The #therecmdmenu command calls getdir() which issues an "in what
direction?" prompt. This allows you to answer with "_" instead of a
regular direction, then it will call getpos() to allow you to move
the cursor and type "," (or ";") to behave as if a left-click had
been done or type "." (or ":") to behave as right-click.
Ordinarily I would think of the 'normal' getpos() response of "."
as suitable for left-click, then one of the other getpos finishers
for right-click, but comma is left of period on a standard keyboard
and that seems useful for remembering which is used for which click.
Left clicking on a spot farther than one step away offers travel,
throw iff lined up, and also click-look as choices. If you right
click farther than one step away, it will only offer click-look.
The look choice for either left or right click isn't inhibited by
having the clicklook option set to False. After all, player is
explicitly choosing the menu entry to look at something.
New getdir.mouse can be bound to some other key than "_" and the
getpos.pick* responses could already be re-bound, but there's no
separate getdir.left/right that could be used to bind different keys
from those used for the four getpos responses.
Still more PR #777. Commit c4c6c3d73a broke #therecmdmenu travel,
throw, and far-look. It was restricting dx and dy unnecessarily
and that resulted in not specifying the correct location when the
destination was farther than one step away.
Testing those properly requires a mouse. I've implemented a way
to simulate a left or right click at getdir()'s prompt (only useful
for #therecmdmenu). That will be committed separately.
More PR #777: there's no need for there_cmd_menu() to pass absolute
<x,y> instead of <dx,dy> for a couple of actions. Those actions can
reconstruct <x,y> by adding <dx,dy> to <u.ux,u.uy>.
For the description of what a keystroke does, augment 'm' (or whatever
key has been bound to #reqmenu) to replace the default description
|m prefix: request menu or modify command (#reqmenu).
with
|m movement prefix: move without autopickup and without attacking
|m non-movement prefix: request menu or modify command (#reqmenu).
The text is delivered by pline so tty will issue --More-- between the
two lines.
Refine the code from pull request #777 by changing act_on_act() to
take 3 arguments instead of 6. x,y and dx,dy are mutually exclusive
so it doesn't need both pairs provided that the caller is adjusted
to pass the ones appropriate for the action, and dir is easily
derived from dx,dy for the couple of cases that use it.
Targetting any of the eight surrounding spots while swallowed will
kill the engulfer. Picking a spot farther away reports "no monster
there" even if there does happen to be one at the chosen spot.
When using 'm #wizkill' to kill monster(s) off without giving the
hero any credit or blame, temporarily force context.mon_moving On
so that collateral damage (other monsters killed by targetted gas
spore's explosion) also don't give the hero any credit or blame.
Make sure that some message identifying the monster is given when
targetted monster is killed even if hero can't see or sense it.
Add a way to get rid of specific monsters in wizard mode without
fighting, zapping, &c. #wizkill command lets you kill creatures by
picking them with getpos().
You can pick multiple monsters by targetting them one after another.
You don't have to be able to see or sense them but if you target a
spot that has no monster, the command ends.
By default, the hero gets credit or blame as if having killed the
targets but #wizkill can be preceded by 'm' prefix to treat their
deaths as if they had been caused by a monster.
When picking an item from inventory and then picking 'I - adjust
inventory by splitting this stack' in the item-action menu,
yn_function("Split off how many?") is used to start getting the
count without needing to wait for <return>. It includes the response
in message history (so review of history will see that first digit).
The code then uses get_count() to obtain any additional digits. Tell
the latter to store "Count: N" in message history if N is different
from the first digit.
That's not as good as updating message history to replace the entry
showing the prompt with the first digit with one that shows the full
count but at least it's accurate when the count is 10 or more.
The change to add a menu choice for naming an adjacent monster via
\#therecmdmenu was unintentionally requiring that the monster have
monst->mextra. So it worked on pets (regardless of whether they
were already named) because they have mextra for 'edog' extension,
but not on the majority of monsters. And when it failed the program
would crash with a 'segmentation fault' error.
Fix the check for whether a target monster already had a name when
deciding to use "name <mon>" or "rename <mon>" in the menu entry.
It's possible to get a strange command result of ECMD_CANCEL+ECMD_TIME.
If/when that happens don't just treat it as cancel, make sure that time
elapses.
If an invalid command is provided, clear the do-again queue so that ^A
won't attempt to try the same invalid command again.
This is enough to prevent abuse by denying access to functions and
denial of service (RAM and instruction step limits), but not enough
to allow restricted use of things that require finer control (e.g.
filesystem access).
If something goes wrong, the whole thing can be turned off, for
now, in config.h (see NHL_SANDBOX).
None of the current functionality requires changes to build systems;
some of the possible future functionality may require some #defines
- TBD. There is lots of dead code (#ifdef notyet) for bits of that
additional functionality; we can rip it out if we don't want those
additions or we can complete (parts of) it depending on our needs.
All current uses of Lua are connected to sandboxes and guarded with
nhl_pcall (sandbox and lua_pcall wrapper); options and limits can
be set at the callsites in the passed nhl_sandbox_info. Some of
the error handling may be wrong - panic() vs. impossible() vs
silence.
Memory and instruction step limits should be tuned prior to release;
there's no point tuning them now.
There's no 'w-' or 'Q-' for alternate weapon, but context-sensitive
inventory is starting from the object rather than the command so can
finesse that. 'A' does allow alternate weapon to be directly unset
(aka reset to bare/gloved hands) but is not friendly to being passed
queued input.
This adds an extra internal command which only handles unset uswapwep,
even though that is something which is awfully specialized to get it's
own command. Users don't see this command so that shouldn't matter.
More context-sensitive inventory support. While examining inventory,
if you pick an item other than gold and it has a quantity of more
than 1, "I - Adjust inventory by splitting this stack" will be one
of the menu choices.
Breaking doorganize() into two parts was much easier than expected,
but the new internal command added to be an alternate for the first
part had more niggling details than anticipated.
Message history only shows the first digit with "Split off how many?"
if the player enters more than that.
Mostly attempting to clean up potential error handling but I don't
have any error cases to test with. Doesn't seem to break anything
when there aren't any errors....
This attempts to make item-actions, #herecmd, and #therecmd be more
robust. When rhack() or yn_function() take queued input off cmdq
and get something unexpected, discard the rest of the queue.
It also fixes the two crash cases that entrez reported. There are
bound to be others though.
I think a lot of actions that can be executed by queued input are
going to need nomul(0) calls to handle repeat counts that should be
ended early if something unexpected happens or something expected
fails to happen. But that clears cmdq so may be tricky to decide
where to use.
Using '&#' or '?f#' showed "# perform an extended command (##)".
The "(##)" part looks rather silly and is not helpful. Expand the
text a little and omit command name for that particular command:
"# enter and perform an extended command".
The comment preceding new 'savech_extcmd()' said that the core didn't
care whether it was given the full command name or just enough leading
substring to be unambiguous. Then it tested the string against
"repeat" which contradicts that comment. Didn't seem to be an actual
problem because "#repeat" is not flagged for auto-completion, but fix
the code to match the intent of the comment and reword the comment to
match the code.
Fix '#repeat' for tty; both it and ^A can repeat an extended command.
Fix both for curses; they can repeat an extended command instead of
just repeating the initial '#' to start getting an extended command.
X11 (tested), Qt (tested), and probably Windows GUI (not tested)
behave the same as before: ^A (or #repeat) after an extended command
just repeats the # to run the dialog to get an extended command.
I hope this introduces fewer bugs than it fixes but I don't think I'd
bet on that....
If you used ^A to repeat a command which had taken no time, the
repeat execution would take time. This fixes that. Also, give some
feedback when trying to repeat an invalid command.
Internals bit: don't use 'X == cmd_from_func(do_repeat)' to decide
whether key X is the key for #repeat. Both X and Y might be bound to
that action and cmd_from_func() could return Y rather than X.
There is another ^A bug that I haven't figured out how to fix:
t ESC start to throw but don't finish
^A nothing seems to happen
^A "You don't have that object."
The first ^A repeats 't', doesn't display a prompt for what to throw,
but does request input for it. The second ^A fulfills that input and
doesn't match any inventory item. Either 't' shouldn't have been put
into the do-again buffer or do-again handling should have ceased when
it was taken out and there was no further remembered input, so that
normal prompting would resume. My tentative attempts for both those
approaches didn't work.
Issue #734 reported as "parse function" by Meklon2007: the change
yesterday intended to make ^A work for commands that were preceded
by a prefix was triggering a crash if used after a keystroke that's
not assigned to any command.
'M^A' or '~^A' would segfault by derefencing a null pointer when
checking whether 'M' or '~' was a prefix. This prevents the check
attempt from doing that, but a better fix would be to not put the
invalid command keystroke into the do-again buffer in the first
place.
Fixes#734
Reported by luxidream via the web contact form and also as github
issue #732: using the repeat command after F+direction would take a
step in direction if there was no target to fight.
The direction was being repeated without the F prefix. It wasn't
specific to F; m+dir misbehaved too. This fix seems to work but it
should be replaced with something more robust.
Fixes#732
Allow the player to precede q/#quaff or M-d/#dip with the 'm' prefix
to skip asking about fountains, sinks, or pools if one of those
happens to be present, similar to how using it for e/#eat skips food
on the floor and goes straight to inventory.
If you use it and don't have any potions, you'll get "you don't have
anything to drink" or "you don't have anything to dip into", same as
when there is no suitable dungeon feature present combined with no
potions. However, if an applicable dungeon feature is present and
you don't use the prefix but answer 'no' to drink from fountain,&c
and you don't have any potions, "else" will be inserted into the
message: "you don't have anything else to drink".
A big part of the diff is just a change in indentation level for
code that is now inside 'if (!iflags.menu_requested) {' ... '}'.
rhack() normally calls parse(), parse() sets context.move to True
assuming that the player's next action will take game time, then
when it returns, rhack() sets context.move back to False if the
assumption turned out to be incorrect. But when performing actions
after picking something in inventory, rhack() doesn't call parse()
so context.move is left at False.
This was hidden by making the inventory command take game time if
the player picked an item and set up an action to be done with it
even though the action hadn't taken place yet. So time was being
accounted for but if the hero didn't get consecutive moves then
monsters got their turn between the shouldn't-take-time inventory
command and the ought-to-behave-like-normal-command queued action.
My initial attempt to fix this (before figuring out how context.move
works) by stopping inventory from taking time didn't work because
queued item-actions stopped taking time too, or rather the fact that
they took no time became exposed. This second attempt doesn't have
that problem and I think it is correct.
Normally dipping gets the thing to dip first and what to dip it
into second and the item-action handling knows that. I'm not sure
why that wasn't working as intended and I couldn't figure out how
to make it do that, so went another way: this adds an internal
extended command that executes an alternate dip routine which gets
the potion to dip into first and the thing to dip into it second.
The #dip command should allow an 'm' prefix to skip fountains and
pools, similar to how eating accepts it to skip food on the floor.
But this doesn't implement that.
Picking a corpse while looking at inventory issued a menu that had
entry for eating that and if on an altar another one for offering
that. Picking the eat or offer choice worked as long as there
weren't any other corpses on the ground or altar. If there were
others, they'd be skipped but you'd get prompted for which item in
inventory to eat or offer instead of operating on the one that was
used to initiate the action.
Allow selecting an item from inventory and show a menu of actions
applicable for that particular item. Some of the entries might
be slightly spoilerish (eg. it'll reveal that you can read T-shirts),
but the improved usability for new players is more than worth it.
Generally known as "item actions", this was first implemented
in AceHack by Alex Smith.
Remove redundant 'if (K)' from there_cmd_menu(). The !K case doesn't
get there and that test's presence fools the compiler (oldish clang)
into warning that 'npick' might be used uninitialized.
My fixes to the travel stuck oscillation did not fix all of them,
and I've even seen a 3-step loop - which my fixes cannot detect.
I guess there could be arbitrary-sized loops too.
To definitely fix this, keep track of all the map locations travel
has moved the hero through, and if it tries to go on a location already
used, stop travel and give the unsure -message.
Offer the chance to explicitly hide via #monster when poly'd into a
hides-under creature. hides_under() doesn't pass the is_hider() test
so wasn't being allowed before.
If poly'd hero's monster form is both a webmaker and can hide-under,
have #monster prompt the player for which is intended. When poly'd
hero successfully spins a web, say so.
If poly'd hero deliberately tries to hide under a cockatrice corpse,
turn to stone.
Add a type to force g.{command_count,last_command_count,multi} to have the
same type (because cmd.c: g.multi = g.command_count;) and some resulting
cleanup.
