#bitfighter IRC Log

Index Search ←Prev date Next date→

IRC Log for 2014-06-27

Timestamps are in GMT/BST.

00:06:45no_mu has joined
00:07:58Nothing_Much Quit (Read error: Connection reset by peer)
00:19:09Watusimoto has joined
00:23:22Watusimotohola amigos
00:23:28raptorhello
00:23:32Watusimotoback in luxembourg
00:23:39WatusimotoI am exhausted
00:23:44raptorwelcome back!
00:23:50Watusimotoraptor: I just forwarded you an email you will find interesting
00:23:56raptorchecking..
00:24:13WatusimotoRE our MIA VPS
00:24:58raptoroh my goodness...
00:25:04raptorthe IPMI was hacked
00:25:12raptorthat's nuts
00:25:23raptorthat's basically equivalent to the server BIOS
00:25:29Watusimotointeresting
00:25:53WatusimotoI'm going to redirect bitfighter.net back to the original vps so we can keep track of restoratin progress
00:25:58raptorok
00:26:42raptorthat's actually good news - URPad was forthcoming and informative and responsible
00:26:45WatusimotoI prepaid a year on the second vps -- what I'd like to do, I think, is keep it in reserve, and use it as an active mirror site so if this happens again, we have both a backup and a fallback
00:27:01Watusimotoit's not a very powerful machine, but it was only something liek $10
00:27:01sam686www.bitfighter.org still points to old server (only bitfighter.org was redirected to another temporary server)
00:27:15Watusimotosam686: welcome!
00:27:28raptorthat's a good idea
00:27:29Watusimotoyes, you are right -- I deliberately did that, and promptly forgot
00:27:40Watusimotowhich idea? the fallback machine?
00:27:50raptori can set up an rsync script - do you want to add a semi-permanent DNS like backup.bitfighter.org to it?
00:27:58Watusimotowe could do that
00:28:18Watusimotoor we could have bitfighter.net point to the backup machine
00:28:28raptorrsync to the fallback machine... if they are basically the same OS/arch, then I can even sync config files
00:28:30Watusimotowe'll need to look at how much disk we have vs. need
00:29:01WatusimotoI think I paid a little more to get 25GB, but I looked at a lot of servers so I may be remembering wrong
00:29:12Watusimotoand I had had several tasty belgian ales, so that also muddled things
00:29:16raptorthis new one has 128MB RAM and 10GB disk
00:29:23Watusimotoah, ok
00:29:23raptorheh
00:29:49Watusimotoyeah, the first one I looked at had the 25GB option, but they were "sold out" of the super low end machine
00:30:04sam686the RAM limit is probably a real problem, I might go with at least 512 mb for 32 bit
00:30:23Watusimotofor running the master, 128MB is plenty
00:30:36raptoryeah, think of IPMI has an uber-BIOS - it can control loads of things including multiple blades in a rack
00:30:52WatusimotoI should read up on it, just as a point of curiosity
00:30:54raptorbut it's usually only accessible via serial port
00:31:06raptoror other direct-hardware access
00:31:19raptorso the fact that it was hacked raises some interesting questions...
00:31:23sam686although we could reduce bitfighter.org memory usage by moving some parts to a separate server, like logbot
00:31:30Watusimotoso the second machine will be there to keep the lights on in case the first goes away for some reason, and also to provide a seed to build a new server if we need it
00:31:31raptormaybe a sysadmin was trying to play rogue on it
00:31:57WatusimotoI think in the case like this, logbot can just go on holliday
00:32:29Watusimotosomeone might have been targeting a particular service running on the vps
00:32:47Watusimotoor the sys admin might have screwed up, and this is their cover
00:33:01raptoryeah, but the direct-hardware access requirement is what makes it odd
00:33:01Watusimotoor someone might have been seeking a good attack platform
00:33:34raptorunless the IPMI vulnerability was that there was some entry way from a NIC
00:33:40Watusimotomaybe when things have calmed down, I can ask for more info about the attack
00:34:26Watusimotowhat if they left a direct access cable plugged in, and it was connected to a machine on the internet
00:34:35raptorthat could happen too
00:35:31Watusimotohacking has come a long way since the days when I tried my hand at wardialing
00:35:51sam686whatever IPMI is, its likely won't affect my home server (no IPMI there)
00:35:58Watusimotonor here!
00:36:23raptorhahaha wardialing
00:37:00Watusimotothis from my latest ticket complaining about a lack of info
00:37:01WatusimotoWe are working to restore the effected VPS' and hope to have them all restored within the next 48-72 hrs, but that is only an estimate.
00:37:48raptorok
00:38:23Watusimotomy big releif is that we won't be losing our database
00:38:30fordcarsYay!
00:38:41Watusimotoif the backup actually works
00:58:12sam686I need to go, might be back in 3 hours
00:58:35sam686 Quit (Quit: Page closed)
01:06:26no_mu is now known as Nothing_Much
01:12:25Nothing_Much Quit (Quit: Konversation terminated!)
01:42:23Nothing_Much has joined
02:06:11WatusimotoLatest news from urpad
02:06:12Watusimotohttp://pastie.org/9328186
02:06:52raptorhmmm
02:07:07raptorlet's hope we're not among the 'few failed backups'
02:33:04raptor Quit ()
02:44:36sam686 has joined
02:44:36ChanServ sets mode +v
03:13:05Watusimoto Quit (Ping timeout: 252 seconds)
07:36:52Nothing_Much Quit (Quit: Konversation terminated!)
07:37:03sam686I guess more updates on urpad problems is here http://urpad.net/blog/
08:19:42Nothing_Much has joined
08:30:22Flynnn has joined
10:28:31fordcars Quit (Quit: Page closed)
10:34:35Flynnn Quit (Quit: Leaving)
12:00:13watusimoto has joined
12:00:13ChanServ sets mode +o
12:08:53LordDVG has joined
13:21:29ChumpBot Quit (Ping timeout: 248 seconds)
15:01:29Watusimoto_ has joined
15:46:33Nothing_Much Quit (Quit: Konversation terminated!)
15:50:32Watusimoto_ Quit (Ping timeout: 260 seconds)
17:42:03Watusimoto_ has joined
18:02:57Nothing_Much has joined
18:17:41YoshiSmb_m has joined
18:18:06YoshiSmb_mHi Everybody!
18:20:45LordDVG Quit (Quit: Leaving)
19:14:32raptor has joined
19:14:33ChanServ sets mode +o
19:16:10raptorgood morning!
19:47:15Watusimoto_ Quit (Ping timeout: 240 seconds)
19:59:37Watusimoto_ has joined
20:00:39YoshiSmb_mMorning Raptor! (what a bad night with the childs)
20:00:55raptorgood morning
20:01:51raptorthat means you need a nap during the day (me too..)
20:01:52raptor:)
20:04:11YoshiSmb_myea...
20:07:42Watusimoto_ Quit (Ping timeout: 245 seconds)
20:28:37YoshiSmb_mhttps://www.dropbox.com/s/0595hilyzw8lj0z/menu.ogg
20:29:20YoshiSmb_mcan you test it out raptor?
20:29:52raptorspooky
20:30:03YoshiSmb_m?
20:30:38raptorit sounds scary
20:31:27YoshiSmb_moh, it's doesnt sound scary for me
20:33:57YoshiSmb_mit's sounds...
20:34:04YoshiSmb_mSad...
20:36:59YoshiSmb_mhow scary it's is?
20:37:06YoshiSmb_mfor you?
20:37:17raptornot very scary
20:38:10YoshiSmb_mat least it's does fit on bitfighter menu... (i think)
20:45:25Watusimoto_ has joined
20:47:47watusimotohi
20:48:06raptorhi
20:49:06watusimotostill no news on bf.org
20:49:26raptorI finally found the real reason behind our client lib and linking woes
20:49:51raptoryeah, i keep pinging www.bitfighter.org, just in case
20:53:36raptordo you understand what PIC is? (pointer-independent code)
20:54:02raptoroops.. *position-independent
20:54:40raptorsee here: http://stackoverflow.com/a/5311538
20:55:30raptorbasically what happens is this:
20:55:54raptor1. code is by default compiled in a position-dependent way (it's faster)
20:56:28raptor2. an archive is linked together with all the position-dependent objects together
20:56:34raptor* a static archive
20:57:24YoshiSmb_m Quit (Ping timeout: 264 seconds)
20:57:46raptor3. when a static archive is linked into the final binary, only those pieces that fulfill the method calls needed in the binary are linked in, not the whole thing
20:58:20raptor4. since pieces of the static archive may be missing in the final binary, the position-dependent code gets misaligned
20:58:34raptor5. weird errors can ensue when running the resulting binary
20:59:19raptor... There are 3 solutions to this problem:
21:00:07raptor1. compile with -fPIC, or as position-independent code (which Windows doesn't support). This also adds some overhead to every method call
21:00:55raptor2. guarantee the full static library is linked in in the final binary linking. This is what kaen did to our build system that adds some complexity with extra linker flags
21:02:20raptor3. don't use static libraries as an interim collection to link in the final binary; instead, link all original object files together. This is what we did before attempting reduce having to recompile code all the time for the tests
21:03:05watusimotook, I got all that. I have some questions, but I'll save them for later. I got to run!
21:03:10watusimotoI'll be on later!
21:03:10raptorok rats
21:03:13raptorlater!
21:03:24watusimotodid you want to finish your thought?
21:03:34raptori'll wait
21:03:38watusimotoif so, just write it down in a text that you can paste later
21:03:39raptori have another solution i just found
21:03:40watusimotoeasy!
21:03:45raptorok :)
21:03:50watusimotopretend you are live chatting!
21:03:53watusimotobye!
21:03:57raptorbye!
21:08:07watusimoto Quit (Ping timeout: 245 seconds)
21:23:33Watusimoto_ Quit (Ping timeout: 255 seconds)
21:27:00Watusimoto has joined
21:28:32Watusimotohi, back, you can paste your text!
21:28:37raptoroh hello
21:28:42raptorback so soon...
21:29:05Watusimotospeedy gonzales!
21:29:16raptorok, so for solution #3 above (not creating static archives and just linking in the object files in the end)
21:29:40raptorsince cmake 2.8.8, there is the option to create an OBJECT library
21:30:11raptorwhat this is is basically just a collection of object files not linked together, but doesn't need to be recompiled again for each project that contains it
21:31:04raptorso the OBJECT library is basically just a list of object files appended in the final link of a binary, instead of linking to a static archive like .a or a .lib
21:31:56raptorthis seems perfectly ideal for what we want to do, the only problem is that cmake 2.8.8 isn't found on a lot of older linux distros
21:32:29raptorit also means we don't need to have the added complexity for solution #2 with the extra linker flags
21:33:06raptorI hope you've been able to follow my explanations of the problem and solutions OK
21:35:32Watusimotoyes
21:35:34WatusimotoI have
21:35:46WatusimotoI was just trying to explain them to my younger son, but I have given up
21:36:16fordcars has joined
21:36:29Watusimotoso option 4 would be the clear winner if we could use cmake 2.8.8
21:36:35raptoroh yes
21:36:38Watusimotois that right?
21:36:44raptorand we can on windowx and osx
21:36:45Watusimotobut we can't depend on 2.8.8
21:36:50raptorwell...
21:37:12raptorlet me see when that was released
21:37:14Watusimotois the object library a .o or a .so file?
21:37:33raptorit's a list of .o files or .obj (windows)
21:37:42Watusimotoyes, ok
21:37:46raptorso the collection isn't actually linked together
21:38:08raptorcmake defers it until the end, but still let's all the projects know that it was already compiled.. so no duplicate compilation
21:38:33Watusimotowe need a python version of cmake we can bundle with our code
21:39:36raptorcmake 3.0.0 was just released
21:40:00raptorwe already constrain to cmake 2.8
21:40:11Watusimotoso why not 2.8.8?
21:40:26raptorcentos 6 has cmake 2.8.11
21:40:43raptorjust a minor worry that some distros use older cmake
21:40:56raptorand i need to prove to myself that it is a stupid worry
21:41:26raptor2.8.8 was released 18 Apr 2012
21:41:30Watusimotohave you proved it?
21:41:55Watusimotoit would be cool to have a build system written in python; no version worries there!
21:41:57raptorthe bigger question is: will we alienate build bitfighter on older systems by adding this requirement?
21:42:10fordcarsHi
21:42:15raptorhave you heard of scons?
21:42:28Watusimotoapt-get cmake 2.8.9 or whatever
21:42:36Watusimotohi fordcars
21:42:39Watusimotono scons
21:43:20raptorfordcars: what version of cmake is on RPi?
21:43:29raptorscons would fry your cpu
21:43:36fordcarsUh, well now I am on Arch, but I'll check
21:44:31fordcarsBtw building Bitfighter on ArchLinux RPi crashes
21:44:40raptorfordcars: is that the latest?
21:44:43fordcarsIt runs out of ram with the first file
21:44:47fordcarsUhm, 019c
21:44:47raptorbecause if so, yes it will crash
21:44:55fordcarsThe latest release
21:44:55raptorah
21:45:10fordcarsbanList.cpp just takes forever, then runs out of ram
21:45:23raptorok, that's probably because it's using a newer gcc which consumes RAM quite a bit more
21:45:30fordcarsPossibly
21:45:37raptoryou can limit g++, but I forget hwo
21:45:39raptor*how
21:46:09fordcarsI have cmake 3.0.0 here
21:46:20raptoroh wow, you're bleeding edge
21:46:32fordcars:O
21:46:36fordcarsIt worked great though
21:46:41fordcarsI guess :)
21:46:46CrazyLinuxNerd Quit (Ping timeout: 240 seconds)
21:47:37WatusimotoI would vote for whatever is the least painful for us; easy install on centos 6 is a bonus
21:47:48Watusimotoit sounds like 2.8.8 is a reasonable requirement
21:48:33WatusimotoI got to run again... back later
21:48:44raptorlater
21:53:05Watusimoto Quit (Ping timeout: 264 seconds)
21:56:42sam686in linux ping, try: ping -i 5 -a www.bitfighter.org it will beep, ding or some kind of alert when pingable
22:02:22fordcarsvirtual memory exhausted: Cannot allocate memory
22:02:59raptorsee here: http://stackoverflow.com/questions/16371939/reduce-memory-usage-in-boost-build
22:03:24fordcarsOh ahah
22:03:29fordcarswait, I don't have a swap
22:03:41raptorit's a problem with newer gcc + boost on RAM limited systems
22:03:55fordcarsI actually did that
22:04:08fordcarsI added CXXFLAGS="$(CXXFLAGS) --param ggc-min-expand=0 --param ggc-min-heapsize=8192" to the MakeFIle
22:04:35fordcarsBut I don't know if it worked
22:04:42fordcarsI guess it did
22:05:09fordcarsOohhh I do have a swap, it's just small okooko
22:58:08Watusimoto has joined
23:01:57raptorah, now I understand sam686's recent commits - they are fixes because of the PIC/static library issues in our current build system
23:03:13raptoralso on Windows, it looks like we can't use solution #2 so we have to either use cmake 2.8.8 or go back to duplicate compiling of all the object files for each project
23:04:03raptorlooks like debian 6 has cmake 2.8.2
23:04:25raptorand centos has 2.6 by default, but we install 2.8.11 via the EPEL source
23:05:24raptoroh, 2.8.9 is available in debian 6 -backports
23:05:27raptorok that's good
23:18:24Watusimoto Quit (Ping timeout: 264 seconds)

Index Search ←Prev date Next date→

These logs were automatically created by BFLogBot on irc.freenode.net.