing our bug page entries someone
mentioned that buying a minor consultation from the oracle exercizes
wisdom twice. First when getrumor() chooses one, then again when
outrumor(BY_ORACLE) delivers it.
This also simplifies some #if GOLDOBJ conditional code.
