Time  Nick        Message
00:00 kahrl       ok, so I tried to construct an example where the bad check would make the algorithm fail, and I can't because it seems PathStartsWith works regardless of what this check returns
00:00 kahrl       but it should be fixed anyway
00:04 kahrl       another thing that valgrind (still) reports is the uninitialized sin6_flowinfo field in socket.cpp, though fixing it would break pull #862
00:29 VanessaE    regarding the timeouts/unable-to-allocate issue, https://forum.minetest.net/viewtopic.php?pid=104309#p104309
01:16 kahrl       aside from bind/sendto (left out intentionally because of #862), this assortment of fixes removes a lot of valgrind warnings for me: https://gist.github.com/kahrl/6170401
01:53 kahrl       btw, the console hasn't been working for me for the last few days -- the font can't be loaded, using a non-freetype build
01:53 kahrl       "Unable to load font: /home/kahrl/minetest/bin/../textures/fontdejavusansmono.png"
01:54 kahrl       PilzAdam: is it intentional that defaultsettings.cpp looks for textures/fontdejavusansmono.png instead of fonts/fontdejavusansmono.png?
01:55 VanessaE    I don't think that was intentional, as we're told now to put those in fonts/
01:55 VanessaE    but what do I know :P
01:55 kahrl       I'll wait for him to come back
01:55 VanessaE    you'll be waiting a while :)
01:56 kahrl       I have years of youtube videos left to watch :P
01:56 VanessaE    heh
01:56 VanessaE    better:  tvtropes.org
01:56 kahrl       oh god
01:56 kahrl       I successfully managed to close that site a while ago
01:56 VanessaE    LOL
05:57 celeron55   kahrl: oh pilzadam didn't fix the path there? i noticed it and mentioned it here though
05:57 celeron55   you can fix it
06:15 celeron55   VanessaE: if it's not that reproducible on a test server, then you should try the changes
06:21 VanessaE    mmh
06:21 VanessaE    I guess if that's how it has to go
06:24 VanessaE    ok, the first change (comment-out the three-line DenyAccess bit)
06:24 VanessaE    is in place and server is running with it.
06:25 VanessaE    I didn't figure so much that a test server couldn't reproduce it, only that a server on the same box probably won't, since that's the same situation I have - e.g. it never fails for *me*, but for remote clients.
06:25 VanessaE    anyway, it's up with the first change, try to sign in
06:57 celeron55   oh at work? well i guess i can
06:58 VanessaE    didn't realize you were at work (duh :) )
07:04 celeron55   it appears to work roughly like i'd have expected with that change; it just waits at "Connecting to server..." until the player is free
07:06 celeron55   a full log of what i did would be useful, maybe
07:07 VanessaE    hm, log's too big to load, I'll have to delete it and have you try again
07:07 celeron55   just tail -n 100 it
07:07 VanessaE    oh.  duh :D
07:07 celeron55   or 1000, whatever
07:08 VanessaE    now I feel utterly stupid :P
07:08 VanessaE    http://pastebin.ubuntu.com/5957924/
07:08 celeron55   on unix-like operating system one never has too large files 8)
07:09 celeron55   there isn't a single mention of me in there
07:09 VanessaE    lemme expand the search
07:10 VanessaE    *watches pastebin choke on a huge paste* :P
07:10 VanessaE    http://pastebin.ubuntu.com/5957928/
07:10 VanessaE    last 100,000 lines :-)
07:15 celeron55   well it appears to me that it just doesn't receive the disconnect packet from the disconnecting client properly
07:15 celeron55   and then waits for the timeout
07:15 VanessaE    for some users, that timeout never seems to happen, even after waiting several hours (well, so they claim)
07:16 celeron55   that's the odd thing
07:17 celeron55   keep this change in for a while and ask on the forum for anyone who still has the problem to tell
07:17 celeron55   because the server doesn't really do anything differently currently
07:18 celeron55   unless there is some very weird interaction caused by the client getting told about the failure (that i added in the DoS patch)
07:19 VanessaE    someone mentioned earlier that the DoS patch might be the cause
07:19 celeron55   of course it is
07:19 celeron55   but the question is, is this a bug or is this just people freaking out of changedbehavior
07:19 celeron55   +space
07:19 VanessaE    definitely a bug - I've got seasoned users having the same problem, and it happens on multiple servers
07:19 celeron55   i haven't yet seen a bug
07:20 VanessaE    so far only twoelk, bulldog1 and I have managed to avoid it on mine, and I'm not sure how those two managed
07:20 celeron55   anyway, let's collect this data now
07:21 VanessaE    ook
07:21 VanessaE    ok
07:21 * sfan5     is about to commit a small XSS fix for the masterserver list.js
07:21 celeron55   if the result of this is that it works as before, then that is a useful result
07:21 celeron55   but one thing: add a log message to the other thing that i suggested to comment out
07:21 VanessaE    and this is just with that one DenyAccess call commented out
07:22 celeron55   it would be useful to see in logs whether that code gets ever called
07:22 celeron55   (instead of commenting it out)
07:22 VanessaE    sure
07:27 VanessaE    ok, with only that first DenyAccess() call commented out, and an extra log call (I stuck it in infostream) inside the if() block at line 4653.
07:27 VanessaE    AND I nuked the old log.  server's up.
07:29 celeron55   then just wait for actual users 8)
07:29 VanessaE    heh
08:31 Taoki       Was looking at the new feature for seasons. Is this configurable via Lua? Like, do you specify temperatures and features for each time of year there, or is functionality all hard-coded?
08:42 Taoki       Ugh. Not a month passes without some change in GIT breaking mods and requiring changes to them or moving them to a new folder
08:43 Taoki       Fixed all of them though, server working again
09:44 Exio4       Taoki: weather? ask proller
09:44 Exio4       it is overly WIP and he wanted to merged it just for the lulz, even if it isn't finished and needs tweaks, etc
11:36 PilzAdam    <kahrl> PilzAdam: is it intentional that defaultsettings.cpp looks for textures/fontdejavusansmono.png instead of fonts/fontdejavusansmono.png? <- oops, no
11:39 PilzAdam    Ill fix that now
13:59 proller_    update jsoncpp - https://github.com/proller/minetest/commit/ea346f43629c90301267176f2e0b1cde6231d18f
14:00 proller_    trivial
14:00 PilzAdam    whats the benefit from this?
14:00 proller_    fixes some rare bugs
14:01 PilzAdam    did any of these occur in Minetest?
14:02 proller_    maybe no
14:57 celeron55   at least you should add to the commit to which version it updates to
14:57 celeron55   there's no way to know otherwise
14:58 Jordach     celeron55, i just dealt with the crevis team and cleared things up for you
15:00 proller_    celeron55, its latests version from svn
15:01 celeron55   proller_: i don't think we should update to those but rather to actual releases at most
15:02 celeron55   that way there are less meaningless commits and less chance of unstable library code
15:08 celeron55   that commit is just useless; you probably see it yourself too
15:09 celeron55   hmm, well, one line does something: the addition of allocated_ = false; to the else branch
15:09 celeron55   everything else does exactly nothing
15:11 celeron55   throw it in but don't waste your or anyone else's time on these library SVN updates in the future; it doesn't benefit anyone
15:28 proller_    commit or no ?
15:34 Exio4       ...
16:25 kahrl       should I push https://gist.github.com/kahrl/6170401?
17:39 PilzAdam    kahrl, seems good
20:33 Tesseract   kahrl: I fixed #862.
20:42 kahrl       Tesseract: let me see
20:43 Tesseract   Hmmm, that might catch 0.0.0.0 for singleplayer if you use IPv6...
20:43 kahrl       that works, but it might not be the best way of doing it
20:44 kahrl       yeah and it would break a server that uses ipv6_server = true unless it also sets bind_address
20:45 kahrl       perhaps delay the socket() call and decision about m_addr_family until Bind()?
20:45 kahrl       but what to do with client sockets
20:47 Tesseract   kahrl: The check IS done in UDPSocket::Bind().
20:48 kahrl       My original "vision" for this was to allow setting a list of addresses to bind to. Then it would create a separate socket for each bind address, using an appropriate family. When binding to 0.0.0.0 the server automatically creates new sockets for each actual local address that someone connects to
20:49 kahrl       The latter is to fix the case where the server has multiple interface (with IPs A and B), minetest bidns to 0.0.0.0, client connects to IP address A, but server sets the source address of its packets to the client to B
20:49 kahrl       interfaces*
20:49 kahrl       binds*
20:49 kahrl       ... causing a timeout on the client side
20:51 kahrl       If you want a purely ipv4 server, set bind_address = 0.0.0.0, if you want a purely ipv6 server set it to ::, if you want both set it to 0.0.0.0, ::
20:51 kahrl       anyway, that was my vision and it's quite complicated so it never got anywhere
20:51 Tesseract   Is the address family used before bind()?
20:52 proller     kahrl, nono. :: for both must work
20:52 Tesseract   proller: You mean :: == 0.0.0.0?
20:52 proller     0.0.0.0 for v4, :: for other
20:53 proller     Tesseract, no, not == , but when listen :: with v6 - it can recieve v4 connections
20:53 Tesseract   proller: That is how it is now, but the ipv6_server setting is also used...
20:53 kahrl       proller: why must it?
20:54 Tesseract   proller: That is what I meant, it doesn't work like that. (AFAIK)
20:54 proller     kahrl, by v6 design
20:54 proller     it creates v46 socket
20:54 proller     possible to make pure v6, but no sense now
20:54 kahrl       afaik socket implementation vary quite wildly in this area
20:54 kahrl       +s
20:55 * Tesseract sees this spiraling out of his skill zone. ;-)
20:56 Tesseract   Anyway, singleplayer with a IPv6 server should work now. proller: Do you have IPv6?
20:56 proller     yes
20:57 proller     v6 only - its my server
20:57 Tesseract   proller: Can you test my pull with it?
20:57 proller     not now
21:00 kahrl       I know a bit about pure ipv6 including the router solicitation stuff etc. But then I wanted to know how it interacted with ipv4. Saw that there was seemingly at least a dozen competing ways, stopped caring.
21:02 proller     most common way - dual 46 socket
21:02 proller     it now works in minetest
21:02 proller     you listen :: in v6 and it can revieve v4 connections and show it as :ffff:1.2.3.4
21:03 kahrl       proller, yeah, it's good that it works
21:03 kahrl       however I have reason to believe it doesn't work in windows
21:03 kahrl       (I assume your server is running on some unixlike OS)
21:04 Tesseract   man 7 ipv6 mentions something like that. The addresses are mapped to IPv6.
21:04 kahrl       http://msdn.microsoft.com/en-us/library/windows/desktop/bb513665%28v=vs.85%29.aspx <-- says you need to set some socket options to get dual stack sockets
21:04 kahrl       and that only works in vista or later
21:05 Tesseract   Oh, right, windows compatibility. :-(
21:05 proller     unixlike have sysctl for it, and usualy 46 enabled by default
21:07 Tesseract   Looks like the only way is to use dual sockets, unless we drop XP support.
21:07 kahrl       "an application has to create two sockets, one socket for use with IPv4 and one socket for use with IPv6"
21:07 kahrl       ^ works on XP / Server 2003 and is kind of what I described above in my "vision"
21:08 kahrl       but I agree, it's probably not worth to make such a complicated system
21:08 kahrl       worth it*
22:09 kahrl       what do you guys think of the download API for lua in #490?
22:10 kahrl       it got kind of lost in the list of pull requests, but now I could add it easily to the httpfetch branch
22:10 kahrl       (rebasing the scriptapi part and using httpfetch for the core implementation)
22:12 kahrl       the practical difference between that and Jeija's implementation is that httpfetch has a limit on parallel fetches (media_fetch_threads), and uses only at most 1 thread
22:43 PilzAdam    kahrl, there should be a setting that disables this function by default
23:10 kahrl       why? I mean, mods can already do os.execute("wget ...") to upload information to some server
23:10 kahrl       PilzAdam: ^
23:10 PilzAdam    not on windows
23:10 Exio4       i guess a way for getting if "curl" is available should be there
23:11 kahrl       isn't there some wget equivalent on windows?
23:11 Exio4       you can even use things like the irc mod
23:11 Exio4       for running C code
23:12 kahrl       you could write some .vbs file and execute it
23:16 kahrl       or even io.popen("ftp")
23:19 kahrl       so my point is that it's not really worth it to restrict access to a single unsafe function where there are many more out there
23:19 kahrl       restricting access to all unsafe functions might be worth considering (I bet sapier has something to say about this)
23:20 Exio4       in a dream world i would like to see a "permissions-like" system for mods
23:21 Exio4       where they could ask for "System access" and warn the user, and if it doesn't accept; it couldn't use os.xyz
23:21 Exio4       and so on
23:21 Exio4       :P
23:21 Exio4       but that would require one "lua stack" / mod, no?
23:22 kahrl       yeah something like that ^ and what if mods call functions from each other?
23:22 Exio4       good question too
23:25 kahrl       perhaps something like minetest.get_library("os") which would return the full os library which is invisible otherwise; only usable during the registration phase (when the modname is known)
23:25 kahrl       then the mod would do local os = minetest.get_library("os")
23:26 Exio4       it could work too, but then you would need to protect it somehow
23:26 Exio4       and, how does serialize works?
23:26 kahrl       it's the mod responsibility to protect it (e.g. by using local)
23:27 kahrl       what about serialize?
23:27 Exio4       isn't there some usage of os. in builtin for files?
23:28 kahrl       builtin can have full access to everything of course
23:28 kahrl       they would be made invisible at the end of builtin
23:29 Exio4       does lua use some thing like closures? or would be probably to "avoid" all of this by replacing functions that will run under builtin functions, (like the one for checking if the pos is protected, or so)
23:30 Exio4       oh man, my english sucks
23:30 kahrl       it has closures
23:30 Exio4       k
23:30 Exio4       nevermind then
23:31 kahrl       maybe not in the completely same sense as in lisp/haskell/whatever, but the same concept
23:31 Exio4       yes, that is what i meant "some thing like"
23:31 Exio4       something*
23:40 kahrl       I see that #490 returns the cURL error code to Lua. Would that be needed or would succeeded = true/false, timeout = true/false be enough?
23:41 kahrl       I ask because the httpfetch module is designed in a way that cURL could (theoretically) replaced by another library
23:41 Exio4       i guess
23:41 kahrl       so it doesn't expose cURL types or error codes or whatnot
23:45 kahrl       wish Jeija was around so I could ask about his thoughts
23:46 PilzAdam    he made this with it: http://youtu.be/jo_GT6a4M90
23:49 kahrl       ah, I saw that back in the day when I came back
23:49 kahrl       is there source code for it?
23:49 kahrl       ah, it's in the video ;)
23:50 kahrl       his callbacks just ignore response and error code
23:53 Exio4       wow, didn't know about it
23:53 PilzAdam    he won the mesecon machine competition with that