Dungeon Master Forum

By 'occasionally' I think I saw it less than a dozen times over a neophyte to um-master ninja length game... Yeah that's not going to be easily reproducible...

I'll bring it up again if I can figure out a way to reproduce it, otherwise lets ignore it.

There's a glitch where I'm not 100% sure whether its a bug or some 'compatibility with the original' thing...

Occasionally, a critter will move toward the party at the same time as you are firing a projectile at it. If the timing is right (wrong?) the critter can move forward at the same time as ...

Unfortunately, its also still a bit buggy. This one is minorish- it doesn't make the game unplayable, but it can 'unfairly' need a tiny switch to open a mandatory moving wall.

http://en.file-upload.net/download-1131 ... 5.zip.html

http://en.file-upload.net/download-11306830/probablysolvable20160214.zip.html
fixes a nasty crashing bug and some more minor glitches.

Works best with 0.66c or later.

Note that I was able to continue an existing game via:
1) get the new DSB in a new folder
2) setup a new module folder
3) shift my ...

There's a dsb66c since then.

I'm three and a half levels deeper without finding any new problems so that may wind up being 0.67

arch.ARCH_NAME isnt available when I'm setting the initial item configuration up, but that's not a big deal- I need to already know the arch names because I'm in the process of constructing them.

Even better, if there's no curse actually running, then the exvar arch is just a cache. And if I'm just ...

Yuck. This one is messy. Your code was doing something wrong... or, more properly, something a little too clever for DSB. It was a condition that I did not at all anticipate so DSB had no error checking and simply crashed.
My preference for DSB could likely be summarized as "Here is an engine I ...

Ok, with a bit of tweaking around the random number generator, I can semi-reliably reproduce this in a save moved over to 0.66b

http://en.file-upload.net/download-11298182/probablysolvablekp.zip.html

If you stand in front of the gor face for a while, it should zap the party and make some equipment ...

Cool. I'll go find another problem then 8-)

With the new random number generator, the semantics of dsb_rand have changed

Before, it generated a random number between min and max, inclusive. Now, the highest number returned is max-1

Something like

local dist={}
local temp
for i=0,999999 do
temp=dsb_rand(0,3)
if (dist[temp] == nil) then ...

Unfortunately saves dont seem to be compatible

I thought this was a version change issue, but it appears to be a general failure to load any saves


Parsing test_dungeon/dungeon.lua
PROGRAM ERROR!
Reason: Check value overwritten 68 [throwtrolin] @ (1,10,7) 0
Stack Dump:
DSBmain
DSBgameloop ...

Awesome! My party can starve to death without the game crashing!

Now to try to replicate the previous crash! Unfortunately saves dont seem to be compatible, so its going to be a little tricky. Hopefully if I move some of the broken items and set up a teleporter to the right area (and clear out the ...

Hmm.

Let me check if I'm spawning them in dumb initial positions. (But if so, why wasn't 0.65 crashing?)

[hack hack hack]

Hmm

Is this related: I made a tiny test dungeon to try to put critters in wacky locations, including a teleporter to see if spinning them might break stuff. I am not seeing ...

While trying to get a reproducible version of my last crash, I have found another one:


PROGRAM CRASH!
Location: 0 20 1
Reason: Invalid slave pointer to pointerize (inst 10024 dir -2)
Stack Dump:
DSBmain
DSBgameloop
gameloop.process_tick
run_timers
monster_acts
monster_move
lua.sys_ai_far
lua.dsb ...

I managed to trigger a segfault from somewhere in my temporary curses code.


PROGRAM CRASH!
Location: 8 12 26
Reason: Segmentation Fault
Stack Dump:
DSBmain
DSBgameloop
run_timers
deliver_msg(leggings_cursed)
lua.dsb_qswap
luaobjarch


I'll dig around and see if
a) I can reproduce it (preferably ...

Thanks for the fix. Since I was working with lua generated items anyway, my workaround was to manually call the method on each item* at the time it was created. But having an actual fix is always better of course 8-)

* after ensuring it would behave if double-called

And now

http://en.file-upload.net/download-1126 ... c.zip.html

(not major, but cleans up a bit of text glitchiness because on_spawn didn't work in the way I expected)

Hi

on_spawn doesn't appear to trigger if an item is created inside anther item. Should this be the case? Or am I misusing the API?

Here is a simple example, with a gem object that logs it's ID as an on_spawn event
http://en.file-upload.net/download-11263868/noonspawn.zip.html

loggem #20 logs ...

Here is a new version

http://en.file-upload.net/download-11263671/probablysolvable20160131b.zip.html

This one adds secret areas to the dungeon. These are guaranteed* to be non-essential for completing the dungeon, but item drops in the secret areas are calculated as much deeper into the dungeon ...

Cool. I wasn't sure any change was necessary, but if there's a drop in replacement MT in a suitable license, a better RNG is always better 8-)

thanks

Oh, and in case anyone winds up here looking for larger dsb_rands, I'll leave this here. (nb I know there are statistical issues with this style of algorithm, but I'm not using it for cryptography, so it ought to be OK)


-- random number generator suitable for ranges up to a couple hundred million ...

Oh two additional things:

1) Is there sufficient data up there, or would you rather the full reams and reams of it? (I would be looking at putting a zip file somewhere rather than dumping all that to the forum)

2) I'm not sure that any of this actually constitutes a problem with dsb_rand, just a ...

The overlap case for lua random is a bit trickier, so I added a histogram


math.randomseed(34892895636) -- arbitrary seed, for reproducability

local dist={}
local hist={}
local temp
for i=0,999999 do
temp=math.random(0,30000)
if (dist[temp] == nil) then
dist[temp] = 1
else
dist[temp] = dist ...

Lua's random falls through to ANSI C random, which can have a range (depending on the implementation) "as low as 32767"

using

math.randomseed(34892895636) -- arbitrary seed, for reproducability

local dist={}
local temp
for i=0,999999 do
temp=math.random(0,9999999)
if (dist[temp] == nil) then ...

If I put


local dist={}
local temp
for i=0,999999 do
temp=dsb_rand(0,30000)
if (dist[temp] == nil) then
dist[temp] = 1
else
dist[temp] = dist[temp]+1
end
end
for k,v in pairs(dist) do
__log(tostring(k) .. " -> " .. tostring(v))
end


in startup.lua I get


Lua: 1 -> 61
Lua: 2 -> 60
Lua ...

If I put


local dist={}
local temp
for i=0,999999 do
temp=dsb_rand(0,99999999)
if (dist[temp] == nil) then
dist[temp] = 1
else
dist[temp] = dist[temp]+1
end
end
for k,v in pairs(dist) do
__log(tostring(k) .. " -> " .. tostring(v))
end


in startup.lua I get


Lua: 1 -> 34
Lua: 2 -> 35 ...

A correction: while dsb_rand may not *crash* for very large ranges, I wouldn't call the results of any range above about 3200 really viable.

a) dsb_rand seems to use "generate a randomish number in 0 .. 32767 and then modulus it into the required range" which means that:

b) if the range is greater ...

And here's the first inevitable bugfix 8-)

http://en.file-upload.net/download-10810563/probablysolvable20150801.zip.html

If a pressure pad was right next to a moveable wall, then you could step off the pad into the wall and it would close behind you. There was an 'emergency pad' mechanism to open ...

:arrow: The default dungeon is randomly generated and is probably solvable, but I haven't playtested it so I don't know.

So I've been tweaking Joramun' Chaos Hack for a while, but I ran into a problem where I was pretty sure it was possible to wind up with a situation where you needed to use ...

Hi

Under some circumstances, it is possible for teleports and messages to combine to crash the system. The teleport can get delayed indefinitely, and if the target is deleted for any reason then everything falls over due to messages involving a now-invalid id...

simple test case:
http://en.file ...