]> granicus.if.org Git - nethack/log
nethack
5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Mon, 11 Feb 2019 17:03:49 +0000 (12:03 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agofix #H8167 - shopkeeper message after ending game
PatR [Mon, 11 Feb 2019 01:45:26 +0000 (17:45 -0800)]
fix #H8167 - shopkeeper message after ending game

Closing nethack's window sets 'program_state.stopprint' to inhibit
disclosure interaction, but shopkeeper claiming hero's stuff or vault
guard claiming hero's gold didn't honor that and just issued normal
pline messages.  For win32, they got delivered in a popup even though
nethack's window had gone away.

Make those two end-of-game situations honor 'program_state.stopprint'.
[Fix not tested on win32...]

5 years agoGuidebook date update
nhmall [Mon, 11 Feb 2019 00:02:06 +0000 (19:02 -0500)]
Guidebook date update

5 years agoGuidebook consistency improvement
nhmall [Sun, 10 Feb 2019 23:48:50 +0000 (18:48 -0500)]
Guidebook consistency improvement

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Sun, 10 Feb 2019 23:21:43 +0000 (18:21 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoisaac64 on vms
PatR [Sun, 10 Feb 2019 21:54:16 +0000 (13:54 -0800)]
isaac64 on vms

Not tested yet...

5 years agocompiling isaac64.c
PatR [Sun, 10 Feb 2019 21:41:45 +0000 (13:41 -0800)]
compiling isaac64.c

Revise isaac64.c so that it can be compiled unconditionally and
yield nothing if USE_ISAAC64 isn't defined.  Necessary for building
via vmsbuild.com.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Sun, 10 Feb 2019 05:05:44 +0000 (00:05 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agocomment typo
PatR [Sun, 10 Feb 2019 00:27:51 +0000 (16:27 -0800)]
comment typo

5 years agofix #H8124 - interrupted donning gives player info
PatR [Sun, 10 Feb 2019 00:07:18 +0000 (16:07 -0800)]
fix #H8124 - interrupted donning gives player info

about the armor.  Wearing armor sets obj->known, making its enchantment
be shown when it gets formatted, because the AC value on the status line
lets the player deduce what that is.  It was being set at the beginning
of the wear operation.  If the armor got stolen before it became fully
worn, the enchantment was still shown.  Defer that until the end of the
operation.  An attentive player can still deduce the enchantment if the
item is stolen (because its protection starts immediately) but the hero
won't learn that enchantment unless the donning completes.

This might be suboptimal but it isn't qualitatively different from
watching a pet walk/not-walk over items whose bless/curse state isn't
known or dropping unidentified items in a shop to check their price.
The player can deduce something that the hero doesn't know yet.

5 years agominor nitpicks
PatR [Sat, 9 Feb 2019 23:33:16 +0000 (15:33 -0800)]
minor nitpicks

A couple of trivial things I noticed while looking at status cleanup.

5 years agocurses status cleanup
PatR [Sat, 9 Feb 2019 23:30:53 +0000 (15:30 -0800)]
curses status cleanup

Stop the last of the memory leaks occuring during basic usage.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Sat, 9 Feb 2019 00:58:36 +0000 (19:58 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agomore curses memory
PatR [Sat, 9 Feb 2019 00:51:33 +0000 (16:51 -0800)]
more curses memory

Message history now cleaned up at game end.  Status window cleanup is
not taking place because the core is suppressing it #if STATUS_HILITES.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Sat, 9 Feb 2019 00:03:20 +0000 (19:03 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Fri, 8 Feb 2019 23:52:22 +0000 (18:52 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agocurses: plug most memory leaks
PatR [Fri, 8 Feb 2019 23:50:59 +0000 (15:50 -0800)]
curses: plug most memory leaks

This takes care of a lot of the leaked memory in the curses interface.
It still needs to free memory allocated for status fields when the
status window is destroyed at game end; likewise for message history
when the message window is destroyed.

5 years agocurses: getline vs DEL, ESC
PatR [Fri, 8 Feb 2019 22:54:40 +0000 (14:54 -0800)]
curses: getline vs DEL, ESC

Support <delete> (aka <rubout>) during getline().  It doesn't actually
honor the current erase_char value set up for the terminal, just
treats DEL the same as ^H.  (The previous lack of support had nothing
to do with terminfo specifying ^H; the handling is hard-coded.)

tty treats escape while there's already some input as kill_char (erase
the input but get more from scratch) and returns ESC if there isn't.
curses was doing the first half but not the second, so not providing
any way to communicate "cancel" back to the core.  Fix is simple.

Other getline() bug fixes:
1] there was a wprintw("%*something") which was passing the value from
strlen (type 'size_t') to the "%*" argument (type 'int').  That's
always wrong (size_t is guaranteed to be unsigned) and could be severe
(if size_t is different width than int--as on current OSX systems--
depending upon the internals of argument passing).
2] strncpy() only supplies a terminating '\0' if the input is shorter
than the number of characters specified.

A lot of reformatting is warranted but I only did the getline routine
(manually, so might have missed stuff).

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Fri, 8 Feb 2019 19:52:57 +0000 (14:52 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agocurses memory: 'anything identifier'
PatR [Fri, 8 Feb 2019 01:10:55 +0000 (17:10 -0800)]
curses memory: 'anything identifier'

Three or four instances of one simple memory leak.  Allocating a union
'anything' to pass to add_menu(), then not doing anything with it.  The
value gets copied so there's no reason for the original to stick around.

[There are still lots of other memory leaks.]

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Fri, 8 Feb 2019 01:05:02 +0000 (20:05 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agocurses: change from malloc() to nethack's alloc()
PatR [Fri, 8 Feb 2019 00:48:37 +0000 (16:48 -0800)]
curses: change from malloc() to nethack's alloc()

There was no provision for malloc() potentially returning Null and it
wasn't integrated with nethack's MONITOR_HEAP.  'heaputil' shows that
the curses interface is leaking like a sieve.  If some things are
actually being allocated separately and then freed from within curses,
those need to be thoroughly documented and maybe switched back to
malloc().

5 years agocustompline(SUPPRESS_HISTORY,...) for curses
PatR [Fri, 8 Feb 2019 00:04:24 +0000 (16:04 -0800)]
custompline(SUPPRESS_HISTORY,...) for curses

The curses interface already has a hack to keep 'Count: 12', 'Count:
123' intermediate multi-digit counts out of its message recall history
for ^P, but it was flushing real messages when getpos()'s 'autodescribe'
reported what the cursor moved over.  Overload the count hack to support
 putstr(WIN_MESSAGE, ATR_NOHISTORY, text)
(which is what custompline(SUPPRESS_HISTORY, ...) eventually calls).

The conditional logic for when to create the 'count_window' was pretty
convoluted.  This simplification has the same semantics but I don't
have PDCURSES to actually verify that.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Wed, 6 Feb 2019 02:53:19 +0000 (21:53 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoarmor comments
PatR [Tue, 5 Feb 2019 22:49:43 +0000 (14:49 -0800)]
armor comments

I don't see the #H8124 bug getting fixed any time soon, so add a
comment about it since the relevant place is unambiguous.

No change in code except removing a couple instances of 'register'.

5 years agomore message history
PatR [Tue, 5 Feb 2019 02:40:55 +0000 (18:40 -0800)]
more message history

Noticed while investigating the issue with DECgraphics characters in
msg_window:full/combination/reverse output for tty which got fixed
by the previous commit.  There was a discrepancy in DUMPLOG because
the pager code bypasses pline() in order to use putmixed().  tty
puts strings from the latter into ^P recall history (although they'll
only render correctly if nothing after the first character needs
special handling), but nothing was putting that same info into
DUMPLOG.  This fix is pretty clumsy but eliminates the discrepancy.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Tue, 5 Feb 2019 02:39:53 +0000 (21:39 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agotty ^P recall for dolook/whatis feedback
PatR [Tue, 5 Feb 2019 02:24:17 +0000 (18:24 -0800)]
tty ^P recall for dolook/whatis feedback

Noticed while testing the history suppression:  if you have DECgraphics
enabled and look at a graphics character on the map, the topline shows
x      description of x
where 'x' is displayed as it appears on the map (line drawing char).
^P for msg_window:single knows about that and reproduces the effect if
you recall such a line.  But msg_window:full/combination/reverse didn't
know about that and dumped it as-is into text output, ending up with a
strange 8-bit character for 'x' instead of the line drawing one.

I think other rendering schemes will be unaffected by this.  It's just
duplicating what is done for msg_window:single.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Tue, 5 Feb 2019 01:54:05 +0000 (20:54 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agofixes entry for amiga fix
nhmall [Tue, 5 Feb 2019 01:27:18 +0000 (20:27 -0500)]
fixes entry for amiga fix

Also, a pull request 175 showed up after the earlier commit for the missing
continuation lines, so mark the pull request as closed.

closes #175

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Tue, 5 Feb 2019 01:05:28 +0000 (20:05 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoMerge branch 'NetHack-3.6.2-beta01' of https://rodney.nethack.org:20040/git/NHsource...
nhmall [Tue, 5 Feb 2019 01:04:59 +0000 (20:04 -0500)]
Merge branch 'NetHack-3.6.2-beta01' of https://rodney.nethack.org:20040/git/NHsource into NetHack-3.6.2-beta01

5 years agoanother contributed Amiga bit, missing line continuations in Makefile
nhmall [Tue, 5 Feb 2019 01:02:53 +0000 (20:02 -0500)]
another contributed Amiga bit, missing line continuations in Makefile

5 years agotty ^P fixes entry
PatR [Tue, 5 Feb 2019 00:52:37 +0000 (16:52 -0800)]
tty ^P fixes entry

Left this out of previous commit.

5 years agotty ^P message recall
PatR [Tue, 5 Feb 2019 00:46:04 +0000 (16:46 -0800)]
tty ^P message recall

Extend 'putstr(WIN_MESSAGE, attribute, string)'s attribute so that
'custompline(SUPPRESS_HISTORY, ...)' can work with ^P's message
history like DUMPLOG history, in order to keep autodescribe feedback
and intermediate prompts for multi-digit count ('Count: 12', 'Count:
123') prompts out of recall history.  The old autodescribe behavior
could easily push all real messages out of the recall buffer when
moving the cursor around for getpos, and the count behavior looked
silly for a four or five digit gold count if you set the msg_window
option to 'full' or 'combination' and viewed them all at once.

Other interfaces may want to follow suit, but this doesn't force them
to make any changes.  I added a hook for "urgent messages" that might
be rendered in bold or red or some such and/or override the use of
ESC at --More-- from suppressing further messages, but there aren't
any custompline(URGENT_MESSAGE, ...) calls (potentially "You die...",
for instance) to exercise it.  Other people have implemented similar
feature it different ways and I'm not sure whether this one is really
the way to go since the core needs to categorize each message that it
deems to be urgent.  MSG_TYPE:stop may be sufficent, although MSG_TYPE
matching can entail a lot of regexp execution overhead at run-time.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Mon, 4 Feb 2019 23:37:15 +0000 (18:37 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agovariable moved to context long ago
nhmall [Mon, 4 Feb 2019 23:35:17 +0000 (18:35 -0500)]
variable moved to context long ago

just reported, a variable moved to context long ago

5 years agofunction call through pointer
PatR [Mon, 4 Feb 2019 23:01:36 +0000 (15:01 -0800)]
function call through pointer

There be others of these.  Don't for ANSI C for something this simple.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Sun, 3 Feb 2019 22:28:27 +0000 (17:28 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agogoodtype() reversal
PatR [Sun, 3 Feb 2019 22:08:29 +0000 (14:08 -0800)]
goodtype() reversal

Reverse an erroneous part of the vault guard commit.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Sun, 3 Feb 2019 02:09:23 +0000 (21:09 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agofix #H7173 / github #101 - vault exit
PatR [Sun, 3 Feb 2019 01:37:06 +0000 (17:37 -0800)]
fix #H7173 / github #101 - vault exit

Fixes #101

If you tell the vault guard your name, drop carried gold, wait one
turn, then pick up the gold again, the guard will move a step away
during the wait.  If you teleport away, the guard will seal vault
walls and then park himself on the one-square (so far) temporary
corridor adjacent to the vault wall.  Periodically he'll say "Move
along!" and the hero will hear it, regardless of location on the
level.  Unless you dig a corridor to rescue him, or one or more of
the vault's walls get breached again, he will never move.

The report emphasized that you could use this to steal the vault's
gold, but it relies on being able to teleport beyond the gaurd's
reach and if you can do that, you might as well do so before the
guard comes.  The stranded guard, and him saying "Move along!" when
no longer leading hero out of the vault, are more significant bugs.

Bonus fix:  if the game ends and the guard seals up the vault while
the hero is in a spot that gets fixed up (vault wall or temporary
corridor) don't give the "You are encased in the rock" message if
game end was due to death rather than quit.

5 years agoobject classes for wizard mode identify
PatR [Sat, 2 Feb 2019 02:40:45 +0000 (18:40 -0800)]
object classes for wizard mode identify

When ^I was changed to allow picking more than one item to make
its temporary identification become persistent, group accelators got
left out.  So to pick all potions, you needed to select them letter
by letter (or via '.' to select everything, then deselect non-potions
letter by letter).  Now you can use '!' to select all potions.

5 years agocuring deafness
PatR [Sat, 2 Feb 2019 02:24:23 +0000 (18:24 -0800)]
curing deafness

Make healing magic which cures blindness also cure deafness.  So,
drinking non-cursed potion of healing or any extra healing or full
healing; breathing fumes from blessed potion of healing or non-cursed
potion of extra healing or any potion of full healing; prayer reward
to cure blindness as a minor trouble.  (Doesn't affect unicorn horns
which already treat deafness and blindness as two distinct troubles
that are eligible to be cured.)

More of a missing feature than a bug fix, so I listed it in the new
features section of the fixes file.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Sat, 2 Feb 2019 00:23:44 +0000 (19:23 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoFix making fake leash object with leashmon set
Pasi Kallinen [Fri, 1 Feb 2019 16:19:00 +0000 (18:19 +0200)]
Fix making fake leash object with leashmon set

When farlooking at a leash glyph on a map, the fake object should never
have leashmon set. This happened when a mimic was on the same spot
and was mimicing a corpse or statue, due to leashmon and corpsenm
using the same field.

5 years agoFix logic error in stuck_in_wall
Pasi Kallinen [Fri, 1 Feb 2019 11:38:21 +0000 (13:38 +0200)]
Fix logic error in stuck_in_wall

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Fri, 1 Feb 2019 00:56:38 +0000 (19:56 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agomissing maybe_reset_pick update
PatR [Fri, 1 Feb 2019 00:50:17 +0000 (16:50 -0800)]
missing maybe_reset_pick update

Accidentally omitted from the commit for 'stale lock picking context'
8bf16b940e15cac75cea2a4de92738c90fa71688.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Fri, 1 Feb 2019 00:29:55 +0000 (19:29 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agostale lock picking context
PatR [Thu, 31 Jan 2019 23:50:12 +0000 (15:50 -0800)]
stale lock picking context

Lock context wasn't being cleared if it was for a container and that
container got destroyed.  Case discovered was forcelock() ->
breakchestlock() -> delobj() (sometimes the container is destroyed
rather than just breaking its lock) followed by #wizmakemap (replace
current level) and maybe_reset_pick() trying to check whether
xlock.box was being carried.  But being interrupted, destroying the
container or dropping it down a hole to ship it to another level, then
attempting to resume picking the lock would also find a stale pointer.

5 years agoMerge branch 'NetHack-3.6.2-beta01' of https://rodney.nethack.org:20040/git/NHsource...
nhmall [Thu, 31 Jan 2019 23:22:55 +0000 (18:22 -0500)]
Merge branch 'NetHack-3.6.2-beta01' of https://rodney.nethack.org:20040/git/NHsource into NetHack-3.6.2-beta01

5 years agoFix warning
Pasi Kallinen [Thu, 31 Jan 2019 19:31:26 +0000 (21:31 +0200)]
Fix warning

Fix "using integer constants in boolean context"

5 years agoRevert "adjust sample config file when STATUS_HILITES isn't defined"
nhmall [Thu, 31 Jan 2019 18:03:00 +0000 (13:03 -0500)]
Revert "adjust sample config file when STATUS_HILITES isn't defined"

This reverts commit 03e6c26af6de62f6a970d5b0c9845f924bfc93f4.

5 years agofix monstone() ... dealloc_obj() panic
PatR [Thu, 31 Jan 2019 12:22:04 +0000 (04:22 -0800)]
fix monstone() ... dealloc_obj() panic

Fuzzer feebdack.  When turning a monster into a statue, monstone()
builds a linked list of mon->minvent items to put into that statue.
It doesn't use obj_extract_self() to take them off again, leaving
obj->nobj non-Null.  Not noticed for the normal case where each item
gets linked into the container's contents, but triggers panic if an
item merges with something already put inside so gets removed.

Suddenly, the dungeon collapses.
dealloc_obj with nobj
[2] 0x01000c4193 panic + 995
[3] 0x0100155427 dealloc_obj + 71
[4] 0x010021d4de obfree + 686
[5] 0x01000f2f92 merged + 834
[6] 0x010015356e add_to_container + 126
[7] 0x01001628ac monstone + 636

I don't know why the petrified monster's mergeable inventory wasn't
already merged while in inventory.

5 years agoOption handling return values and errors
Pasi Kallinen [Thu, 31 Jan 2019 08:46:41 +0000 (10:46 +0200)]
Option handling return values and errors

Fix some options not stopping when encountering an error,
some didn't show up any errors, and other minor fixes.

5 years agoSilence sprintf format warnings and dehardcode buffer sizes
Pasi Kallinen [Thu, 31 Jan 2019 07:42:03 +0000 (09:42 +0200)]
Silence sprintf format warnings and dehardcode buffer sizes

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Thu, 31 Jan 2019 04:11:49 +0000 (23:11 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoanother warning fix
PatR [Thu, 31 Jan 2019 00:25:31 +0000 (16:25 -0800)]
another warning fix

Only appeared when ISAAC64 was disabled.

5 years agorng bit
nhmall [Wed, 30 Jan 2019 16:08:11 +0000 (11:08 -0500)]
rng bit

5 years agoadd a descriptive comment to integer.h
nhmall [Wed, 30 Jan 2019 13:30:02 +0000 (08:30 -0500)]
add a descriptive comment to integer.h

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Wed, 30 Jan 2019 03:33:18 +0000 (22:33 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agofix a likely typo/transcription error in integer.h
nhmall [Wed, 30 Jan 2019 03:12:09 +0000 (22:12 -0500)]
fix a likely typo/transcription error in integer.h

5 years agofix build
nhmall [Wed, 30 Jan 2019 01:29:56 +0000 (20:29 -0500)]
fix build

../sys/unix/unixmain.c: In function ‘sys_random_seed’:
../sys/unix/unixmain.c:779:29: error: expected expression before ‘long’
         fread(&seed, sizeof long, 1, fptr);

5 years agorandrole() fix
PatR [Wed, 30 Jan 2019 01:14:55 +0000 (17:14 -0800)]
randrole() fix

Give all the calls to randrole() its new argument.

5 years agoMerge branch 'NetHack-3.6.2-beta01' of https://rodney.nethack.org:20040/git/NHsource...
nhmall [Wed, 30 Jan 2019 01:03:21 +0000 (20:03 -0500)]
Merge branch 'NetHack-3.6.2-beta01' of https://rodney.nethack.org:20040/git/NHsource into NetHack-3.6.2-beta01

5 years agoupdate the long version info generated by makedefs
nhmall [Wed, 30 Jan 2019 01:01:31 +0000 (20:01 -0500)]
update the long version info generated by makedefs

5 years agolast warning fix
PatR [Tue, 29 Jan 2019 23:27:22 +0000 (15:27 -0800)]
last warning fix

Compiler gave four diagnostics about 'seed' being used uninitialized
if 'no_seed' were false, but two of those were alternate suggestions
for how to suppress them.

5 years agoanother warning fix
PatR [Tue, 29 Jan 2019 23:18:17 +0000 (15:18 -0800)]
another warning fix

Comparison between signed and unsigned.  Compiler stupidity since
the range of possible values that signed 'i' can take is limited and
never negative.

5 years agowarning fix
PatR [Tue, 29 Jan 2019 23:00:07 +0000 (15:00 -0800)]
warning fix

The new code provoked several warnings; this fixes one of them.
Moving the declaration of 'rolecount' would have been sufficient,
but I've gone another way.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Tue, 29 Jan 2019 22:23:03 +0000 (17:23 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agomore window port code uses of obj_to_glyph()
nhmall [Tue, 29 Jan 2019 22:21:55 +0000 (17:21 -0500)]
more window port code uses of obj_to_glyph()

5 years agotry fix for Qt4 using a core macro
nhmall [Tue, 29 Jan 2019 22:07:00 +0000 (17:07 -0500)]
try fix for Qt4 using a core macro

../win/Qt4/qt4inv.cpp:41:26: error: macro "obj_to_glyph" requires 2 arguments, but only 1 given
  glyph=obj_to_glyph(nhobj);
                          ^
../win/Qt4/qt4inv.cpp: In member function ‘void nethack_qt4::NetHackQtInvUsageWindow::drawWorn(QPainter&, obj*, int, int, bool)’:
../win/Qt4/qt4inv.cpp:41:8: error: ‘obj_to_glyph’ was not declared in this scope
  glyph=obj_to_glyph(nhobj);
        ^

5 years agobuild bits - mostly for Windows
nhmall [Tue, 29 Jan 2019 16:31:41 +0000 (11:31 -0500)]
build bits - mostly for Windows

Allow sys/share/random.c to be included in the build
always, even if USE_ISAAC64 is defined, by making most
of its contents conditional in that case.

That avoids Makefile tinkering when going back and
forth between USE_ISAAC64 and not during testing.

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Tue, 29 Jan 2019 14:27:17 +0000 (09:27 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoWorkaround for including stdint.h on older GCC versions
Patric Mueller [Tue, 29 Jan 2019 14:03:07 +0000 (15:03 +0100)]
Workaround for including stdint.h on older GCC versions

GCC doesn't set __STDC_VERSION__ for older versions. It is verified to
be set by default on 5.3.0.

Older versions set __INT64_MAX__ though (oldest version tested is
4.5.4).

If compilation with older GCC versions is required, set -std=c99. Even
3.4.6 is supported with this setting.

5 years agofixes36.2 update for added isaac64 prng
nhmall [Tue, 29 Jan 2019 12:38:57 +0000 (07:38 -0500)]
fixes36.2 update for added isaac64 prng

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Tue, 29 Jan 2019 12:22:23 +0000 (07:22 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agoMerge branch 'alex-and-isaac' into NetHack-3.6.2-beta01
nhmall [Tue, 29 Jan 2019 12:20:17 +0000 (07:20 -0500)]
Merge branch 'alex-and-isaac' into NetHack-3.6.2-beta01

5 years agoConsistently use rn2 in trap detection
Alex Smith [Tue, 29 Jan 2019 03:37:08 +0000 (03:37 +0000)]
Consistently use rn2 in trap detection

It's a gameplay-affecting action (it updates the character's
memory of the map), and there's no way to spam it without using
resources, so rn2 is safe.

5 years agoinclude the rnd.c bits
nhmall [Tue, 29 Jan 2019 00:54:18 +0000 (19:54 -0500)]
include the rnd.c bits

5 years agomerge Alex's dual rng proposal with the isaac64 rng code and adjust
nhmall [Tue, 29 Jan 2019 00:43:55 +0000 (19:43 -0500)]
merge Alex's dual rng proposal with the isaac64 rng code and adjust

This is branched from Alex's hallu-rng-stability branch,
with two build corrections (detect.c, zap.c), and merged
with  the isaac64 branch that we have ready to go.

Alex's dual rng is supported by setting up the array
of multiple isaac64 contexts.

I stuck with Alex's approach of passing the rng function
name around as the parameter (rng or rn2_on_display_rng)
for the new additional parameter needed for
set_random(), init_random(), reseed_random(),
and init_isaac64().

5 years agoMerge branch 'countermeasures' into alex-and-isaac
nhmall [Mon, 28 Jan 2019 23:37:50 +0000 (18:37 -0500)]
Merge branch 'countermeasures' into alex-and-isaac

5 years agoMerge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2
nhmall [Mon, 28 Jan 2019 21:36:23 +0000 (16:36 -0500)]
Merge branch 'NetHack-3.6.2-beta01' into NetHack-3.6.2

5 years agofix #H8072 - failing wish segfaults
PatR [Mon, 28 Jan 2019 17:10:52 +0000 (09:10 -0800)]
fix #H8072 - failing wish segfaults

Having an artifact wish be refused uses zeroobj and code which
followed was attempting to update its weight, triggering a segfault
now that zeroobj is 'const'.

5 years agoMerge branch 'countermeasures' of https://rodney.nethack.org:20040/git/NHsource into...
nhmall [Mon, 28 Jan 2019 14:45:31 +0000 (09:45 -0500)]
Merge branch 'countermeasures' of https://rodney.nethack.org:20040/git/NHsource into countermeasures

5 years agoWindows build bit
nhmall [Mon, 28 Jan 2019 14:44:35 +0000 (09:44 -0500)]
Windows build bit

5 years agoworkaround for C89 not supporting inline
Patric Mueller [Mon, 28 Jan 2019 14:00:04 +0000 (15:00 +0100)]
workaround for C89 not supporting inline

5 years agontconf.h ensure Rand is always something
nhmall [Sun, 27 Jan 2019 14:34:18 +0000 (09:34 -0500)]
ntconf.h ensure Rand is always something

order of preference:
USE_ISAAC64
RANDOM
C routine

5 years agodetect DEV_RANDOM fopen failure and fall back, noting it in paniclog
nhmall [Sat, 26 Jan 2019 16:26:02 +0000 (11:26 -0500)]
detect DEV_RANDOM fopen failure and fall back, noting it in paniclog

5 years agodetect rng seed strength at runtime based on algorithm not compile time based on...
nhmall [Fri, 18 Jan 2019 23:42:42 +0000 (18:42 -0500)]
detect rng seed strength at runtime based on algorithm not compile time based on platform features

5 years agoevery platform provides sys_random_seed() and SYS_RANDOM_SEED goes away
nhmall [Thu, 17 Jan 2019 23:30:56 +0000 (18:30 -0500)]
every platform provides sys_random_seed() and SYS_RANDOM_SEED goes away

5 years agoDon't define Rand() if isaac64 is used
Patric Mueller [Thu, 17 Jan 2019 10:48:58 +0000 (11:48 +0100)]
Don't define Rand() if isaac64 is used

5 years agodefine Rand() in isaac4 config
nhmall [Tue, 15 Jan 2019 01:36:50 +0000 (20:36 -0500)]
define Rand() in isaac4 config

Rand() was typically defined to random() or to rand().

gcc seems to provide a random() to link to on linux
when sys/share/random.c is linked in, but other platforms
such as Windows got an undefined refence to random()
when RANDOM wasn't defined.

The only direct use seems to be in get_rnd_txt() these
days, in rumors.c

Under the USE_ISAAC64 config, neither srandom()
nor srand() are being invoked to seed those routines,
and it really should be using isaac64 when USE_ISAAC64
is defined anyway.

5 years agosome system-specific adjustments for RNG routines
nhmall [Mon, 14 Jan 2019 01:54:11 +0000 (20:54 -0500)]
some system-specific adjustments for RNG routines

move some system-specific seed-related stuff from hacklib.c to
a system-specific source file and #define SYS_RANDOM_SEED to
utilize it during build.

Windows changes for random seed generation using
crypto next gen (CNG) api routines.

Corresponding vms changes due to disentangling of VMS and
unix when the unix seed bits got moved (untested).

5 years agointegrate isaac64 into nethack
Patric Mueller [Sun, 13 Jan 2019 15:36:55 +0000 (16:36 +0100)]
integrate isaac64 into nethack

Also removed the float code from isaac64 as they are not used in
NetHack.

5 years agoadd file for the isaac64 random number generator
Patric Mueller [Sat, 12 Jan 2019 20:36:08 +0000 (21:36 +0100)]
add file for the isaac64 random number generator

This is the version from the Comprehensive C Archive Network, licensed
under the CC0 "No Rights Reserved" Creative Common License.
http://ccodearchive.net/info/isaac.html

5 years agofix check for stdc version in include/integer.h
Patric Mueller [Sun, 13 Jan 2019 14:17:59 +0000 (15:17 +0100)]
fix check for stdc version in include/integer.h

5 years agoreseed during level change to prevent deduction of rng state
Patric Mueller [Fri, 11 Jan 2019 18:23:01 +0000 (19:23 +0100)]
reseed during level change to prevent deduction of rng state

For platforms that read from the system's random number generator,
reseed during level change, before the map of a new level is created and
after level creation has finished.

5 years agoread rng seed from random number source device
Patric Mueller [Fri, 11 Jan 2019 18:08:22 +0000 (19:08 +0100)]
read rng seed from random number source device

Linux and BSD system have random number source devices that can be used
as source for a unguessable seed source.

Other platforms fall back to generate the seed with gettime().