Time  Nick        Message
02:07 hmmmm       alright
02:07 hmmmm       i came up with a much more sane mapgen param setup
02:07 hmmmm       MapgenParams is not a polymorphic class
02:08 hmmmm       MapgenParams has a pointer to a MapgenSpecificParams, which is the polymorphic class
02:08 hmmmm       instead of reallocating the whole MapgenParams structure and updating pointers to everything that stays the same, the only thing that is modified is the specific params
02:09 hmmmm       there is *only* one instance of the specific params, which is statically allocated and has the hardcoded parameters set as default
02:10 hmmmm       every modification just overwrites this single data structure
02:10 hmmmm       there is no more any mapgen noise param related things in defaultsettings.cpp
02:11 hmmmm       i added a new set of Settings methods called tryGet*() that returns true/false if it was able to get that setting, and the variable to get the setting is apassed by reference
02:11 hmmmm       this way if you really don't give a shit if the setting is present or not you don't have to do a try/catch statement for each
09:57 sapier      https://github.com/minetest/minetest/pull/1131 ShadowNinja I merged your comments is it ok now?
12:20 sapier      missing TOSERVER_RECEIVED_MEDIA in 0.4.9 seems to be more of a problem then I initially expected. as 0.4.9 client does spam console with error messages in "half initialized" state
12:20 sapier      I can imagine various options to fix it:
12:20 sapier      1) bump protocol and handle "old" clients in a special way
12:21 sapier      2) add a connection timeout ... bad option as it'd kick clients on slow lines too
12:21 sapier      3) a server notification about "possible old client" but don't do anything else ... almost as doing nothin
12:21 sapier      4) do nothing
12:21 sapier      does anyone have a better idea?
12:28 sapier      http://i.imgur.com/RM7HIa0.png this is what will happen to 0.4.9 clients on connecting without having all textures. It's permanent for this session but wont happen again on next connect (unless server got new media in between)
14:07 sapier      https://github.com/minetest/minetest/issues/1127 comments?
16:14 ShadowNinja sapier: https://github.com/minetest/minetest/pull/1122/files#diff-2543b17841165757ac531ffedd0b5fddR169 Space after , here and { on the same line.
16:20 sapier      nope { on same line isn't used anywhere?
16:27 sapier      "...m_ for members.  The latter is discouraged for newer code (since one can  simply notice there is no variable by that name declared in that  method's scope)..." ???????
16:28 sapier      is there anyone out there who memorized full class definitions of minetest?
16:28 sapier      that's from coding style
16:30 hmmmm       intellisense bro
16:31 sapier      I don't consider hinting to a special tool to be a valid reason for a coding style decision ... especially if it's a commercial quite expensive one
16:31 hmmmm       besides, you just double click on the variable name and it gets highlighted everywhere throughout the rest of the function so that's a pretty good visual aide
16:31 hmmmm       well I mean intellisense in the general class member hinting popup thing
16:31 hmmmm       some people call it code completion but it's not quite that
16:32 hmmmm       about the code style, it's not like it matters too much because everybody ends up doing their own thing anyway
16:32 sapier      tools change but code stays sometimes for decades so relying to a certain feature isn't a good idea
16:32 hmmmm       like the no-whitespace-around-control-statements thing
16:32 hmmmm       if(foobar)
16:32 hmmmm       and then an opening brace on a new line
16:32 hmmmm       {
16:33 hmmmm       eww
16:33 sapier      well maybe because we didn't ever have a real complete coding style ... current one does look like a collection of random thoughts
16:33 hmmmm       what I'm asking for something like the Linux kernel style
16:33 hmmmm       that's because the Linux kernel style is the first link on the page, that's the completed thing
16:33 sapier      maybe but kernel style doesn't match as we do have c++ code not c code
16:33 hmmmm       yeah
16:34 hmmmm       I try to fill in some things for the C++ specific things
16:34 hmmmm       s/things/constructs/
16:35 sapier      and m_ notation has another benefit, you don't have to name your parameters size_ in setSize() function
16:35 hmmmm       alright, maybe I should attach an importance level to each of the items (rule -> shall, guideline -> should, suggestion -> may)
16:36 hmmmm       when I wrote that up also I tried to make it match my own style so I didn't have to go back and rewrite my own stuff
16:36 hmmmm       but
16:36 hmmmm       i'd rather work on the mapgen param fixing
16:36 sapier      well it's not a "must" rool so I guess it's not really necessary to discuss g_ m_ ... it to death :-)
16:36 sapier      rool --> rule :-)
16:37 hmmmm       i think g_ should stick around because there really aren't too many globals
16:37 ShadowNinja sapier: It is. And it's in the style guidelines. With the exception of functions. (And classes, namespaces, and the like)
16:37 hmmmm       and they're indistinguishable from members in the same sense of not being in a local scope
16:37 sapier      actually we should try to get rid about globals ... at least if it's possible
16:38 hmmmm       well all of the global nparams_<mgname>_def_<noisename>s are going
16:38 sapier      well ShadowNinja I'm looking at style guideline for exactly that reason, there's not a single comment about where to place {
16:38 hmmmm       like the Linux kernel style
16:38 hmmmm       or BSD KNF
16:38 hmmmm       K&R
16:39 sapier      well code most of time does exactly opposite
16:39 hmmmm       void Function(void *param1, char *param2, ...goes over 80 chars...
16:39 hmmmm       \tint param5)
16:39 hmmmm       {
16:39 hmmmm       if (!param1)
16:39 sapier      as I saif, code doesn't stick to that rule
16:39 hmmmm       \tparam1 = something;
16:39 sapier      so do we change 90% of minetests code ?
16:39 hmmmm       if (morecomplexifstatement()) {
16:40 hmmmm       do stuff here;
16:40 hmmmm       } else {
16:40 hmmmm       so on and so forth
16:40 sapier      look at code!
16:40 hmmmm       no we don't change 90% of minetest's code
16:40 hmmmm       you just try to do the 'right' thing from now on
16:40 sapier      why do we define a style we'd have to change all of our code?
16:40 ShadowNinja We update things as we go along ^.
16:40 hmmmm       you don't need to change things
16:41 sapier      come one guys do you really not understand my point?
16:41 hmmmm       I think the rule of thumb is that you typically stick to the module's existing code style unless you're changing over 1/3rd of it
16:41 hmmmm       soooo if it's a brand new source file why not just do it in the new style
16:42 sapier      ok if you want to get a new style we need a way to automate style fixes.
16:42 hmmmm       non ononnonoon
16:42 hmmmm       astyle is out of the question
16:42 hmmmm       there have been people who brought this up before and it's definitely not going to be automated
16:42 sapier      noone will fix all of those things and I'm really anoyed about discussing when is 1/3 changed or not
16:43 hmmmm       right
16:43 hmmmm       nobody discusses that though
16:43 hmmmm       because it doesn't really matter that much
16:43 sapier      see discussion with ShadowNinja, that file uses quite mixed style
16:43 hmmmm       lemme see
16:43 PilzAdam    sapier, astyle misses some options to configure it to match our style
16:43 hmmmm       jesus christ
16:44 hmmmm       ShadowNinja, it'd be really great if you stop micromanaging people
16:44 ShadowNinja Hmmm?
16:45 hmmmm       for people making lots of changes, unless the style is wayy out of wack, just let it go
16:45 hmmmm       focus on the big things like program logic
16:45 hmmmm       this is what matters most^
16:45 ShadowNinja Alright.
16:45 hmmmm       reviewing a pull request doesn't mean you need to scruitinize the spacing or whatever
16:45 hmmmm       that's just mostly frustration
16:46 hmmmm       i mean if they get it right that's awesome
16:46 sapier      well don't blame ShadowNinja basically he's right but how to handle files like that without really style in there
16:46 hmmmm       especially for modules with the old irrlicht style
16:46 hmmmm       like if somebody were to add shit like
16:46 hmmmm       if( 0 != fooSomething() )
16:46 hmmmm       { doBlah();
16:46 ShadowNinja sapier: https://github.com/sapier/minetest/commit/695230db86918929c04aead76931f21b3e6d3b73#diff-18513665750ef5adf42b5ec29e14162eR1036 Um...  You're missing a number of combinations, I think you should just remove the check as resolving works properly with those addresses.
16:46 hmmmm       into emerge.cpp
16:47 hmmmm       I think i'd have a fit
16:47 sapier      maybe we should add a function isNullAddress(std::string) :)
16:48 ShadowNinja sapier: Use resolveString and check if it's null.  But that's what is does if the conditional succeeds.
16:48 sapier      ok you're right keep it simple
16:48 hmmmm       i know MIRSA says that you should compare explicitly against a value but it's really not unclear in 99% of cases
16:49 ShadowNinja And it looks like you're using the bind address on the client still.
16:49 hmmmm       this is why i like to do if (!foo) instead of if (foo == NULL)
16:49 hmmmm       but it doesn't matter, that's up to personal taste because it's unregulated
16:50 ShadowNinja sapier: You also do the check here, although this patch shouldn't touch the meinmenu at all.
16:50 ShadowNinja https://github.com/minetest/minetest/pull/1131/files#diff-16525c326f5842e04cbdf0b3ae8254fcR961
16:50 sapier      if(!foo) may fail dependent on truth value of architecture ... but I don't know any where it may fail ... gues that's theoretical case only
16:51 hmmmm       no it wont
16:51 hmmmm       because
16:51 sapier      well that's the addition I made now you can control the binding from main menu too once you set it
16:51 hmmmm       you're implicitly casting foo to an integer
16:51 ShadowNinja NULL == 0L == false == 0 And 0 is the only value that fails.
16:52 hmmmm       straight from the standard, "the integer representation of a null pointer shall compare true to 0"
16:52 sapier      you know as good as me compilers not always correctly implement standard, but that's more of a problem in embedded development then in application dev
16:52 ShadowNinja sapier: And this is only available in the server tab, and is shown as the "Bind address" and preferable marked as advanced.
16:53 hmmmm       yes and like i said in an embedded environment where that happens to *not* work, that's not standard C anymore
16:53 ShadowNinja s/\./?/
16:53 ShadowNinja I don't think such an advanced option as the bind address should be in the menu.
16:53 hmmmm       and most of them aren't standard C anyway, they don't even claim to be, they're a "C-like language"
16:53 sapier      everyone capable of enabling this already prooved to have advanced knowledge
16:55 ShadowNinja As for https://github.com/minetest/minetest/issues/1127 I guess the default should just be raised to max_clients * 4 or so.
16:56 hmmmm       ShadowNinja, see: http://c-faq.com/null/ptrtest.html
16:57 VanessaE    max_clients * 4 would be twice my current setting, which would be overkill probably in my case, but (currently connected) * 3 or 4 would be pretty close to "just right"
16:58 ShadowNinja hmmmm: What about that?  I know that.
16:58 hmmmm       [11:51 AM] <ShadowNinja> NULL == 0L == false == 0 And 0 is the only value that fails.
16:58 ShadowNinja Yes, and?
16:58 sapier      well if you set it to max_clients * 4 you may increase lag more then required for e.g. 2 player games
16:59 hmmmm       int foo = (NULL == 0); printf("%d\n", foo); on ANY ISO 1989:9989 compliant compiler, and you're going to get 1
16:59 ShadowNinja That's a max though, right?  Setting it very high shouldn't cause any lag.
16:59 ShadowNinja hmmmm: Yes, of course.
17:00 hmmmm       so then maybe I misunderstood what you were trying to say in that sentence
17:00 ShadowNinja Yes, I think so.  :-)
17:00 sapier      it will cause lag as this is the number maximum blocks processed in server thread. a large number may cause quite a lot of blocks initially queued
17:00 ShadowNinja I meant (NULL == 0L == false == 0) == true.
17:01 sapier      well it's a design issue ... I really don't know why actually cyclic tasks are called async in minetest engine
17:02 sapier      what do you think about splitting server thread in a cyclic and async thread? cyclic is doing all that stuff required to be done regularly while async does user interaction
17:03 hmmmm       user interaction, like handling packets?
17:03 sapier      yes
17:04 sapier      basicaly
17:04 hmmmm       yes that would be great
17:04 hmmmm       back when celeron was suggesting that minetest be rewritten this is what really needs to happen
17:04 sapier      we'd get best result if we added thread priorities along with this change
17:04 hmmmm       the entire queue-it-up-and-process-it-in-one-huge-loop thing isn't very efficient with resources
17:04 sapier      well that's not a rewrite it's a minor change
17:04 hmmmm       well
17:05 hmmmm       it's the beginning of a huge change
17:05 sapier      I already started doing this some time ago ;-)
17:05 hmmmm       there are thread priorities
17:05 sapier      where?
17:05 hmmmm       except they don't work on Linux
17:05 hmmmm       porting.cpp
17:06 hmmmm       thread priorities only mean anything if it's using SCHED_RR or SCHED_FIFO
17:06 hmmmm       and to change the policy you need to have root permissions
17:06 sapier      hmm I remember dark I may have even been the one adding that code long long ago
17:07 sapier      but I'd move it to jthread as that code is now ours
17:07 hmmmm       yeah
17:07 hmmmm       i remember you did try adding thread priorities but that was a pull request that never made it
17:07 hmmmm       and here i actually did add it along with thread affinities for emergethread but i realized what the problem was
17:07 sapier      could be true too
17:09 sapier      but if we have separate threads for cyclic as well as acylic things we can only gain full benefit by doing cyclic things low prio. It's useless to generate map at full speed if network thread is starved and can't send it to clients
17:09 sapier      just to name one example
17:10 hmmmm       i think linux takes this into account actually
17:10 hmmmm       if something is cpu bound it lowers its priority slightly
17:10 hmmmm       shrug
17:12 sapier      I think so too but it's quite limited
17:16 hmmmm       it'd be pretty annoying to have the user enter their root pw every time they run minetest
17:16 hmmmm       and weird
17:17 hmmmm       it wouldn't really make a difference either unless all cores are pegged at 100% cpu usage, and even then, the thing that it really makes a difference with is latency
17:17 sapier      yes ... the nice value is honored for SCHED_OTHER ... is nice usable thread specific?
17:17 hmmmm       i'd hope so
17:18 sapier      running minetest at root is not an option so we can forget about RR and FIFO
17:19 hmmmm       well
17:19 hmmmm       if you can figure something out, by all means...
17:20 sapier      :-( nice seems to be process specific
17:31 sapier      https://github.com/minetest/minetest/pull/1131 ShadowNinja fine this way?
17:33 sapier      https://github.com/minetest/minetest/pull/1129 is ok now
17:33 sapier      at least as of my point of view
17:37 sapier      https://github.com/minetest/minetest/issues/1132 comments?
20:14 emptty      Hello there
20:15 emptty      I have a bug reported against the debian package stating that the game segfaults when browsing the public serverlist
20:15 emptty      does it sounds like something known to you guys?
20:15 sapier      yes
20:15 sapier      well maybe
20:15 sapier      is luajit involved?
20:16 emptty      luajit is compiled in, yes
20:16 sapier      what version?
20:16 VanessaE    debian doesn't have a luajit package I thought?
20:16 emptty      there is a backtrace and a strace in the bug report, if it helps: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=733974#17
20:16 emptty      VanessaE: yes we do, but not on all architectures
20:16 sapier      2.0.2 hmm that's not the bug I expected
20:16 emptty      some of our rare architectures don't have luajit, but nobody has a luajit on powerpsce or so
20:17 sapier      we know about a bul in bet version of luajit
20:18 VanessaE    however, sapier.... look at event #7 of the backtrace... :)
20:18 ShadowBot   https://github.com/minetest/minetest/issues/7 -- Fix build with Pathscales EKOPath by Merkil
20:18 VanessaE    shaddup ShadowBot :P
20:18 sapier      looks like the bug xyz mentioned for beta4
20:18 sapier      beta7
20:21 sapier      well that's great everything is attached except the relevant information
20:22 emptty      what would you need?
20:23 sapier      well the script error has a error message attached I can't find it in this traces
20:24 VanessaE    sapier: it wouldn't have.
20:24 VanessaE    it threw a C++ exception
20:24 VanessaE    it got bit by the "errors don't get logged" bug
20:25 sapier      whoo
20:25 VanessaE    emptty: any chance of a console capture of the crash?
20:25 sapier      it's not supposed to crash there
20:25 emptty      you have a console capture for the same issue above on the same page
20:25 VanessaE    oh yeah, so i do
20:25 VanessaE    derp
20:25 emptty      but the first capture was produced with version 0.4.8
20:26 emptty      that's why I pointed you guys directly on the bottom of the page, with recent info
20:26 VanessaE    yeah, I'm just blind, don't mind me :P
20:27 sapier      well that issue seems to be what xyz mentioned but for some reason he didn't get that error on 2.0.2 but on 2.0.0beta7 only ... at least that's what I understood
20:28 VanessaE    well wait a sec here
20:28 VanessaE    wtf am I using then
20:28 VanessaE    *looks*
20:29 VanessaE    my copy of luajit is at commit 6964a7752ae314dcae693abcb0c1175c95ad22e0 ...
20:30 VanessaE    minetest is within one or two commits of HEAD
20:30 sapier      it's only relevant on client
20:31 VanessaE    oh right,  I tend to forget about the client-side Lua in the menu.
20:32 VanessaE    lesse.  client side is within one or two of minetest HEAD also, luajit there is 2.0.0 as included in Xubuntu's repo
20:32 VanessaE    (12.04)
20:32 sapier      and it does work?
20:32 VanessaE    yup
20:33 VanessaE    no problems here.
20:33 sapier      well not sure if this is a good or bad sign
20:33 VanessaE    wait
20:33 VanessaE    specifically,
20:33 VanessaE    2.0.0 beta 9
20:35 sapier      emptty did I understand correct you are able to reproduce this error?
20:36 emptty      let me try
20:36 emptty      yes :)
20:36 sapier      that's good
20:36 hmmmm       watch the error be caused by some guy missing the mainmenu .lua files
20:36 hmmmm       :\
20:36 emptty      If I tick the "public list" box, it lasts maybe one second and crashes
20:36 VanessaE    hmmmm: wouldn't be the first time that's been the case :P
20:36 hmmmm       oh
20:37 hmmmm       serverlist, hahaha, that thing :p
20:37 sapier      no this seems like some lua error occuring in async thread
20:37 emptty      yeah, that's not as soon as I click
20:37 emptty      that's a delayed death :)
20:37 sapier      xyz reported a bug about serialization not working correct on luajit 2.0.0beta7
20:37 hmmmm       shouldn't errors in the async thread be handled and not fatal?
20:38 sapier      actually yes but ShadowNinja made it a fatal error on catching luajit errors
20:38 hmmmm       i'd call that a regression
20:38 sapier      still there shouldn't happen an error in this code
20:38 hmmmm       sure, but there are two problems at least we know about now
20:39 emptty      let's call it *two* bugs, then ;)
20:39 emptty      I can run whatever tests you want from me, but I'm clueless to investigate the issue myself
20:40 sapier      no problem I know where to start searching async_env.lua
20:41 emptty      highlight me when you need me, then.
20:41 emptty      many thanks for digging into that issue
20:41 sapier      https://gist.github.com/sapier/8774597
20:41 sapier      can you add those two lines right at beginning of engine.job_processor function?
20:43 emptty      FCT: "6\2{\0\0\0\27LJ\1\0/@/usr/share/games/minetest/builtin/mainmenu.luaD\0\1\3\0\3\0\4\13�\1\0024\1\0\0007\1\1\1%\2\2\0@\1\2\0\11online\18get_favorites\11engine\1\1\1\1param\0\0\5\0\0\0\0\0\0"
20:43 emptty      DATA: "0"
20:43 emptty      (but it didn't crash this time oO )
20:44 hmmmm       what happens when you comment out the first line of those two?
20:44 emptty      unticked the box, reticked it, got exactly the same output, and a crash
20:44 sapier      try again :-) this looks fine
20:44 emptty      the FCT ?
20:44 hmmmm       yea
20:44 hmmmm       try it a few times
20:45 hmmmm       does it consistently crash?
20:45 emptty      yeah
20:45 emptty      3 crashes out of 4 tests
20:45 emptty      w/o FCT, I get DATA "0" and crash
20:45 hmmmm       hmm
20:46 hmmmm       try adding the following:
20:46 hmmmm       local t = clock()
20:46 hmmmm       err
20:46 hmmmm       local t = os.clock()
20:46 emptty      I'm by 9 crash out of 10 or something
20:46 hmmmm       while os.clock() - t <= 5 do end
20:46 hmmmm       and then see if it crashes then
20:47 emptty      btw, I put the print before the local lines, right?
20:47 hmmmm       it doesn't really matter
20:47 hmmmm       local line, i meant the latter of the two
20:47 emptty      should I put the waiting loop before the decode lines or after?
20:47 sapier1     put it before so we can be sure we read it if there's a unmarshall error
20:48 emptty      crash after 5sec
20:48 hmmmm       ah okay
20:48 sapier      what did you do now?
20:49 hmmmm       lol
20:49 sapier      can you gist the code?
20:49 hmmmm       I tried to diagnose if it was a race condition
20:49 hmmmm       without even looking at the code in question
20:49 emptty      (3 times on 3 tests)
20:49 sapier      hmmmm race condition ... ok I gues that's an issue we can solve
20:50 hmmmm       no it's not
20:50 hmmmm       see he just said it crashes even after delaying 5 seconds
20:50 emptty      State of my code: https://gist.github.com/mquinson/8774730
20:50 sapier      and it doesn't crash if you print the fct?
20:51 hmmmm       it does crash, but just not all of the time
20:51 hmmmm       you realize you're not going to get anywhere without debugging right?
20:51 sapier      ok good but we need to know if the serialized_fct is broken in this "crash case"
20:51 emptty      Nah, it crashes with the print FCT uncommented too
20:52 sapier      that's the information I need
20:52 emptty      Ran 3 tests with the print (but w/o the delay), and all of them crashed
20:52 sapier      yes hmmmm but the more information I have the better I can reproduce
20:54 sapier      emptty? whats printed in a crash situation?
20:55 emptty      sapier: http://paste.debian.net/79842/
20:56 sapier      ok seems like serialized data isn't messed up
20:56 sapier      bug has to be hidden somewhere in (un)marshalling code
21:48 hmmmm       what do you guys say about removing the math mapgen?
21:48 VanessaE    proller will kick your ass :P
21:48 blaise      lmao
21:48 hmmmm       he's not around
21:48 blaise      people tend to time out on my server... and I can't figure out which mod is doing it..
21:48 hmmmm       and he's not even associated with minetest anymore
21:48 blaise      I actually have no issues connecting..
21:49 blaise      do you guys think it's due to impatience?
21:49 VanessaE    frankly, I have no use for it, and in the end, the mapgen is math ANYWAY
21:49 VanessaE    (what is perlin noise if it isn't a math formula?)
21:50 VanessaE    blaise: the biggest issue with timeouts now is tablet users versus servers with more mods than tablets have memory to handle the media for, or so I woudl have to guess.
21:50 VanessaE    would*
21:51 emptty      sapier: I go to bed now. If you need to contact me, I'm very very often connected to irc.debian.org (I don't even know which irc network we are on here)
21:53 sapier      ok thanks emptty I havent found it by now
21:53 hmmmm       oh a debian guy
21:54 blaise      hrmm
21:54 hmmmm       debian ships with PermitRootLogin yes
21:54 hmmmm       lol
21:55 VanessaE    blaise: and a lot of THAT can be blamed on those clients being, as far as I can tell, unable to disable the "preload item visuals" feature.  So they try to preload hundreds, if not thousands of textures and boom, crash + timeout.
21:55 sapier      well for servers root login is at least comfortable ... but not very safe :-)
21:57 blaise      anyone care to log in to my server and take a guess at why people time out?
21:59 blaise      I tried to shut off all the mob mod's to see if it was better...
22:02 blaise      oh wow, I'm in -dev
22:02 blaise      sorry..
23:23 hmmmm       so like
23:23 hmmmm       who ever said that assert() from assert.h is always a macro
23:24 hmmmm       nevermind
23:24 hmmmm       was going to say something about debug.h
23:33 hmmmm       ugh
23:34 hmmmm       i hate how Map needs a pointer to emergemanager but it needs to stay in there until Map only does Map stuff and absolutely nothing else
23:35 hmmmm       come to think of it, ServerMap has a gamedef too... this is not really good design
23:38 sapier      well hmmmm there are a lot of bad design decisions as of data separation. If there's a chance to hide data from other subsystems I suggest doing it
23:39 hmmmm       sapier, what i'm doing now is an improvement, but only an improvement...
23:39 sapier      yes but a lot of small improvements will still still help to reach our goal
23:39 hmmmm       but at least I'll be able to die knowing that MapgenParams isn't a trainwreck
23:40 sapier      e.g. my connection fixes hide "peer" from everyone except connection classes I'd like to hide the RemoteClients too but by now I didn't get that far
23:41 hmmmm       who made environment take an emergemanager
23:41 sapier      :-) why does server need environment to know a player ;-)
23:42 hmmmm       well my question is a real question because I didn't make it like that
23:42 hmmmm       i know i didn't because the member is named m_emerger, i would've made it m_emerge
23:42 sapier      there's a lot of cleanup missing :-) ... or why does connection need emerge mangager to cleanup peer ;-)
23:43 hmmmm       in any case it's being used in exactly one spot
23:43 sapier      well if it's only one location you could try to get rid of it ... or add a sane interface if something really needs to be done
23:45 hmmmm       oh nice, and that bit of code is a bug anyway
23:45 hmmmm       it passes a NULL MapBlock * to activateBlock
23:45 sapier      a reason more to clean it
23:46 hmmmm       i think people ought to reexamine their assumptions and take a look at functions they're calling because it might not handle NULL appropriately
23:47 sapier      well I usually check any parameter I get from some other class and only consider class internal pointers which are initialized by constructor and never changed as not null :)
23:48 sapier      everything else is at least worth and assert(param != null)
23:48 sapier      well ... that's how I try to do it :-) sometimes I miss it
23:48 hmmmm       that's a good scheme
23:49 sapier      I have to get rid of that damn "well" again recently I use it permanently
23:51 hmmmm       heh.  for a while i had this verbal tick where i'd end my sentences with "...and what not" whenever i'd get nervous
23:51 hmmmm       i didn't realize how stupid i sounded until afterward
23:57 sapier      I can't reproduce that damn error emptty found
23:58 sapier      I forgot to ask him what machine he did use ... seems my good old phenon II is getting a little bit slow recently